Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The relative effect of clipping, neighbours, and fertilization on the population dynamics of Lupinus… Graham, Stepanie Ann 1994

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

Item Metadata

Download

Media
831-ubc_1994-0428.pdf [ 3.39MB ]
Metadata
JSON: 831-1.0087483.json
JSON-LD: 831-1.0087483-ld.json
RDF/XML (Pretty): 831-1.0087483-rdf.xml
RDF/JSON: 831-1.0087483-rdf.json
Turtle: 831-1.0087483-turtle.txt
N-Triples: 831-1.0087483-rdf-ntriples.txt
Original Record: 831-1.0087483-source.json
Full Text
831-1.0087483-fulltext.txt
Citation
831-1.0087483.ris

Full Text

THE RELATIVE EFFECT OF CLIPPING, NEIGHBOURS, AND FERTILIZATION ON THEPOPULATION DYNAMICS OF Lupinus arcticus (Family Fabaceae).bySTEPHANIE ANN GRAHAMB.Sc., The University of Calgary, 1990A THESIS SUBMITTED TN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of BotanyWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAJune 1994© Stephanie Ann Graham, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of Bntish Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)____________________________Department of /O74IV VThe University of British ColumbiaVancouver, CanadaDate vZj2 c,2, /‘)DE-6 (2188)ABSTRACTA demographic study was conducted in 1991 and 1992 on replicated field populations ofLupinus arcticus, near Kluane Lake, Yukon. The relative effects of herbivory, neighbours, andsoil fertility level were assessed using a factorial experiment of +1- clipping, +1- neighbourremoval, and +1- fertilizer (NPK). The main population experiment monitored the dynamics ofleaves, however, data on reproduction, survival, and size were also collected from the permanentquadrats. Clipping reduced leaf cohort survivorship, total leaf density, and the incidence ofdisease on leaves, but resulted in an increased standing crop of leaves. Removing neighboursincreased the percent cover of L. arcticus and decreased total leaf mortality. Fertilizing increasedthe incidence of disease on leaves, and reduce the standing crop of leaves. Significant three-wayinteractions between treatments affected the plasticity of petiole length distributions for L.arcticus. Between-year differences in the responses to the treatments were also detected,particularly for reproductive investment and output in L. arcticus.Although a number of significant responses to treatments were detected, nevertheless, theoverall tendency was for a lack of response, especially pertaining to leaf population dynamics.This low response to the treatments imposed is consistent with Grime’s (1979) arguments thatplants growing in low productivity, infrequently disturbed habitats (i.e. stressful sensu Grime1979) should show little response to short-term changes in local environmental conditions.iiTABLE OF CONTENTSPageABSTRACT iiTABLE OF CONTENTS iiiLIST OF FIGURES viLIST OFTABLES viiLIST OF ABBREVIATIONS xiACKNOWLEDGMENTS xiiDedication xiiiINTRODUCTION 1Background 3Thesis problem 4CHAPTER 1 LITERATURE REVIEW 7The effect of herbivory 7The effect of neighbours 14The effect of fertilizer 23CHAPTER 2 DEMOGRAPHY OF LUPINUS LEAVES 33METHODS 33Study Area 33Experimental Design 36Leaf Demography Study 38Statistical Analyses 38Standing crop available 40RESULTS 41Leaf Demography 1991 41Leaf Demography 1992 41iiiStanding crop available.50DISCUSSION 53Leaf demography 54Standing crop available 57CHAPTER 3 LEAF COHORT SURVIVORSHIP 59METHODS 59Statistical Analysis 60Parametric Model Fitting 60Weibull Model Description 61Nonparametric Statistics 61RESULTS 62Weibull Model Analyses 62Nonparametric Analyses 65DISCUSSION 67Effects of clipping 68Effect of fertilizer and removing neighbours 69Comparison of cohorts 70CHAPTER 4 VEGETATIVE GROWTH 74METHODS 74Size structure 74Cover 75RESULTS 75Size structure of petioles 1991 75Size structure of petioles 1992 78Changes in percent cover 81Changes in Vegetation Composition 82DISCUSSION.86Petiole length distribution.86Percent Cover 89CHAPTER 5 REPRODUCTION 94METHODS 95Statistical Analysis 95RESULTS 96Yearly variation 96Reproduction 1991 97Reproduction 1992 105DISCUSSION 111CHAPTER 6 SUMMARY AND CONCLUSIONS 115APPENDIX A 119Characteristics of the Weibull Distribution 119BIBLIOGRAPHY 121VLIST OF FIGURESFigure 1: Bud production I m 2 over time by treatment in 1992. 51Figure 2: Mean (±SEM) live summer leaf-days (C1) by treatment and year. 52Figure 3: Leaf survivorship curves for cohort 1 and cohort 2 by treatment in 641992.Figure 4: Mean (±SEM) change in percent cover for Lupinus arcticus and 85Festuca altaica from 1991 to 1992.FigureS: A comparison of total number of racemes, flowers, and pods 99produced across all treatments in 1991 and 1992.Figure 6: Snowshoe hare spring densities on CSP treatment grids from 1987- 991994. The control treatment shows the beginning of the crash in1991.Figure 7: Reproductive effort in 1991 and 1992 as measured by mean (±SEM) 101raceme production I m 2 for each treatment.Figure 8: Summary of mean (±SEM) number of flowers initiated / m2 and 102mean (±SEM) number of fruits matured (pods) / m2 by treatment.Figure 9: Survival curves generated by the Weibull distribution when 2=1 and 120o is allowed to vary. Redrawn from Lee (1992).viLIST OF TABLESTable 1: Summary of mean leaf density, bud density, leaf mortality and incidence 42of disease by treatment in 1991.Table 2: Source of variation for three-way ANOVA results of bud density, leaf 43density, incidence of mortality and disease in 1991.Table 3: Summary of mean leaf density, bud density, leaf mortality and incidence 44of disease in 1992.Table 4: Source of variation for three-wayANOVA of leaf density, incidence of 45disease and mortality in 1992.TableS: a) Source of variation for one-way analysis of leaf density and 46incidence of disease by treatment in 1992.b) Tukey comparison of one-way analysis of leaf density by treatment 46in 1992.c) Tukey comparison of one-way analysis of disease by treatment in 461992.Table 6: Source of variation for one-way repeated measures MANOVA of bud 47production in 1992.Table 7: Summary of multivariate test statistics for one-way and three way 47repeated measures analysis of bud production/m 2 in 1992.Table 8: Three-way repeated measures MANOVA of bud production/rn 2 48Table 9: Source of variation for the polynomial test of order (quadratic) for 49three-way repeated measures analysis of bud production in 1992.Table 10: Source of variation table for a 3-way ANOVA of log transformed live 52summer leaf days (Ci) by treatment in 1991 and 1992.Table 11: Scale (?) and shape (ce) parameter estimates for the Weibull 63Distribution (1992).viiTable 12: Combined source of variation for the one-way ANOVAs for cohort 1 63and cohort 2 comparing the log transformed (scale) Weibullparameter by treatment.Table 13: Combined source of variation table of one-way ANOVAs for cohorti 66and cohort 2 comparing the log transformed shape (x) parameter bytreatment.Table 14: Combined source of variation table for two-way ANOVAs testing for 66the effect of cohort within treatment on log transformed shape and scaleparameters.Table 15: The nonparametric log rank statistics testing for between treatment 66effects in cohort 1 and cohort 2 leaf survivorship curves.Table 16: Chi square test for independence of fertilizer, clipping and neighbour 77removal on petiole length distribution using log linear analysis.Table 17: Length distribution of petioles by treatment reported as frequency and 77percent occurring in each class.a) Survey5, 1991.b)Survey7, 1991. 78Table 18: Chi square test for independence of fertilizer, clipping and neighbour 79removal on petiole size distribution using log linear analysis.Table 19: Distribution of petiole lengths by treatment reported as size class 80frequency and percent occurring in each class.a) Survey 4, 1992.b) Survey 6, 1992. 81Table 20: Summary of mean changes (± SEM) by treatment of percent cover of 83species occurring in quadrats from August 1, 1991 to July 21, 1992.Table 21: Summary of ANOVA results for arcsine transformed changes in percent 84cover in Lupinus arcticus.Table 22: a) Univariate source of variation comparing total raceme, flower bud 100and pod production I m 2 in by treatment 1991 and 1992.vii ±b) Multivariate test statistics comparing reproductive variables by 100year.Table 23: a) Univariate source of variation comparing total raceme, flower bud 103and pod production/rn2 by treatment in 1991 and 1992.b) Wilks’ Lambda and F statistics from MANOVA of 1991 estimates of 103raceme and flower bud production.Table 24: a) Three-way ANOVA of square root transformed pod production per 104m2in 1991.b) Summary of Kruskal-Wallis non-parametric test results for some 104reproduction variables.Table 25: Three way ANOVA test of arcsine transformed reproductive efficiency 105/m2 in 1991.Table 26: a) Univariate statistics by treatment of transformed raceme production 107and flower bud initiation in 1992.b) Wilks’ Lambda and F statistics from MANOVA of 1992 estimates of 108log transformed raceme and flower bud production in 1992.Table 27: a) Three-way ANOVA of log transformed pod production / m2 in 1081992.b) One-way ANOVA of pod production / m 2 in 1992 to determine the 108effect of the intensified clipping treatment on reproductive output.Table 28: a) Three-way ANOVA of arcsine transformed relative success in 1992. 109b) One-way ANOVA of arcsine transformed relative success in 1992. 109c) Post hoc Tukey grouping of relative success by treatment in 1992. 109Table 29: Mean size (±SEM) of reproductive structures by treatment for 1992. 109Table 30: a) Univariate (three-way) test statistics of length of racemes, and length 110of peduncles in 1992.b) Multivariate test statistics from three-way MANOVA of length of 110racemes, and length of peduncles in 1992.ixc) Univariate (one-way) test statistics of length of racemes and length 110of peduncles by treatment in 1992.d) Multivariate test statistics from one-way MANOVA of length of 111racemes and length of peduncles by treatment in 1992.Table 31: Summary of significant main effects detected in this experiment for 1151991 and 1992.xLIST OF ABBREVIATIONSTo save space, the following abbreviations are used in all tables or figures unless otherwisespecified in the table or figure legend.TREATMENT TABLES FIGURESCONTROL CONT CONTFERTILIZER FERT FCLIPPING CLIP CNEIGHBOURS REMOVED REM NRFERTILIZED/CLIPPED PERT/CLIP F-CFERTILIZED/NEIGHBOURS PERT/REM F-NRREMOVEDCLIPPED/NEIGHBOURS CLIP/REM C-NRREMOVEDFERTILIZED/CLIPPED! PERT/CLIP/REM F-C-NRNEIGHBOURS REMOVEDEXTRA CLIPPING EXTRA CLIP or CLIP+STANDARD ERROR ±SEM ±SEMSHAPE PARAMEThRSCALE PARAMETERxiACKNOWLEDGMENTSI wish to extend my sincere appreciation to Dr. Roy Turldngton for his friendship, constantencouragement and support, both emotional and financial, during the supervision of this thesis.Thank you for giving me the opportunity to experience the Yukon as a field botanist. It was theopportunity of a lifetime that few will ever experience, and I will never forget. I am also verygrateful for his diligence as an editor, constructive criticism, and amazing ability to pull muchneeded references out of the air at the appropriate moment. Thanks, boss. I would also like tothank Dr. Elizabeth John, Dr. F.E.R. MacCauley, Dr. Jack Maze, Dr. Gary Bradfield, and Dr.David Hik for their helpful suggestions, observations, theoretical and statistical assistance duringthe course of this project.I am very grateful for the dedication and perseverance of many field research assistants andfellow graduate students who endured monster mosquitoes, black ffies, and the risk of bear attackwhile remembering how many hundred lupines they had just counted. They include Elena Klein,Franklin Dlott, Brian Leung, Sue Crites, Shane Kirby, and Meagan Williams.A very special thanks must go to Mr. Ji Yong Yang, an outstanding research associate andtrue friend, who both inspired and endured a great deal throughout this thesis. Your friendship,patience, unusual dreams, frequent bouts of insanity, romantic woes, and cannibalistic cravingswere refreshing. I am grateful beyond words, Ji. You are one of a kind and I am glad to call youfriend.I wish to acknowledge the technical assistance of the Collaborative Special Project, headedby Dr. Charles Krebs, who provided field assistants, equipment, and vehicles, when required. Iwould also like to thank the Arctic Institute of North America, and Andrew and Carol Williamsfor making my stay in the Yukon truly memorable. I must also thank the senior CSP technicians(Mark O’Donoghue, Sabine Schweiger, Cathy and Frank Doyle), for their patience and good-natured tolerance of a bunch of lazy botanists. These people must be credited with transforming abunch of ‘green’ students (myself included) into seasoned field ecologists in a few short weeksevery year. You have done an excellent job. I am also grateful to frene Wingate, who hasfrequently come to the rescue with equipment, and information on very short notice.I cannot say enough about the role my family has played in my education, but this thesiswould not have been possible without their love, and life-long support. To my mother, AudreyGraham, who has been a wonderful role model, and has always encouraged me work hard toachieve my best, you are the greatest! To my father and stepmother, Barry and Dar Graham, whotaught me to pursue my dreams and let the sky be my limit, they have always been there when Imost needed them. Together, they have all instilled in me inner strength, and an appreciation forthe value of education. Thank you all, I could not have done this without you.xiiDedicationFinally, I have a special thanks for Shane Kirby, whose contribution to the completion ofthis thesis cannot be easily stated. Over this long, and sometimes rough road, we have climbedmountains, stumbled down unstable rockslides, passed through many valleys, listened to the rain,and drank orange juice in the sunshine. Many moments. Through it all, you encouraged me topersevere, even when obstacles stood in the way. Together, we have overcome and grown. Thisthesis is dedicated to you, with friendship, respect, and much love.xiilINTRODUCTIONA number of abiotic and biotic factors are known to influence plants on an individual,population, and a community scale. A viable seed may be prevented from germinating by theplacement of a stone or leaf that blocks its access to sunlight or moisture, whereas such factorsmay have little impact on a mature plant. Similarly, a seedling, by virtue of its size, may beundetectable by herbivores or insensitive to wind yet these are important mortality factors on full-sized plants. This is significant because the relative importance of different factors on thedynamics of individual plants and populations change over time, over stage, over a season. Theeffects of these factors may be localized or patchy within a habitat. Although a myriad of factorsinfluence plant dynamics, few are capable of limiting or regulating plant populations on their own.Rarely, if ever, are plants in natural environments influenced by one factor alone to the exclusionof all others. A challenge for ecologists is to untangle and measure the relative strengths ofdifferent factors and determine how they interact to produce observed patterns of abundance(Hunter & Price 1993). Experiments that manipulate a number of potential limiting factorsconcurrently and measure the response of individual plants and populations permit a betterunderstanding of how different limiting or regulatory agents interact in natural communities andgive insight at the individual, population, and community levels. The most effective way toachieve this is to conduct replicated multi-factorial field experiments.Historically, three factors have been investigated as important agents determining plantperformance and population dynamics: disturbance (including herbivory), competition, andnutrient availability. Larger disturbances, are not considered here because they frequently operate12on a scale and an intensity that eliminates entire communities or populations as opposed tooperating within an existing community. Another potentially important limiting agent that is notconsidered here is the effect of the soil microbial community. At present, the implications of themicrobial community as a limiting agent in plant dynamics can only be inferred from a few studiesalthough this will likely change in the future as more data are collected.Much of the previous research concentrates on how these agents interact to structurecommunities. This focus on community level responses to these factors has fueled the debateabout top-down or bottom-up regulation of communities. Depending on the community beingstudied, different researchers have come to the conclusion that either top-down forces (herbivoryor predation), or bottom-up forces (nutrient availability and primary production) are moreimportant in determining structure of that particular community. A recent study (Hunter & Price1993), has tried to synthesize the two opposing views with a more realistic model incorporatingboth bottom-up and top-down forces. This model proposed a compromise such that the relativeroles of top-down or bottom-up forces in a community vary within and among the systems inquestion. The Hunter & Price model proposes that bottom-up abiotic factors such as nutrientstatus set absolute limits upon which organisms can exist in a certain environment. However, thisbottom-up component of the model is also subject to top-down forces generated by consumers inthe community. Variability or heterogeneity in the response of consumers in terms of theirinteractions with other members of the community (predators, competitors, or prey) act as topdown feedback mechanisms that act in conjunction with bottom-up forces. The concession thismodel makes is that if either top-down and bottom-up forces can operate and prevail in acommunity depending on local conditions, any species in the food web (consumer or producer)can act as a keystone species depending on the outcome of the interactions both between andamong levels of the food web. This model is flexible enough to support both top-down and3bottom-up regulation of communities on the basis of interactions between and among differentlevels in the food web as well as due to environmental variability.The Hunter & Price’s model necessitates more experiments to investigate the effect ofdifferent limiting or regulatory agents and especially their interactions in communities. However,to interpret the response of communities to these factors, it is necessary that we first understandhow these factors operate at the level of the population and the individual. Few studies examinethe effect of these factors and their interactions on plant population dynamics, and those that doare commonly performed in pot or common garden experiments. However, the results of suchexperiments can rarely be extrapolated to natural communities and thus the need for manipulativefield studies is evident.BAC KGROUNDAs part of an NSERC Collaborative Special Project (CSP), community dynamics in a borealforest ecosystem are being investigated with emphasis on testing the relative importance ofvarious interactions between top predators, herbivores and the plant community on the structureof the community (Krebs 1988). Specifically, Krebs and co-workers are investigating the linkagesbetween the various trophic levels in the forest ecosystem with particular emphasis on thevertebrate food web. One major aim of the study is to test a number of hypotheses to explain the10-year cycle of snowshoe hare (Lepus americanus) population densities that have been observedacross northern Canada and Alaska. In this work, top-down (predators, herbivores), and bottomup (nutrients) factors are being manipulated at an ecosystem scale with experimental grids as largeas 1 km 2 The vegetation component of this study is concerned with dynamics of the understoryherbaceous vegetation as a potential summer food resource for hares and other herbivores. Twohypotheses are being tested regarding dynamics in the plant community:i) Vegetation amount and composition is regulated by nutrient availability alone.4ii) Vegetation amount and composition is regulated by nutrient availability and herbivory.To test these hypotheses, experiments on the understory herb community are being conductedthat focus on fine-scale analysis of changes in composition, and more recently abundance(biomass), under treatments manipulating soil nutrient availability (fertilizer) and herbivore density(fencing). In spite of its importance in structuring some plant communities, there are currently nostudies investigating the role of competition on the composition and structure of the plantcommunity.THESIS PROBLEMAlthough it is generally acknowledged that there is a need to understand how differentfactors interact to influence the structure and dynamics of plant communities, there is also a needto understand how these factors interact at the scale of individuals and populations whichcomprise the communities. Understanding how populations are limited and regulated in naturalcommunities is important in the formulation of conservation strategies that are becoming anincreasing part of ecological research world-wide. Surprisingly, few studies are available thatexamine the effect of different limiting factors and their interactions on the dynamics of naturalpopulations. If we are to gain a better understanding of how populations are regulated, it isimperative that studies are performed on natural populations exposed to the full range ofenvironmental forces and heterogeneity.In this study, three factors commonly influencing the growth of plants were chosen forinvestigation: herbivory, interference from neighbours, and soil nutrient availability. The specieschosen for study was the Arctic Lupine, (Lupinus arcticus Lindi.: Family Fabaceae), a relativelyabundant understory herb in the boreal forest. The effect of these three factors on the populationdynamics of L. arcticus will be determined by assessing the relative impacts of different5combinations of clipping, neighbour removal, and fertilizer in a factorial design. The reasons forchoosing this particular species are as follows:1. Lupinus arcticus occurs in high enough abundance to obtain adequate densities forpopulation experiments.2. Lupinus arcticus is known to be a summer food resource for many herbivores in the borealforest including snowshoe hares and ground squirrels.3. Members of the Genus Lupinus are nitrogen fixers and are also capable of producing anti-herbivore defense chemicals. This may result in unique responses to the different limitingfactors being tested, particularly nutrient availability and herbivory.Ideally this study would have preferred to examine the effects of natural levels and types ofherbivory on plant population dynamics. However, the population density of snowshoe hare, theprimary herbivore in the study area (near Kluane Lake, Yukon), was in the decline phase of itscycle in 1990/91. As a result, herbivore densities were not sufficient to measure the effect ofnatural grazing on L. arcticus populations. Therefore, herbivory was simulated by clipping to aconstant height to control the amount and type of herbivory experienced by the replicatepopulations.The objectives of this thesis are to determine the direct and interactive effects of clipping,neighbour removal, and fertilization on population dynamics of modules of L. arcticus, as well asdetermine their relative importance in limiting the growth of L. arcticus.The experimental component of this thesis is divided into four sections. The first iscomposed of leaf demography studies that comprise the main body of data. This experiment wasperformed for two reasons. First, L. arcticus grows clonally by sending out undergroundrhizomes. Because of this, it was not possible to identify genets or ramets for a populationexperiment. Second, in both 1991 and 1992 there was insufficient recruitment from seed to6monitor the dynamics of new recruits (genets). Therefore population dynamics in L. arcticuswere investigated by counting leaves rather than ramets. The second section presents the resultsof a leaf cohort survivorship study done in 1992. Section three summarizes the growth datacollected as part of the leaf demography study (section 1). The final section analyzesreproductive data also collected as part of the main experiment (section 1). Because of thedetailed methods and statistical analyses associated with the different components of this study,each of the four sections is structured to include its own methods, results, and discussion section.The final chapter attempts to synthesize the results into a more cohesive framework.CHAPTER 1LITERATURE REVIEWThis review is divided into three parts that examine ecological studies pertaining to theeffects of the major factors being investigated in this study: herbivory, neighbours, and soilfertility. The body of literature available for each of these factors is very extensive and cannot befully addressed in one chapter, therefore this review concentrates primarily on terrestrial fieldstudies that manipulate herbivory by simulation or herbivore exciosure; neighbour density byremoval of species; and soil fertility through fertilization. Although the focus is on fieldpopulation studies, other studies (pot, common garden) are incorporated in some cases if theymanipulate similar treatments, or if they demonstrate potential responses not yet reported for fieldpopulations. Because the response of populations to manipulations in the field is influenced byboth individual plants as well as to community level effects, some studies are included particularlyif they show pertinent results that may explain some population level responses observed in thisstudy.THE EFFECT OF HERBIVORYHerbivores have a broad range of effects on plants originating from both the consumptionof plant material, as well as through activities not associated with direct removal of plant biomass.Direct effects incurred as a consequence of herbivores feeding include: (i) reductions in fitness,(ii) subordination of position in the canopy, as well as changes in (iii) seed number and size, (iv)growth rate, and plant architecture, (v) root growth, (vi) flowering time, (vii) the outcome ofcompetition in mixture, (viii) rates and patterns of dispersal, (ix) the rate of population increase,and (xi) the amount of genetic variation within a population, and (xii) the stimulation of anti78herbivore defense compounds. Indirect effects that are the product of plant- herbivorecoexistence have also been shown to affect plant dynamics. Some of these include trampling,deposition of faeces, urine, or saliva, transmission of disease, pollen and seed dispersal, and plant-animal mutualisms (e.g. ant-acacia, Crawley 1983). The combination of these direct and indirecteffects may have positive or negative repercussions for the plant and the dynamics of plantpopulations. As a result, the question of herbivory as a beneficial process to plants has beenraised by some investigators (see Owen 1980, Beisky 1986). Although herbivores have the abilityto heavily damage plants, there is still some debate in the literature on the relative importance ofherbivory. Belsky (1987), in a review of the effects of grazing at organismic, community, andecosystem scales, suggested that there is evidence that low to moderate levels of herbivory haveno measurable effect on plant dynamics. Crawley (1988) further submits that there is littleevidence in support of herbivores as a regulatory agent in spite of their ability to affect plantgrowth in such major ways.Herbivore exclusion studiesPrevious studies designed to measure the effect of herbivores on plant dynamics in the fieldhave commonly used two methods: (i) herbivore exclusion with fencing, pesticides, insect traps,and (ii) simulated herbivory. Studies that use herbivore exclusion usually compare plant dynamicsin an area that is inaccessible to herbivores for a period of time relative to a control area whereherbivores are permitted unrestricted access (Tansley & Adamson 1925, Andersson & Jonasson1986, Coppock et al. 1983, Pyke 1986, Crawley 1990). The advantage of this method is that it isthe only one that allows the researcher to measure the effect of natural levels of herbivory (i.e. interms of intensity, frequency, and type of damage) on plant dynamics. The drawbacks are that theamount of herbivory among replicates is not controllable. Some experimental plots that areaccessible to herbivores are not necessarily grazed, or they may not all be grazed with the sameintensity, or for the same duration. Different species of herbivores may graze replicate plots9differently. This can result in a great deal of variability in the measured response within atreatment, and it may make it difficult to quantify and discern the effects of herbivory from otherfactors affecting plants. This could be particularly important when herbivore densities are low,and the impact of grazing on plants is patchy or infrequent. The level of replication required tocompensate for this problem may be more than is manageable in an average study with limitedresources.Pacala & Crawley (1992) in a review of the effects of herbivores on plant diversitysuggested that previous experiments removing herbivores have had little effect on plantcommunities. As a result, they suggested that a more appropriate means of studying herbivorywould be to combine herbivore exclusion or removal with some form of experimentalperturbation. This has frequently involved manipulating the competitive environment in which theexperimental defoliation occurs.Grazing simulation studiesIn simulation studies, the effects of herbivory on plants is determined by controlled tissueremoval experiments where investigators decide the amount and type of herbivory to be imposedon a plant population or community. Baldwin (1990) reviewed this method and presented anumber of inadequacies associated with this procedure. He suggested that while simulatedherbivory may adequately mimic the effect of biomass loss on plants, it is not possible forresearchers to simulate the myriad of other effects that herbivores have such as their selectivity forcertain tissues and patterns of damage, the effect of saliva, timing of attack, stimulation ofchemical response in plants, or the exact nature of the mechanical damage (tearing, suckingchewing, stripping). Several studies were cited documenting the difference in response of plantsto mechanically damaged plants relative to those subject to real herbivory. In Baldwin (1988), acomparison of Nicotiana sylvestris’s response to real and simulated herbivory reported that thealkaloid response to the simulation was significantly different than that observed for actual10herbivore attack. Mauricio et al. (1993) also reported that the pattern of leaf damage(concentrated vs. dispersed damage) in an annual, Raphanus sativus L., had different impacts onreproduction and biomass. Therefore, it is evident that simulation studies must be undertakenwith caution, with special attention paid to the nature of the question being addressed. It isobvious that studies of the indirect effects of herbivores are not adequately simulated with thistechnique (Baldwin 1990). In spite of this, simulated herbivory experiments have someadvantages over exclusion experiments in that the magnitude, and the type of damage inflicted canbe controlled and measured. Pathogen spread that may confound the results due to alterations ofhost-plant resistance is minimized, and simulated herbivory can distinguish between theuncontrolled effects of non-random selection of tissue and plants (Baldwin 1990).Community studiesOne of the earliest experiments that manipulated the abundance of herbivores and measuredthe response in a plant community was a study by Tansley & Adamson (1925) on the effect ofrabbit grazing on British chalk grassland. Within six years, the exclusion of rabbits resulted in theelimination of a number of plant species to produce a community dominated by a single grassspecies. This new community was itself eventually replaced by a shrub community (Harper 1977).Crawley (1990) in a similar study excluded rabbits from acid grassland in Britain, and reportedthat palatable grasses increased in abundance when protected from grazing. Species richness wasnot affected in this experiment, however this study was only maintained for three years. Areduction in species richness may have occurred if the experiment was maintained longer. Othercommunity studies that have manipulated rodent densities and examined the effect on plantcommunities have reported changes in species diversity, and increased plant biomass in absence ofherbivores (Coppock et al. 1983, Andersson & Jonasson 1986).Studies have also examined the effect of the removal of non-rodent herbivores on plantcommunities. Bakker & Ruyter (1981) found that re-introduction of grazers (cattle) to a salt-11marsh changed patterns of succession in the plant community. Areas remaining ungrazed showedprogressive succession (increasing species richness and complexity of structure) and canopyclosure with an increase in herb biomass and litter cover, whereas areas that were newly grazedoccasionally showed retrogressive succession (decreasing species richness and structuralcomplexity) to a more open canopy. Bazely & Jefferies (1990) excluded Lesser Snow Geesefrom a salt-marsh in LaPerouse Bay, Manitoba. They reported an increase in species diversity,changes in species composition and increased standing crop inside herbivore exciosures.Population experimentsMany studies have examined the effects of herbivory from a population perspective. Anexperiment that excluded rabbits in the Breckland showed a change in the population agestructure of Hieraceum pilosella L. as a result of increased plant lifespan (Davy & Bishop 1984).This same study also reported a reduction in the rate of flowering when rabbits were excludedfrom the area.Kotanen & Jefferies (1989) excluded Lesser Snow Geese from a tidal flat on Hudson’s Bayand reported that release from grazing decreased leaf production and turnover, and increased leaflifespan of Carex Xflavicans. Based on these results they also suggested that C. Xflavicans hasthe ability to modify its leaf demography in response to grazing. Jonsdottir (1991) compared thepopulation dynamics of Carex bigelowii at grazed and ungrazed sites along an altitudinal gradientin the Icelandic Highlands. This study reported that the effect of sheep grazing varied withaltitude, and that grazing resulted in an increased tiller density and reduced the survivorship,turnover time, rate of flowering, and age of the population.The response of plant populations to herbivore exclusion is generally variable. Somesuggest that this is due to past grazing history (Detling & Painter 1983, Jaramillo & Detling1988), while other studies have shown that the response to herbivore exclusion depends onextrinsic factors such as soil nutrients and faecal deposition (Berendse 1985; Kotanen & Jefferies121987; Polley & Detling 1989), changes in the light or competitive environment (Polley & Detling1989; Reader 1992 ), or intrinsic factors such as below ground reserves (root storage), andgrowth substances (Kotanen & Jefferies 1987). Some studies have also shown that the impact ofdefoliation does not always depend on the competitive environment, and the loss of position in thecompetitive hierarchy due to herbivory does not always occur (Fowler & Rausher 1985).Similarly, plants with defoliated neighbours do not always show competitive release (Lee &Bazzaz 1980). Crawley (1990) and Pacala & Crawley (1992) have also suggested that the effectof release from grazing within a community will be determined by the interaction between aspecies’ palatability and its competitive ability.Grazing pressure has been shown to affect plant populations in other ways. Comparisons ofgrazed and ungrazed populations of some species have shown variation in the incidence of diseasedepending on the amount of herbivory experienced within a population. In some cases grazingreduced the level of infection, while in others it increased the incidence of disease found within apopulation (Bradshaw 1959; Clay 1988; Wennstron & Ericson 1991).Reproduction in plants is frequently altered by herbivory, however, the effects of herbivoryon all aspects of reproduction depends on many factors including the timing, intensity andfrequency of the disturbance, the type of tissue removed (vegetative vs. reproductive), and thenature of the damage (defoliation, frugivory, sap-sucking, etc.) (Crawley 1983). The effects ofdamage or removal of reproductive tissue on reproductive output is usually negative, but notexclusively so. There have been cases reported where herbivore damage (simulated or natural)does not significantly reduce a plant’s reproductive output. In some cases, these species arebelieved to be selected for over-initiation of reproductive structures beyond what a plant isnormally capable of maturing (Stephenson 1980; Lee & Bazzaz 1982). Several reasons have beenproposed to explain this apparent over-initiation of reproductive structures. Lee & Bazzaz (1982)suggest that an over-initiation of reproductive structures may be an attribute of certain plants thathave been selected to persist in environments where there is a high degree of resource13unpredictability and an abundance of fruit predators. In good years (high resource availability), allfruit could be filled, whereas during unfavourable periods, structures lost to frugivores could bereplaced. Alternatively, if all flowers weren’t fully pollinated, a plant could mature those fruit thathad been fertilized to maximize its reproductive output (Lee & Bazzaz 1982).Herbivore exclusionA brief summary of some of the effects reported for herbivore exclusion experiments are:increased reproduction (Louda 1984; Andersson & Jonasson 1986; Pyke 1986; Jonsdottir 1991;Karban & Strauss 1993); increased recruitment (Swank & Oechel 1991; Karban & Strauss 1993);increased leaf or tiller survivorship (Rausher & Feeny 1980; Louda 1984; Kotanen & Jefferies1987, 1989; Jonsdottir 1991; Reader 1992); increased biomass and growth (Rausher & Feeny1980; Jaramillo & Detling 1988; Jonsdottir 1991; Fox & Morrow 1992); an increase in tissuenitrogen or phosphate (Jaramillo & Detling 1988; Fox & Morrow 1992); decreased initiation ofnew leaves or turnover of leaves (Louda 1984; Kotanen & Jefferies 1987, 1989); and increasedplant height (Detling & Painter 1983; Louda 1984).ClippingThe response of plants to clipping has also been extensively studied. Johnson et al. (1987)in a study of the effect of defoliation (clipping), and nutrients on the growth, and alkaloidproduction of Lupinus succulentus found that defoliation reduced tissue biomass, total nitrogencontent and alkaloid content. The effect of nutrients in this study are discussed later. Otherstudies have also reported significant changes in biomass, total nitrogen, and/or alkaloid content inresponse to simulated herbivory (Stephenson 1980, 1984; Ruess et al. 1983; Fowler & Rausher1985; Polley & Detling 1989; Baldwin 1988; Abul Fatih & Bazzaz 1984; Holland et al. 1992;Doak 1992). Other commonly seen responses to defoliation are increased leaf production (AbulFatih & Bazzaz 1984; Polley & Detling 1989; Ruess et al. 1983), increased leaf and seedling14mortality (Abul Fatih & Bazzaz 1984; Kotanen & Jefferies 1987), reduced levels of reproduction(Abul Fatih & Bazzaz 1984; Doak 1992), and reduced plant height (Lee & Bazzaz 1980). Inother cases, clipping was not shown to significantly reduce reproduction (Lee & Bazzaz 1980;Ruess et al. 1983; Fowler & Rausher 1985). In some cases, the lack of effect on reproductioncould be attributed to the short duration of the experiment. Frequently, the effects of clipping arenot detected until the following growing season (Pyke 1986; Karban & Strauss 1993).Herbivory is a complicated process and the ability of plants to respond to changes in theintensity of herbivory they experience is a function of both physiological and environmentalfactors. Jaramillo & Detling (1988) list some of the pertinent physiological changes a plant canexperience when levels of herbivory change. These include: (i) changes in photosynthetic ratesand balance of assimilates, (ii) changes in nutrient allocation patterns, (iii) differential balance invegetative vs. reproductive tissues, and (iv) changes in nutrient uptake and hormonal balance.Because the range of responses to herbivory is so broad, it is difficult to predict with certainty itseffect on plants, populations or communities. As a result, future studies will likely concentrate ondetermining how herbivory in the field interacts with other ecological forces operating in theenvironment.THE EFFECT OF NEIGHBOURSThe presence of neighbours is an important factor influencing plants at all levels oforganization from the individual to the community. The components and structure of theneighbourhood that a plant inhabits can determine its access to resources, its susceptibility toherbivory and disease, exposure to harsh environmental conditions, and even the nature of thelocal disturbance regime (fire, wind, water). Because plants are sessile, the presence ofneighbours is presumed to have primarily negative consequences on plant performance as a resultof neighbours competing to acquire the resources necessary for growth. However, neighboureffects extend beyond competition for resources. Interference among neighbours can occur as a15result of non-competitive interactions (Harper 1977). For instance, Reader (1992) demonstratedthat neighbouring plants provided the habitat required to maintain herbivore population densities,thus increasing risk of attack on a target species. Christie et al. (1978) also reported that theabundance of soil microbes in a plant’s rhizosphere varied according to the presence and identityof neighbouring plant species. The consequences of neighbours influencing a plant’s rhizospherecould be either positive or negative depending on the nature of the relationship between the plantand soil microbes (i.e. pest or symbiont).Neighbours have been shown to be beneficial to plants (reviewed in Hunter & Aarssen1988), in some cases by providing shelter from harsh environmental conditions in severe climates(reviewed in Callaghan & Emanuelsson 1985; Callaghan 1987), in others by improving localconditions by increasing moisture and nutrients (i.e. nitrogen fixation) (Bliss & Svoboda 1984).Another study has also reported that close proximity to an unpalatable neighbour reduced the riskof herbivory for some species (Holmes & Jepson-Innes 1989). Similar results have been reportedsuch that the risk of herbivory was reduced when plants were concealed or shaded by neighbours(Huffaker & Kennet 1959; Rausher 1981). Consequently, although competition is an importantcomponent of neighbour interactions, the presence of neighbours in a community can be shown toaffect plants and populations in ways not related to competition for limited resources.This review demonstrates that the effect of neighbours on plants and populations arenumerous, widespread and varied. Another good review on the response of natural and semi-natural vegetation to different types neighbour manipulations in the field is provided in Aarssen &Epp (1990). A brief summary of some commonly observed effects attributed to the presence orabsence of neighbours include changes in: (i) plant biomass (Mack & Harper 1977; Campbell &Grime 1992; Rees & Brown 1992), (ii) patterns of survivorship and establishment of seedlings(Gurevitch 1986), (iii) branching patterns or tillering rates (Lee & Bazzaz 1982; Fetcher 1985),(iv) individual reproductive output (Paimblad 1968; Lee & Bazzaz 1982; Campbell & Grime1992), (v) population size hierarchy (Weiner 1985), (vi) rhizome length and growth patterns16(Schmid & Bazzaz 1992), (vii) competitive exclusion from optimal habitats (Gurevitch 1986),(viii) physiology (Wilson 1989), and (ix) root: shoot ratio (Gurevitch et al. 1990).Although interference from neighbours (intra- or interspecific) can regulate or limit thegrowth of plants and populations, the ability of neighbour interactions to determine the structureof plant communities is often uncertain and difficult to measure (Wailer 1981). The evidenceavailable indicates that the intensity of interference in plant populations on both a temporal andspatial scale is subject to variation. Certain investigators have argued that interference is not asignificant factor for communities existing in high stress environments (Grime 1979). Otherauthors suggest that the level of competition does not change across environments, while thelimiting resource does change (Tilman 1987). The effects of interference on plants are evidentlyconfounded by other agents operating concurrently or intermittently in communities such asdisturbance and herbivory. Depending on the nature and strength of theses forces, they maydisrupt or supercede the effect of competitive interactions as structuring agents within thecommunity.Methods of studySeveral studies have attempted to measure the effect of interference on plants andpopulations in the field by manipulating the density of neighbours by thinning, transplanting, orremoval (Fowler 1981; references in Connell 1983; Fetcher 1985; Holmes & Jepson-Innes 1989;Keddy 1989; Gurevitch & Unnasch 1989; Aarssen & Epp 1990). Experiments that removegroups of species study the effect of diffuse competition, whereas experiments that manipulate asingle species at a time assess specific differences in competitive ability between species. Diffusecompetition measures the cumulative effect of competition from all neighbours on a target speciesand is frequently studied in field experiments. Single species competition experiments oftenemploy pot or common garden experiments. In these experiments, plants are grown in mixture ormonoculture at different densities and the effect of neighbours are then assessed on the target17species (i.e. de Wit 1960; Fowler 1982; Firbank & Watkinson 1985; Gurevitch et al. 1990). Thescale of response tinder investigation determines what measurements are recorded. Studies ofindividual plant performance frequently measure survivorship, growth, or physiological responses,whereas population experiments tend to measure changes in density, biomass, or cover.Neighbour interactions in communities are measured by changes in relative biomass, speciesdiversity and/or species richness (Goldberg & Barton 1992).ConsequencesAccording to Mack & Harper (1977), interference among coexisting plants can have threepossible consequences: (i) failure to germinate, (ii) death, or (iii) survival combined with a plasticgrowth or reproductive response. The relative importance of neighbours on the performance of atarget plant has been correlated to the size, proximity (or density) and spatial arrangement ofneighbours (Mack & Harper 1977; Silander & Pacala 1985). Weiner (1982) and Aarssen & Epp(1990) suggested that an additional factor such as plant age, or stage of development (i.e. prereproductive vs. mature) can also affect response. Gurevitch et al. (1990) have also demonstratedthat the identity of neighbours was an important factor, such that some species of neighbourswere shown to have greater impact on the performance of a target plant than others.Individual-level responsesIndividuals and populations respond to their local environments by varying patterns ofrecruitment, mortality, growth, and reproduction (Harper 1967; White & Harper 1970; Grace &Wetzel 1981; Weiner 1985; Snow & Whigham 1989; Lieffers & Titus 1989; Schmid & Bazzaz1990; Jonsdottir 1991). Most studies of neighbour interactions tend to concentrate on measuringplastic changes in plant growth and reproduction, although changes in survivorship due to thepresence of neighbours also occur. A few studies that have sought neighbourhood predictors ofplant performance tend to show that at low neighbourhood densities, plants show a plastic18response to interference, whereas mortality occurs primarily when crowding is intense, plants areyoung (Silander & Pacala 1985; Weiner 1982, 1984), or the habitat is productive.Modifying individual plant size through the birth and death of modular units of construction(vegetative or reproductive) is a common response to changes in a plant’s local environment(Harper 1977). Determining how individual plant size varies as a response to neighbours isinteresting because it has often been observed that plant performance, as measured byreproductive output and survival of an individual, is correlated to its relative size within apopulation (Soibrig 1981; Meagher & Antonovics 1982; Wolfe 1983). Plant populationsgenerally show a skewed size distribution comprised of a few large (dominant) individuals that arereproductive, and many smaller individuals that often fail to reproduce and experience high levelsof mortality (thinning) due to suppression from neighbours (Yoda et al. 1963; White & Harper1970; Harper 1977). In these cases, relative size in the hierarchy is an indicator of competitivesuccess and the ability to acquire resources from the environment. Because reproduction is oftenpositively correlated with size in plants, individual size is often used as a fitness correlate (Harper1977; Grace & Wetzel 1981).Several researchers have reported plastic changes in reproductive output due to theinterference of neighbours. (Palmblad 1968; Mack & Harper 1977; Barkham 1980; Lee &Bazzaz 1982; Fowler 1984; Zimmerman & Weis 1984; Whigham 1984; Silander & Pacala 1985).Palmblad (1968) reported that the effect of experimental thinning on reproduction in two speciesof Senecio varied depending on the timing and intensity of the treatment. Most of the differencesobserved were in the number of flower heads produced, total seed production and date of flowerinitiation. Silander & Pacala (1985) determined that local interactions of neighbours accountedfor 70% of the variation in reproduction in Arabidopsis thaliana (L.) Schur.A study of population dynamics in daffodil (Narcissus pseudonarcissus) found that sexualreproduction predominated at lower field densities, whereas vegetative growth occurred at higherdensities (Barkham 1980). This study also found that the number of flowers produced to be19correlated with the amount of growth in the previous year. Similarly, Whigham (1984) in a studyof the effect of competition and nutrient availability on growth and reproduction in Ipomeahederacea 0 found that competition had a negative effect on number and weight of flowers andflower buds, and on seeds and other vegetative characters. In contrast, soil fertility had no effecton number or weight of flower buds.Measuring the plastic growth response to neighbours in clonal plants has been difficultbecause plasticity in clonal organisms is manifested at two levels: (i) changes in the size ofindividual ramets, and (ii) alterations in the rate of production of new ramets and the survival ofexisting ramets. Although Hutchings & Slade (1988) did not study the effect of neighbours onthe dynamics of a clonal plant Glechoma hederacea (L.) they did report that patterns ofhorizontal space acquisition, architecture, and biomass acquisition in G. hederacea varied with theamount of light and nutrients supplied.Population-level responsesPopulation experiments have investigated the effects of neighbours in a variety of ways.Cavers & Harper (1967) followed the fate of seeds and transplants of two species of Rumexplaced in different environments. In habitats where neighbours formed a closed canopy, seedsthat germinated failed to survive the growing season, and transplants showed poor growth. Bothseedlings and transplants did better at more open sites. Although this experiment was notdesigned to test the effect of neighbours on seedling recruitment and the survival of young plants,it is possible that interference from neighbours at the closed sites was responsible for poor growthand survivorship.Gurevitch et al. (1990) studied the effect of inter- and intraspecific neighbours on individualplants of different species, at different fertility levels, in a pot experiment. This study reportedthat the presence of neighbours had a greater effect than simple reduction in available space andthat the effect of neighbours on the target varied with the neighbour’s identity.20Fetcher (1985) studied the effect of removing moss and shrubs on the populations dynamicsof cotton sedge, Eriophorum vaginatum in central Alaska. Cotton sedge tussocks in neighbourremoval treatments had more daughter tillers and smaller adult tillers than controls. The responsewas largely attributed to changes in irradiance when neighbours were absent rather than changesin nutrient status. The presence of neighbouring species in a grassland habitat in south-easternArizona was shown to limit the distribution of Stipa neomexicana by limiting seedlingestablishment and survival, as well as reducing the rate of flowering and growth of mature plants(Gurevitch 1986). The act of removing neighbours allowed S. neomexicana to invade th morefavourable habitat previously occupied by neighbours. Lee & Bazzaz (1982) investigated theeffect of competitor removal, nutrients, and water on reproduction in an annual legume, Cassiafasciculata. In the absence of neighbours, biomass and branching increased relative to thecontrols, however, when neighbou Lemoval was combined with fertilizer and/or water, plantbiomass increased beyond neighbour removal alone. The number of fruit matured per plant alsoincreased in the combined treatment,Community-level responsesStudies at the community level show a variety of responses to species removal. In a 4-yearstudy, Jonasson (1992) examined the effect of removing a dominant shrub on various types oftundra communities. Few changes in species diversity or various estimates of cover weredetected in this study unless the removal of the dominant shrub was combined with the addition offertilizer. The combined effects of removal and fertilizer increased diversity in the stable tundraand decreased diversity on the disturbed (frost-heaved) tundra. In this case, disturbance wasshown to increase the soil nutrient pool relative to the stable communities. Fertilizer was reportedto increase graminoid cover substantially. Jonasson (1992) concluded on the basis of these resultsthat the dominant shrub species was not likely to be competing with the herbaceous (vascular,bryophyte and lichen) community.21Keddy (1989) performed a similar 4-year removal experiment on wetland vegetationoccurring along an environmental gradient. In this case, removal of a dominant shrub producedhighly significant increases in cover, richness and diversity in the community, although less than25% of the species showed a significant individual response to the experimental treatments. Thesingle species that did respond to the removal treatment were reported to comprise a largeproportion of the seed bank. In areas of low fertility and high susceptibility to wave disturbance,no competitive release was detected following shrub removal.Gurevitch & Unnasch (1989) studied the combined effect of the removal of a dominantgrass, and application of fertilizer on an old field community in Long Island, New York. Fertilizeralone was reported to reduce species diversity and richness, while fertilizer combined with theremoval of the dominant maintained species richness. Removal of the dominant species increaseddiversity particularly at high levels of soil fertility. This study concluded that competition (fromthe dominant) was more important in structuring the community at high levels of fertility than atlow levels.Fowler (1981) investigated the effect of three types of species removal on percent cover ofindividual species occurring in a North Carolina grassland community. These included (i)graminoids, (ii) dicots, or (iii) removal of a number of single species. Removal of single specieshad less effect than removing groups of grasses or dicots. This may have been the result ofgreater biomass removed when groups of species were eliminated. Evidence was presented thatsome grasses limit some dicots and vice versa. Based on these results, Fowler (1981) suggestedthat there is indirect support for the hypothesis that this grassland community is characterized byweak and comparatively equal competitive relationships among its component species. The abilityof neighbours to have a strong effect on the structure of this community was not clearly shown.Other factors such as herbivory may be more important in this community, although this was notstudied.22The results of several experiments investigating the effects of removing a dominant specieson the community structure in different habitat types indicate that neighbour interactions arevaried and habitat-specific to a certain degree. These three experiments provide good evidencethat the effect of a dominant neighbour varied with the disturbance regime and soil fertility.LimitationsAlthough many studies have shown significant responses in plants when neighbours areremoved, Waller (1981) and Campbell et al. (1991) urged caution when interpreting these results.Campbell et al. (1991) suggested that the responses reported in some experiments may be anartefact of the disturbance imposed during the treatment application (puffing up roots, residualherbicide effects) rather than a product of growing without neighbours. These disruptions couldbe important in infertile communities where the disruption associated with removing neighboursalters the nutrient regime. However, Aarssen & Epp (1990) suggested that the additionalnutrients released were resources that would have been available if they hadn’t been usurped byneighbours. More specifically, these nutrients were likely to have been the objects of pastcompetition and therefore do not pose a problem when assessing the results of these experiments.Campbell et al. (1991) also suggested that seasonal variability in resource capture andgrowth may make the timing of removal critical in determining results of the experiment. Waller(1981) suggested that a significant response may not be an important factor in plant communitiesif the effect of neighbours is subject to temporal or spatial variability. It was suggested that somefactors that may prevent neighbours from exerting a significant effect include: (i) recentdisturbance, (ii) frequent herbivory, and (iii) general stress that is capable of reducing growth rate(Waller 1981). Wiens (1977, 1984) suggested that the critical structuring agents in a communitymay operate for only short periods of time and are therefore not easily detected in the time frameof most experiments. It was also argued that competitive release may not be the reason forincreased plant growth following neighbour removal in these experiments. In spite of the23complications associated with this method, it is unlikely that all significant responses reported todate are the artefacts of treatment application and uncontrolled density manipulations. Pot andgarden experiments that examine the effect of neighbours on plant dynamics with density as avariable have shown significant responses in plant growth depending on whether the species wereplanted in monoculture or mixture (Banyikwa 1988; Holmes & Jepson-Innes 1989; Wilson 1989;Gurevitch et al. 1990). These experiments neither disturb nutrient conditions, nor vary plantdensity in an uncontrolled fashion. It remains true however, that removing neighbours is one ofthe few methods that can be used to study neighbour effects in the field in an established plantcommunity.THE EFFECT OF FERTILIZERPlants require a variety of nutrients to grow and reproduce, and as a result, the availabilityof nutrients is a key factor in determining the productivity of a habitat. Other factors consideredto be important determinants of productivity are soil moisture content and aeration (Trudgill1979), light and temperature (DiTommaso & Aarssen 1989). Although the ability of fertilizer toenhance growth in plants is not disputed, the effect of fertilizing a natural community is notnecessarily predictable. For instance, Grime (1977) suggested that plants adapted to toleratestressful environments should not necessarily respond to short term changes in resourceavailability when fertilizer is applied. Mellinger & McNaughton (1975) also suggested that olderplant communities may show greater resistance to perturbations (fertilizer) than youngercommunities. A review of the effects of addition of inorganic fertilizer will be presented withemphasis on studies pertaining to NPK (nitrogen:phosphorus:potassium) additions. Anotherextensive review on the effect of resource manipulations in natural vegetation is in DiTommaso &Aarssen (1989).Few studies in ecology are interested in studying the effect of fertilizer alone on plantdynamics. Most recent studies focus on determining how changes in soil fertility affect the24dynamics of plants and populations when they are subjected to different levels of interference,herbivory, and disturbance. The question of how these factors interact under field conditions isfrequently addressed in order to ascertain which factors are most important in limiting populationsand determining the patterns and structures of communities. Manipulating resources in fieldstudies is also used to provide data on the mechanisms and consequences of species interactionsand other community processes that cannot be obtained through green-house or descriptivestudies (DiTommaso & Aarssen 1989). Although there has been much debate in communityecology as to whether top-down factors are more important than bottom-up forces such asnutrient regime, Hunter & Price (1993) proposed that both operate simultaneously to regulatecommunities. Hence, studies of how soil nutrient availability interacts with other factors limitingplants and populations are of interest.The addition of inorganic fertilizer to a habitat can have a number of potential effects onorganisms within the community beyond simply increasing the availability of nutrients. Some ofthese include: (i) changing ambient soil chemistry that may result in changes in the availability ofother required nutrients (Wilson 1987). This may occur through changes in pH or salinity of thesoil. (ii) Fertilizer may change the composition or activity of the soil micro- and meso-fauna. Forexample, Halvorson et al. (1992) state that high levels of soil nitrogen can have inhibitory effectson nitrogen-fixing activity and nodule formation in lupines. It is not clear whether increases insoil nitrogen concentration mediate this symbiosis by suppressing the microbe, or by acting on thehost. (iii) Application of fertilizer may change the rate of nutrient cycling through the habitat(Gosz 1987). This may occur through changes in patterns of litter deposition in response tofertilizer. Tanner et al. (1992) reported that fertilization (N+P) of tropical montane forestincreased trunk growth and leaf production. This response was detected after an initial lag andwas measured as increased trunk diameter and litterfall in response to fertilizer. The increase inlitterfall was greatest when both nitrogen and phosphate were added together in contrast totreatments of either nutrient added alone. The quality of the litter after fertilization reportedly25changed when phosphate was added to the system such that both treatments that appliedphosphate as P, or N+P, showed increased levels of phosphate in litter. In spite of this, changesin the quality of live leaves in response to fertilizer were not detected and the authors postulatedthat much of the nitrogen and phosphate added as fertilizer was not likely incorporated intoabove-ground biomass.Individual-level responsesAt an individual level, changes in ambient soil fertility can have a variety of effects on plantdynamics. For instance, the addition of fertilizer may change a plant’s nutrient status and hencetheir palatability to herbivores (Wilson & Stinner 1984; Coley et al. 1985; Mihaliak & Lincoln1985, 1989; Larsson et al. 1986; Bryant et a!. 1987a, b; Wilcox & Crawley 1988; Loader &Damman 1991). In some cases fertilizer increases plant quality by increasing the concentration ofnutrients in the tissue (Fox & Morrow 1992). This change in physiological status, measured asincreased nutrient concentration, makes a plant a more attractive food source to herbivores(Mihaliak & Lincoln 1985 & 1989; Bryant et al. 1987a, b; Loader & Damman 1991). Thealternative is that fertilizer may stimulate the production of anti-herbivore defences that reducepalatability (McClure 1980). Anti-herbivore defences include a variety of structures andcompounds including thorns, spines, resins, digestion inhibitors, and toxins (Crawley 1983). Thetwo commonly discussed groups of chemical defences known to vary with soil fertility are: (i)carbon-based compounds such as resins, tannins, terpenes, sesquiterpenes, and phenolics, and (ii)nitrogen-based compounds including alkaloids and cyanogenic glycosides. Carbon-baseddefences are considered quantitative defences because their inhibitory properties are moreeffective in large doses or at high concentration. These compounds often act to physically inhibitor interfere with the digestive activity of enzymes in the herbivore (Crawley 1983). Nitrogen-based defences are generally highly toxic compounds that occur in plants in low concentration26relative to carbon based defences. Many specialist herbivores have shown evolutionaryadaptations to the presence of these compounds in contrast to carbon-based defences.According to Bryant et al. (1983), nitrogen based defences are more common than carbon-based defences in dicotyledonous herbs growing in nutrient rich sites, or that are nitrogen fixers.The concentration of nitrogen based defences in Lupinus succulentus has been reported to varywith the level of nitrate fertilizer (Johnson et al. 1987). This study suggested that the quantity ofalkaloids in nitrogen fixing lupines depended on levels of soil nitrate because defoliated plants haddepressed rates of nitrogen fixation due to reduced levels of carbon available to symbionts as aresult of herbivory.Wilcox & Crawley (1988) also reported changes in alkaloid concentration, and amino acidconcentration in response to fertilizer in Senecio jacobaea L. foliage. Fertilizer (ammoniumsulphate) slightly increased alkaloid concentrations and decreased amino acid concentrations. Thedecrease observed in amino acid concentration following nitrogen fertilization contrasts withother studies that show increased nitrogen or amino acid concentration with nitrogen fertilization(Onufet al. 1977). However, fertilizer may rapidly produce a flush of new growth in plants withno associated increase in nitrogen content of tissues detected (Crawley 1983).Population-level responsesChapin & Bliss (1989) studied the population-level response of two herbs (Eriogonumpyrolifolium Hook and Polygonuin newberryii Small) to treatments of fertilizer and water. Theresponse to fertilizer was measured as seedling growth and survivorship. Polygonum sp. showeda substantially greater increase in dry mass in response to fertilizer+water relative to Eriogonumsp. The authors concluded that Eriogonum sp. that grew slower under high nutrients was moretolerant to nutrient stress. Another study examined the effect of NPK fertilizer (withmicronutrients added) and shading on seedling dynamics and herbivore resistance of Betulapendz4a Q(Rousi et al. 1993). Fertilizer was reported to increase growth rate and palatability of27seedlings to voles, but only one out of four families of Betula pendula became more palatable tohares following fertilization. The response of birch to fertilizer in this study did not concur withthe results of similar experiments on birch (see refs in Rousi et al. 1993). Lieffers & Titus (1989)also reported larger seedlings when fertilizer was applied to white spruce and lodgepole pineseedlings. Fertilizer in this experiment was also reported to increase differences in growth ratethat resulted in size inequalities within the population.When a nutrient in limiting supply is added, it is expected that a plant will increase itsgrowth rate until other factors become limiting. This increase in growth rate can be beneficial toplant’s in some environments, but Sarukhan & Harper’s (1973) study of Ranunculus spp.dynamics also report that the greatest risk of death for plants occurred when they were growingmost rapidly. This was also noticed by Sukatschev who reported a 6% death rate for Matricariainodora L. on unfertilized soil and a 25% death rate on fertilized soil (described in Crawley 1983).Further evidence supporting this hypothesis is found in studies of self-thinning along aproductivity gradient (i.e. Lieffers & Titus 1989). Comparison of thinning rates along a gradientof nutrients (NPK) showed that self-thinning was most intense in the high fertility treatments(Yoda et al. 1963). Some authors suggest that increasing the soil fertility in a populationexperiencing self-thinning likely results in increased growth in some individuals that results in thesuppression and death of others. Fertilizer may act by decreasing the time required before apopulation becomes light limited and self-thinning begins (Harper 1977; Crawley 1983).The effect of fertilizer on tissue longevity has also been studied. Watkinson et al. (1979)reported that fertilization of Carex arenaria L., a sand dune sedge, increased both birth rate anddeath rate of tillers. This increased turnover reduced the longevity of tillers, resulting in a changein the age structure of the population. Increased turnover of leaves in response to elevated levelsof soil nutrients has been reported or proposed as an explanation of elevated turnover rates inother studies (Bryant et al. 1983; Shaver 1983; Kotanen & Jefferies 1987, Diemers et al.1992).28Several studies have reported changes in the allocation of carbon to roots and shoots underdifferent fertility regimes. It is generally observed that application of fertilizer results in adecreased allocation to root biomass (Chapin & Bliss 1989; Bastow Wilson 1989; Lieffers &Titus 1989). However, Gurevitch et al. (1990) reported an increased root:shoot ratio in a potexperiment in response to NPK fertilizer in three species of herb (Achillea millefolium L,,Dactylis glomerata L., and Vicia cracca L.).Fertilizer effects on reproduction in plants have been described in some studies. Davy &Bishop (1984) reported that fertilizer resulted in a more rapid initiation of flowering as well as afive fold increase in the number of flowers produced by Hieraceum pilosella. However,flowering in H. pilosella stopped after one year of fertilization. This was likely due to theincreased activity of competitors. Stephenson (1984) in a study of maternal investment in Lotuscorniculatus noted that NPK fertilizer produced an increase in seed production per ramet, as wellas increased the number of flowers and matured fruit produced per plant. Fertilization of alegume, Cassiafasciculata, was also reported to increase fruit production per plant (Lee &Bazzaz 1982). In a study of Ipornea hederacea, an old field perennial reported that fertilizer hada significant effect on biomass of leaves, stems, roots, seeds and fruit, but no effect on the numberor biomass of flowers (Whigham 1984).The effect of fertilizer under different conditions of available space has been investigated bya few authors. In Gurevitch et al. (1990), available space was manipulated by growing plants ineither monoculture or mixture in different sizes of pots. They reported that the impact of fertilizeron various measures of plant performance was greater in large pots compared to small pots.Donald (1963) did not manipulate available space directly, but rather manipulated density of asward. He reported that nitrogen uptake by swards was reduced at high density, and postulatedthat high density conditions result in shallower rooting depth and less efficient soil exploration bythe population. It was further suggested that as the fertility of an environment improves, a greaterdensity is required to achieve maximum yield. Banyikwa (1988) also reported that defoliated29grasses showed an interaction between nitrogen fertilization and density that decreased yield perplant.Competition experiments have frequently manipulated soil fertility and measured theperformance of species growing in mixtures. In Wilson (1989), root competition in pairwisemixtures of three species of upland grasses was assessed. Soil fertility was manipulated withapplication of sodium nitrate. Nitrogen application to different species mixtures was found todepress growth in one species and increase growth in the remaining two species. In thisexperiment increased nitrogen availability allowed two species to outcompete a third species.Similarly, nuthent additions to tundra communities have sometimes shown that within a season,the response among species varies; this between-species variation in the pattern of response tofertilizer also varies between years although overall community productivity remained stable(Chapin & Shavers 1985).Schmid & Bazzaz (1992) tested the effect of changing patterns of resource availability withNPK fertilizer on genet architecture. They hypothesized that rhizomatous perennials shouldincrease rhizome numbers but decrease their length in fertilized soil compared to unfertilized soil.Several species were tested including graminoids and dicots. They ranged from spreading tocompact growth forms. The results were mixed and no clear evidence was found to support theirhypothesis. Only one species (Aster lanceolatus) showed an increase in rhizome number in thefertilizer treatment. Changes in rhizome length were not significant for any species. It wassuggested that either response to fertilizer was delayed and not seen in this short term experiment,or that rhizome architecture is subject to some constraints in terms of plasticity of response.DiTommaso & Aarssen (1989) suggest that age and/or successional stage may influence theresponse of some plants to nutrient additions, although the evidence to clearly show this is notavailable.Nitrogen fixers have been reported to respond differently to nitrogen based fertilizers thannon-fixing species depending on whether they are grown in mixture or pure stand. Dennis &30Woledge (1984) studied the effect of nitrogenous fertilizer on growth of mixed swards ofTrifolium repens L. (white clover), and Loliurn perenne L.(ryegrass). Application of nitrogenfertilizer when potassium and phosphate were maintained in adequate supply resulted in asubstantial increase in leaf area index and yield in the sward. This increase was attributed entirelyto the response of ryegrass. Clover was reportedly reduced to one fifth its leaf area index andyield relative to the unfertilized control. No differences were observed in petiole length, laminaarea, or specific leaf area of newly expanded clover leaves. The number of live leaves on cloverstolons was significantly greater in the absence of nitrogen fertilization. Early in the experimentmeasures of photosynthesis in clover were higher in the control relative to the nitrogen treatment.This difference was lost as the experiment progressed. Other experiments reporting growthreduction in clover grown in mixed sward have attributed the suppression of clover to gradualshading by neighbouring grasses as the season progressed (described in Donald 1963). Incontrast, this study concluded that shading from ryegrass did not reduce photosynthesis in cloverby the end of the experiment as would be predicted if ryegrass suppressed clover by shading(Dennis & Woledge 1984). Although clover did show growth reduction in the nitrogentreatment, the authors attribute this to early season effects rather than a progressive increase inshading as ryegrass overtopped clover (Dennis & Woledge 1984).Community-level responsesDiTommaso & Aarssen (1989) in a review of the effect of several types of resourcemanipulations (water, nutrients, light) on various types of plant communities from agroecosystems to arctic tundra concluded that the wide range of responses to fertilizer reported infield studies depends on the habitat type, duration of resource enrichment, and the resourcesadded. This review also describes several fertilizer studies of communities containing leguminousspecies. Thurston (1968), Zarzycki (1983), Traczyk et al. (1984), and Hobbs (1988) reportedthat while fertilizer stimulated grass growth, leguminous species were significantly reduced or31eliminated. Henry et al. (1986) also noted that increased tiller growth in response to increasednutrient availability is common in northern rhizomatous graminoids.At a community level, Jonasson (1992) found that the effect of fertilization (NPK) of tundracommunities varied with the disturbance regime. On stable tundra, species diversity increasedwhen fertilized, whereas frost-heaved communities showed a decrease in diversity when fertilizerwas applied. Another consequence of fertilization was increased above-ground biomass. Incontrast, Gurevitch & Unnasch (1989) reported that fertilization of a stable, old field herbaceouscommunity led to increased productivity associated with a reduction in both species richness anddiversity. In this same study, application of fertilizer was combined with the experimentalremoval of a dominant grass (Daclylis glornerata). The response to fertilizer (NPK) changed sothat in the absence of D. glomerata, species diversity increased more in the high fertility treatmentthan in the low fertility treatment. Hence the response of populations to fertilizer in an old fieldcommunity was mediated by the response of a dominant species.Some studies have investigated the interaction between fertilizer, defoliation andcompetition. Banyikwa (1988) in a pot experiment examined the growth response of twoperennial grasses in a factorial experiment that manipulated nitrogen level, defoliation, pure ormixed culture, and density. In this experiment, nitrogen fertilizer substantially increased plantyield, and increased the shoot:root ratio. An interaction between defoliation and nitrogenfertilization was reported to vary between the two test species (Sporobolus ioclados() andDigitaria macroblephara ()) such that Digitaria sp. showed an increased yield when defoliatedand fertilized with nitrogen, whereas Sporobolus sp. showed a decreased plant yield. Thisexperiment provides additional evidence that response to herbivory is modified by localenvironmental conditions such as nutrient availability and competition.Fowler (1982) in a study of competition and coexistence in a North Carolina grasslandspecies found that the intensity and outcome of competition in a pot experiment varied with soilfertility (NPK fertilizer application). She reported that fertilizer was the most important factor in32determining overall yield in species grown in mixture or monoculture. It was noted thatmanipulating soil fertility in the experiment resulted in reversal of competitive dominance in twoof six pairs of species.There are currently opposing views regarding the intensity of competition alongproductivity gradients. Grime (1979) and Keddy (1989) supports the view that competition alonga productivity gradient is not constant, and unproductive habitats are dominated by stress tolerantspecies that are not subject to competition. In contrast, Tilman (1987), Newman (1973), andGrubb (1985) maintain that competition occurs equally at all positions along a gradient althoughthe limiting resource may change. Regardless of these conflicting predictions soil fertility hasbeen demonstrated to play a key role in many communities (see DiTommaso & Aarssen 1989).Soil fertility plays a key role in many communities through its ability to limit bothindividuals and populations, and in its ability to modify the response of plants and populations toother environmental variables. It is through these complicated interactions with other limitingagents that makes soil fertility of interest in current ecological research.CHAPTER 2DEMOGRAPHY OF LUPINUS LEAVESPlants, as modular organisms, have several levels of population organization. These includepopulations of genets (genetic individuals), ramets (clonally produced individuals), or reiteratedplant parts (e.g. leaves, buds, flowers). As it is often difficult to measure population dynamics ofgenets and ramets in rhizomatous perennials, the relative effects of clipping, neighbours and soilfertility level on the demography of Lupinus leaves (modules) were examined to determine howthese factors affected the birth and death of plant parts (i.e. leaves), and how these factorsinteracted at this level of population organization. Data on the amount of leaf damage (by insector microtine herbivores) and leaf disease were measured as additional information to explainpotential differences in leaf mortality. Tagged sub-populations of lupine clumps were also used tocalculate an index of standing crop available to herbivores through the season and determine howthis differed between treatments.METHODSSTUDY AREAThe experiments were conducted in the Kluane Game Reserve, located in the southwestcorner of the Yukon Territories, approximately 3 km south of the Alaska Highway at BoutelierSummit (138° 22’W, 610 02’N). This area is presently being used as part of a long-term study ofa northern boreal forest ecosystem led by Dr. C.J. Krebs.The dominant vegetation in this tract of northern boreal forest consists of a closed canopyforest dominated by white spruce (Picea glauca Voss), as well as open areas of shrub habitatdominated by grey willow (Salix glauca L.), dwarf birch (Betula glandulosa Michx.), and to a3334lesser degree Shepherdia canadensis (L.)Nutt., and Potentillafruticosa L. The understorythroughout the region is dominated by the woody vines Arctostaphylos uva-ursi (L.) Spreng. andLinnea borealis L., and by herbaceous species such as Festuca altaica Torr., Lupinus arcticusLindi., Mertensia paniculata (Don), Achillea millefoliwn L., Anemone pari4flora Michx.,Epilobium angustifolium L, Solidago canadensis L., Senecio lugens Rich., and a number ofmoss species. The primary mammalian herbivores are snowshoe hare (Lepus americanus), andarctic ground squirrels (Spermophiles parryii). A detailed description of the study area is foundin Krebs et al. (1986).SpeciesThe target species in this investigation was the Arctic Lupine, Lupinus arcticus Lindil.(Family Fabaceae). It is a nitrogen-fixing perennial with a rhizomatous growth form that has beenobserved to grow in a variety of habitats in Kluane including closed canopy spruce-moss forest,open birch-willow shrubland, river flood plains, and alpine tundra. Leaves are palmatelycompound with leaflet number ranging from 6-10. Leaves emerge in the spring following snow-melt from underground stems with only the petioles and lamina breaking the soil surface. Verticalgrowth is obtained primarily through petiole extension which, in some instances, exceeds 20 cm.Flowers occur on indeterminate racemes, and colour ranges from bluish to pinicish-violet.Racemes are observed to emerge from the soil following spring thaw concurrently with petioles.The fruit are dehiscent legumes capable of producing up to 5-8 seeds. A full description of themorphology of L. arcticus is found in Dunn & Gillet (1966). Lupinus arcticus is distributedthroughout the Yukon, Northwest Territories, Alaska, British Columbia and Washington and iscapable of hybridizing with other lupines when distributions overlap (Dunn & Gillet 1966). Somemembers of this genus are known to be produce toxic alkaloids that can deter herbivores (Johnson& Bentley 1988; Dolinger et al. 1973). It is not currently known if L. arcticus produces anti-35herbivore defences. In addition to being a constituent of the summer diet of snowshoe hares,ground squirrels have also been observed to feed on L. arcticus in Kluane.Site SelectionSites for the experimental quadrats were chosen using three criteria: (i) a minimum of 10clumps of Lupinus arcticus occurring in a 1 m2 area, (ii) the minimum distance separatingquadrats must be greater than 1.5 m, and (iii) avoidance of natural hazards such as stumps, boles,and deadfall. These criteria were used to ensure a minimum population density, as well as toreduce the possibility that individuals of L. arcticus in neighbouring quadrats were connected byrhizomes. This method of selection was biased towards (i) selecting high density lupine sites thatwere easier to find in early spring, (ii) selecting plants that tended to break bud earlier and wereeasier to locate as snowmelt occurred, (iii) selecting larger or older clones as opposed to seedlingsand recent recruits which do not generally appear until later in the spring and early summer.Quadrat ConstructionFollowing site selection, 32- 1 m2 areas were each permanently marked with four 20 cmx 5 cm x 5 cm wooden posts inserted into the ground to a depth of 15 cm. The perimeter of theexperimental area was marked with cotton string fastened to each corner post. Orientation of thequadrats was such that the maximum number of lupine clumps were included. All 1 m2 quadratswere then surrounded by a fence, 1.5 mon a side to a height of 60 cm, using 2.5 cm diametergalvanized chicken wire to limit uncontrolled natural grazing by mammalian herbivores thatfrequent the area. The total fenced area was 2.25 m2 to include a 25 cm buffer zone between the1 m2 experimental quadrat and the fence. Treatments were applied over the entire 2.25 m2 area.The herbivore exclosure was supported with four 1 m steel fence posts pounded 20 - 30 cm intothe ground. A 15 cm skirt of chicken wire was folded outward on the ground and secured with15 cm wire staples to deter animals from penetrating under the fence. Fences were checked36regularly to determine if herbivores had penetrated the experimental area and no such evidencewas found during the study. The outer perimeter of the fences were spaded to the depth of aspade blade to sever rhizome connections growing beyond the treated area. This was done onceat the onset of each growing season.EXPERIMENTAL DESIGNThe primary experimental design used for the duration of this field study was a factorialcross of three treatments with two levels in each treatment: +1- fertilizer (F), +1- neighbourremoval (NR), +1- clipping (C). This factorial design allowed the comparison of populationresponses of L. arcticus to individual treatments as well as how main effects interact to modifypopulation dynamics. In 1992, an additional treatment consisting of an elevated intensity ofclipping was imposed, raising the number of treatments from 8 in 1991, to nine treatments in1992. Each treatment was replicated 4 times at one field site for a total of 32 experimentalquadrats in 1991 and 36 in 1992. All quadrats were located in an area less than 1 ha.TreatmentsThe protocols used for treatment application were as follows:FertilizerFertilizer was applied at the start of each growing season following spring thaw withHutchinson’s NPK (35-10-5 percent by weight) fertilizer mixture. On June 1, 1991, theapplication rate per m2 was 9 g nitrogen, 2.5 g phosphorus, and 1.27 g potassium. The chemicalcomponents of this mixture were ammonium nitrate ((NH4)2NO3,super phosphate (H2PO4),and potash (K20). In 1992, the amount of fertilizer applied was doubled due to a low responseby L. arcticus in the first year. Fertilizer was applied at the same rate as 1991, but was appliedtwice in the season (May 27 and June 30). Fertilizer was applied dry, sprinlding evenly over theentire area within the fence. The rate of application in 1992 corresponded with other fertilizer37udies in the area and is within established rates of forest fertilization projects (Binkley 1986;Nams et al. 1993).ClippingTo control the amount of leaf tissue removed in this experiment and ensure consistencybetween treatment replicates, a simulated grazing treatment was used in place of natural grazing.Petioles and/or stems were clipped to a constant height of 8 cm using scissors and maintained atthis height using a press design such that plants were continually clipped as re-growth occurred(Bender et a!. 1984). This pattern of leaf removal is similar to the pattern commonly attributed tohare grazing in the field. The initial clipping treatment was imposed mid-season (July 4, 1991,and June 29, 1992) when the plants reached a height of approximately 10 cm. Leaf tissueremoved was collected and dried at 50 °C for 5 days and then weighed. In the elevated clippingtreatment done in 1992, (designated as C+ to distinguish it from the main clipping treatment C)plants were clipped to a height of 4 cm. All other aspects of this clipping treatment wereunchanged.Neighbour RemovalTo examine the effects of the presence of neighbours on populations of Lupinus arcticus,all herbaceous neighbours as well as woody ground-growing vines, mosses, and spruce seedlingswere removed. Neighbours were removed at the start of the growing season (May 26, 1991),immediately after quadrat erection, by pulling them out of the ground. Care was taken to avoidexcessive disturbance of the soil and of neighbouring lupine clones. Neighbours were continuallyremoved throughout the season in 1991 and 1992 as regrowth or reinvasion occurred.38LEAF DEMOGRAPHY STUDYSampling RegimeFollowing quadrat construction and treatment application in May 1991, biweekly surveysof L. arcticus populations commenced for a total of 7 surveys in 1991, and 6 surveys in 1992. Atthe height of the growing season, a survey took up to 5 days to complete. To control the lengthof the sampling interval, quadrats were surveyed in the same order each time. The followingvariables were recorded at each survey:Population variables Growth variablesleaf density number of racemes petiole lengthbud density number of flower buds raceme lengthleaf mortality number of fruits (pods) peduncle lengthnumber of diseased leaves seedling recruitmentnumber of damaged leavesThe category of damaged leaves measured the number of leaves per quadrat showing visible signsof tissue removal or herbivore damage. The disease category measured the number of leaves perquadrat showing signs of infection including discolored leaves, necrotic spots, the presence of andfungi colonies on the leaf surface. The analyses of the above variables are divided into differentsections for convenience. This chapter deals with the analyses of leaf and bud density, leafmortality, and number of diseased leaves. Chapter four presents the results of a separateexperiment on leaf cohort survivorship. Chapter five, analyzes vegetative growth data such aspetiole length. Chapter six summarizes all of the reproductive data from this experiment includingflower and raceme production, and fruit (pod) production, seediing recruitment and size ofreproductive structures (raceme and peduncle lengths).STATISTICAL ANALYSESThese experiments were based on a repeated measures design such that the same variableswere measured on the same quadrats at bi-weekly intervals throughout the season. To analyze39the population data collected, a Pearson Multiple Correlation (Systat 1992) was first done usingSystat for Windows 5.0 on individual population variables to determine if these variables showeda serial correlation in time. The variables analyzed included leaf density, bud density, number ofmissing leaves, and incidence of disease. This analysis was done by generating a Pearsoncorrelation matrix for each of the four variables after they had been pooled within a single survey,across treatments. The matrix generated in this analysis showed the correlation within a variable(e.g. leaf or bud density) between surveys when the effects of treatment removed. Afterexamining the partial correlation matrix, if no correlation in time were detected, time was not usedas a factor and all measurements of these variables over time were considered as replicates(White, R. personal communication). Population variables were then analyzed individually with athree-way or a one-way ANOVA by treatment. A three-way analysis was used to test the effectof the main factorial design in 1991 and 1992, and the one-way analysis was used for unbalanced1992 design to test the effect of the additional clipping treatment. Post hoc Tukey comparisonswere performed to ascertain where differences occurred. With the exception of bud density in1992, correlation in time was not found for any of the population variables. To analyze buddensity data in 1992, a repeated measures MANOVA by treatment was performed with survey asthe repeated measures variable. The data on incidence of leaf damage were not analyzed and isnot presented in this thesis for two reasons: (i) because herbivores were excluded from theexperimental quadrats the level of leaf damage detected in the field was minimal, and (ii) t wasdifficult to distinguish between damage caused by insects and damage from disease. As a result,most damage was classified and analyzed as incidence of disease. Similarly, the category ofmissing leaves was also not analyzed or presented because little evidence was found for theremoval of leaves by insect or microtine herbivores in 1991 or 1992.40STANDING CROP AVAILABLEAs a supplementary experiment, the response of individual plants to treatments wasmeasured and used to calculate the standing crop of leaves available as a potential food source forherbivores (live summer leaf-days, Ci). Ten lupine clumps were tagged in each quadrat at thebeginning of the first growing season so that each of the 8 treatments consisted of 40 taggedclumps of lupines. Clumps were selected to include a range of sizes in each quadrat. Taggedlupines were numbered and measured as individuals for two seasons. In the second season,lupines reappeared at the same locations marked in the previous field season. From 320 taggedclumps, only one cluster died during the study and did not reappear in the second year. Thereforesurvivorship curves were not generated for these clumps. Reproduction (measured as racemeproduction, flower initiation, and pod production) in the tagged population was negligible in 1991and low in 1992.The formula used to estimate the standing crop available (live summer leaf-days - Ci), wascalculated for each treatment as the sum (over all surveys done in one summer) of the meannumber of leaves available at during each survey interval multiplied by the length of time that theywere available (i.e. alive) to herbivores during that particular time period (i.e. survey interval).For example, the mean number of leaves available during the first survey interval would becalculated as the leaf density at survey 1 + survey 2 divided by 2. This method of calculating thenumber of leaves available during a survey interval uses the mean number of leaves between twosurveys in order to reduce any possible bias (over-estimate or under-estimate) that is the result ofinstantaneous birth or death of leaves at the onset of each survey. This mean number of leavesavailable during a survey interval would then multiplied by the number of days between survey 1and 2 to get an estimate of the number of live leaf-days available for one survey interval. Thisnumber was calculated, and summed for all surveys done in a summer for each quadrat andtreatment to obtain an estimate of standing crop of live leaves available for consumption for the41entire season. Simply, live summer leaf-days (C1), takes into account the number of leaves thatare available as food, as well as how long they are available for consumption.Live summer leaf-days C1 = EkEjj [O.5*(Lj+Lj+1)*(At)]jk (1)Time available At = tj1-t (2)L1=number of leaves alive at survey i, At=no. of days between survey i and i +1, j=quadrat,k=treatmentStatistical analysisThe variable live summer leaf-days, C1 , was log-transformed to make variancehomoscedastic, and analyzed using a three-way ANOVA by treatment in 1991 and 1992.RESULTSLEAF DEMOGRAPHY 1991In 1991, all population variables were analyzed using a three-way ANOVA. Populationdata collected in 1991 are summarized in Table 1. In 1991, clipping significantly reduced leafdensity (P=O.015), and incidence of disease (P=O.009) in leaves, while removing neighbourssignificantly reduced (P=O.OO1) leaf mortality (Table 2). Application of fertilizer resulted in asignificant (P=O.047) increase in the incidence of disease (Table 2).LEAF DEMOGRAPHY 1992To determine the effect of the additional clipping treatment (C+) added in 1992 , one-wayANOVAs were run in conjunction with the original three-way factorial ANOVAs. Data on thepopulation variables analyzed in 1992 are summarized in Table 3. Three-way ANOVA results ofleaf density, leaf mortality and incidence of disease are reported in Table 4. As in 1991, clipping42reduced leaf density and incidence of disease (P=O.05, Table 4). No other significant treatmenteffects were detected including the elevated fertilizer treatment.One-way ANOVA indicated significant treatment effects for both leaf density (P=O.004)and disease (P=O.026, Table 5 a). A post hoc Tukey comparison of the leaf density resultsindicated that intense clipping, and the interaction fertilizer by clipping had significantly lower leafdensities than populations where neighbours were removed (P<O.05, see Table 5 b). Tukeycomparison indicated that populations that were intensely clipped (C+) had less disease than thosefertilized (F) (P<O.05, Table 5 c). Tukey comparison also indicated a moderately non-significanttreatment effect for regular clipping (C) (P<O.067, Table 5 c).Table 1: Summary of mean leaf density, bud density, leaf mortality and incidence of diseaseby treatment in 1991. Note that n=32 as replicates were pooled across surveys.Treatment Mean ± Mean bud ± Mean ± Mean ±leaf SEM density SEM leaf SEM incidence SEMdensity /m 2 mortality of disease/m2 /m2 /m2Control 454.7 42.0 125.8 26.6 76.2 18.6 208.5 37.7Fertilizer 591.3 57.5 146.6 25.9 118.1 16.9 393.6 61.3Clipping 426.4 30.5 128.3 21.6 62.74 11.0 211.5 34.5Removal 476.9 62.9 94.5 24.4 49.4 14.3 271.1 48.5Fert/Clip 425.1 47.7 135.4 13.9 69.0 12.8 261.9 49.2Fert/Rem 568.3 59.8 138.4 26.1 55.0 11.9 262.9 50.9Clip/Rem 426.4 30.5 128.3 21.6 62.7 11.0 211.5 34.5Fert/Clip/Rem 464.6 51.4 180.6 30.3 47.4 9.2 239.9 44.943Table 2: Source of variation for three-way ANOVA results of bud density, leaf density, incidenceof mortality and disease in 1991.Variable Source dF MS F-ratio Prob.Bud density Fertilizer 1 51656.813 3.263 0.072Clipping 1 17057.054 1.077 0.300Removal 1 235.718 0.015 0.903Fert*Clip 1 210.164 0.013 0.908Fert*Rem. 1 15073.700 0.952 0.330Clip*Rem. 1 26679.233 1.685 0.196Fert*Clip*Rem 1 1325.612 0.084 0.773Error 216 15833.494Leaf density Fertilizer 1 241630.248 3.488 0.063Clipping 1 420288.378 6.068 0.015Removal 1 5818.714 0.084 0.772Fert*Clip 1 130540.41 1.885 0.171Fert*Rem. 1 212.396 0.003 0.956Clip*Rem. 1 6260. 122 0.090 0.764Fert*Clip*Rem 1 23863.012 0.345 0.558Error 216 69265.93Dead Fertilizer 1 9723.691 1.911 0.168Clipping 1 17675.961 3.473 0.064Removal 1 55487.714 10.903 0.001Fert*Clip 1 6236.706 1.226 0.270Fert*Rem 1 6708.858 1.318 0.252Clip*Rem 1 10204.707 2.005 0.158Fert*Clip*Rem 1 2943.616 0.578 0.448Error 216 5089.071Disease Fertilizer 1 236831.97 3.988 0.047Clipping 1 162944.66 2.744 0.009Removal 1 31168.73 0.525 0.470Fert*Clip 1 30739.12 0.518 0.473Fert*Rem 1 155601.40 2.620 0.107Clip*Rem 1 6072.24 0.102 0.749Fert*Clip*Rem 1 108049.761 1.819 0.179Error 216 59388.11644TaN 3: Summary of mean leaf density, bud density, leaf mortality and incidence of diseasein 1992. Note that n=32 as replicates were pooled across surveys.Treatment Mean ± Mean ± Mean leaf ± Mean ±leaf SEM bud SEM mortality SEM incidence SEMdensity density un 2 of disease1m2 /m2 /m2Control 505.13 49.3 98.96 26.01 80.71 16.35 267.08 49.51Fertilizer 540.50 62.6 115.13 17.77 116.13 23.05 386.67 66.27Clipping 386.21 47.15 79.38 15.47 68.04 14.32 175.63 38.93Removal 600.13 100.6 122.38 36.19 119.58 34.31 316.00 73.97Fert*Clip 313.67 31.6 98.00 12.95 73.5 16.62 215.13 34.31Fert*Rem 529.04 55.7 103.78 18.58 62.5 14.28 256.63 55.62Clip*Rem 469.92 49.3 103.25 21.96 69.33 13.54 191.54 37.16Fert*Clip*Rem 457.58 50.5 121.38 17.39 74.04 15.66 246.58 46.26Extra Clip 327.92 47.4 90.54 18.29 55.54 13.97 144.38 25.3345Table 4: Source of variation for three-way ANOVA of leaf density, incidence of disease andmortality in 1992.Variable Source dF MS F-ratio ProbabilityLeaf density Fertilizer 1 20 155.964 0.236 0.628Clipping 1 693738.713 8.111 0.005Removal 1 99072.234 1.158 0.283Fert*Clip 1 47026.124 0.550 0.459Fert*Rem 1 11973.008 0.140 0.709Clip*Rem 1 43076.778 0.504 0.479Fert*Clip*Rem 1 54865.83 0.641 0.424Error 184 85528.818Dead Fertilizer 1 186.445 0.020 0.888Clipping 1 18190.379 1.948 0.164Removal 1 756.684 0.08 1 0.776Fert*Clip 1 1403.158 0.150 0.699Fert*Rem 1 20098.605 2.153 0.144Clip*Rem 1 717.173 0.007 0.782Fert*Clip*Rem 1 28965. 179 3.102 0.080Error 184 9336.539Disease Fertilizer 1 72734.167 1.117 0.292Clipping 1 412207.455 6.331 0.013Removal 1 114.366 0.002 0.967Fert*Clip 1 6354.721 0.098 0.755Fert*Rem 1 81358.365 1.249 0.265Clip*Rem 1 36692.861 0.564 0.454Fert*Clip*Rem 1 140543.972 2.158 0.143Error 184 65113.83AAAAAAAAl,444+dIUODNJ%J]IIN/D/dDkITNJD3+5j661U’iuomauAqsisTpjosts(juiM-uoJouosudmoo‘ni:(oçjqjAAAAAAAA4+4JNTWST.kT1U03JN/DWJJD/.I3+3Z66TU!iuwiiauAqXiisupjiJosrsApuii-uojouosuduio3i(jnJ:(qçjqj166TUIpppiuwnan&nddijosuutiiwoiiippiqipwIotho3uIsist(pusIq1661rniuunaiiCqJO3UTOWputI1!SUpJUJJOsisXtiiMM-UOojUOUJAjoatnog:(ç‘[qj9IL9869E6cLOOO968TJ01.I9W0c17vz8EV9ZEET8cLoT99oT11JSUS!Uz%ccL66LLOOO69T1OJ.IP000OO6c6c8I76Tz86cL88ccc8TW1k.L1!SUpJ’I£1qq0JJO!11U-SI’SlIPSSZUflO47Table 6: Source of variation for one-way repeated measures MANOVA of bud production in1992.Source SS dF MS F-ratio ProbabilityBetween treatmentsTreatment 1855588.8 8 231948.6 0.913 0.521Error 6860454.0 27 254090.9Within treatmentsTime 7101515.4 5 1420303.1 107.882 0.000Time*Treatment 815643.6 40 20391.2 1.549 0.034Error 1777320.0 135 13165.3Polynomial testof order (linear)Time 2284348.4 1 2284349.4 205.622 0.000Time*Treatrnent 358017.9 8 44752.2 4.028 0.003Error 299954.9 27 11109.4Table 7: Summary of multivariate test statistics for one-way and three way repeatedmeasures analysis of bud production/rn 2 in 1992,Variable Test Statistic dF dF F ratio Probability(hyp) (error)ONE-WAYTime Wilks’ Lambda 0.063 5 23 67.927 0.000Time*Treatment Wilks Lambda 0.098 40 103 1.813 0.009THREE-WAYTime Wilkst Lambda 0.060 5 20 63.042 0.000Time*Clip Wilks’ Lambda 0.506 5 20 3.927 0.01248Table 8: Three-way repeated measures MANOVA of bud production/rn 2•Source SS dF MS F-ratio ProbabilityBetween treatmentsFertilizer 43621.021 1 43621.021 0.169 0.685Clipping 898995.021 1 898995.021 3.474 0.075Neighbour Removal 290474.083 1 290474.083 1.123 0.300Fert*Clip 7252.083 1 7252.083 0.028 0.868Fert*Removal 6417.187 1 6417.187 0.025 0.876Clip*Removal 62280.021 1 62280.021 0.241 0.628Fert*Clip*Removal 83333.33 1 83333.33 0.322 0.576Error 6210177.17 24 258757.4Within treatmentsTime 6971265.9 5 1394253.2 103.019 0.000Time*Fertilizer 109268.0 5 21853.6 1.615 0.161Time*Clipping 187830.2 5 37566.0 2.776 0.021Time*Removal 68154.4 5 13630.9 1.007 0.417Time*Fert*Clip 25556.1 5 5111.2 0.378 0.863Time*Fert*Removal 41453.5 5 8290.7 0.613 0.690Time*Clip*Rernnial 63123.5 5 12624.7 0.933 0.462Time*Fert*ClipRem 15232.5 5 3046.5 0.225 0.951Error 1624067.8 120 13533.949Bud density showed a serial correlation in time (Fig. 1) and was analyzed separately using arepeated measures MANOVA (one-way and three-way). No significant treatment effects weredetected in either the one-way or three-way analysis, however, the repeated measures componentof these analyses detected a significant within treatment Time effect in both cases (Tables 6, 8).The multivariate test results for both one-way and three-way analyses also detected significantTime, and Time by Treatment interactions (Table 7). In the three-way analysis, the Time byTreatment interaction occurred only for clipping (C) (Table 7). The polynomial test of order(Gurevitch & Chester 1986, Systat 1992) of the within treatment effects for the one-way analysisin 1992 indicated that this effect was significant and linear (P=0.003) (Table 6). In contrast, whenthe extra clipping treatment was removed and the data were analyzed with the three-way design,the polynomial test of order of the within treatment effects showed that the significant Time andTime by clipping effects were quadratic (P<0.037, Table 9). The discrepancy between the resultsfor the polynomial test of order for the one-way and three-way analysis is unusual. These resultsindicate that when the intense clipping treatment is incorporated into the experimental design, thewithin treatment repeated measures effect is best described with a linear model.Table 9: Source of variation for the polynomial test of order (quadratic) for three-wayrepeated measures analysis of bud production in 1992.Source dF MS F-ratio ProbabilityTime 1 3858099.3 143.926 0.000Time*Fertilizer 1 41125.1 1.534 0.227Time*Clipping 1 130510.8 4.869 0.037Time*Removal 1 0.595 0.000 0.996Time*Fert*Clip 1 13626.0 0.508 0.483Time*Fert*Removal 1 165.0 0.006 0.938Time*Clip*Removal 1 52417.0 1.955 0.675Time*Fert*Clip*Rem 1 13464.4 0.502 0.485Error 2450STANDING CROP AVAILABLEStanding crop available in 1991 and 1992 calculated as live summer leaf-days (C1) issummarized by treatment in Figure 2. In 1991, clipping and the interaction between fertilizer andneighbour removal had significant effects on live summer leaf-days (C1) (P=0.05, Table 10).Clipping in 1991 increased mean C (±SEM) to 5421.5 ±420.22 relative to the control of 5131.0±555.74. The interaction between fertilizer and neighbour removal was such that C1 was greatestin the control (no fertilizer or neighbour removal). However, the reduction in C1 as aconsequence of treatment was less when fertilizer was combined with neighbour removal, thanfertilizer alone, or neighbour removal alone (Fig. 2).In 1992, the only significant effect on standing crop available detected was fertilizer(P=0.005, Table 10). Mean (±SEM) live summer leaf-days for the fertilized treatment was4516.25 ±734.11 relative to the control that was calculated as 8075.25 ±216.56 (Fig. 2). In 1991,live summer leaf days violated the assumption of normality in this analysis.51800 -___________________________________ __________________• CONTROL700 -• F600-5OO- A NR400- . C0300- o C+200 -100 -0- I I I I0 10 20 30 40 50 60 70 80Time (days)800 -______________ _______* F-C-NR700 --I- F-C500- C-NR400- * F-NR300- • CONTROL200 -100 -(1-I I I I0 10 20 30 40 50 60 70 80Time (days)Figure 1: Bud production I m2 over time by treatment in 1992.Time 0 is the date of the first survey (June 2), which wasapproximately 7 days after snowmelt and 3 days after treatments(except clipping) were first applied. The control has been shownin both panels for comparative purposes.100000 52C,, • 19911992100000 1L) 9L)0Figure 2: Mean (±SEM) live summer leaf-days (Ci) bytreatment and year.Table 10: Source of variation table for a 3-way ANOVA of log transformed live summer leafdays (Ci) by treatment in 1991 and 1992.Variable Treatment dF MS F-ratio Probability1991 Fertilizer 1 0.247 3.511 0.074Clip 1 0.3 14 4.476 0.045Removal 1 0.263 3.753 0.065Fert*Clip 1 0.001 0.007 0.933Fert*Rem 1 0.403 5.738 0.025Clip*Rem 1 0.050 0.7 17 0.406Fert*Clip*Rem 1 0.002 0.024 0.879Error 24 0.0701992 Fertilizer 1 1.943 9.593 0.005Clip 1 0.013 0.062 0.805Removal 1 0.310 1.531 0.228Fert*Clip 1 0.210 1.037 0.319Fert*Rem 1 0.314 1.549 0.225Clip*Rem 1 0.241 1.192 0.286Fert*Clip*Rem 1 0.069 0.34 0.565Error 24 0.20353DISCUSSION“The growth and development of most plants depends on the accumulation of reiteratedelements (e.g. leaves, shoots and flowers) ... Because reiterated elements possess demographicproperties like natality and mortality, plants can be considered as metapopulations (sensu White1979) of parts.” (Maillette 1992). In designing plant population studies, some problems areencountered in determining the unit of census (leaves, stems, shoots, buds), as plant architectureis quite variable both within and between taxonomic groups (reviewed in White 1979). Maillette(1992) suggests a number of criteria for selecting census variables based partly on their“demographic competence”, (i.e. their ability to satisfy the equation N t+1 = N t + B-D, andproduce daughter units). Although leaves, as demographic units, do not directly producedaughters, they are still used because of their value in determining aspects of plant growth(Harper 1989)Although the number of demographic studies done on plants are accumulating rapidly,many of these are comparative or descriptive studies of dynamics in the field or greenhouse(Sarukhán , Harper 1973; Sarukhán 1974; Barkham 1980; Ford 1981; Hartnett , Bazzaz 1985).Relatively few studies monitor population dynamics in response to experimental manipulations,and even fewer study clonal perennial species (but see Hawthorne, Cavers 1976; Bazzaz , Harper1977; Noble et al. 1979; Jonsdottir 1991). The lack of studies available for rhizomatous perennialsis partly attributed to the difficulty in identifying genetic individuals, and in the case of somespecies such as L. arcticus, difficulty in identifying ramets or other modular units withoutexcavation. This study of the response of L. arcticus to treatments by-passed these problems bymeasuring populations of leaves per unit area. Other studies have measured the response ofplants using the demography of leaves, but this new approach to measure the demography ofpopulations is advantageous in situations where the identification of individuals (ramets or genets)is not feasible without destructive sampling. Although this method is not capable of detectingchanges in ramet populations, Maillette (1992), in a comparative study of the plasticity between54different levels of modular reiteration in Potentilla anserina L., reported that the dynamics of thesmallest units of reiteration (leaves) were more sensitive to treatments (fertilizer), than eithermodules or stolons. The different levels of modular reiteration compared in Maillette’s study areconsidered to be hierarchical units of reiteration where the smallest units are compounded to formlarger units (White 1979). The results of Maillette (1992) suggests that while this method ofmonitoring leaf demography is not without flaws, it may have merit in some unique cases, as forL. arcticus.LEAF DEMOGRAPHYBecause no correlation in time was detected for most variables in this experiment, theanalyses of leaf demography variables in response to the treatments were done in such a way thatthese data were pooled over time and analyzed globally by treatment. The final result showed thatcomparing leaf demography as a metapopulation (sensu White 1979) of leaves per unit area wassensitive enough to detect treatment effects in this experiment.In 1991, fertilizer and clipping both significantly affected incidence of disease occurring inthe populations. Clipping reduced the incidence, whereas fertilizer increased it. The reductionassociated with clipping is quite surprising as it is often reported that herbivore damage increasessusceptibility to disease due to their acting as vectors, or by increasing the vulnerability of tissuesto infection after herbivore damage (Crawley 1988, Baldwin 1990). The reduction of diseasemay be the consequence of reduced crowding of leaves in the clipped population decreasing therate of spread of disease between remaining leaves. However, under conditions of naturalherbivory, parts of leaves or stems may be removed and the pattern of thinning may be quitedifferent from the experimental clipping. In addition, incidence of disease may have been greaterin the taller leaves so that the clipping of leaves may have purged the population of disease tosome extent. In any case, reduction in incidence of disease in clipped plants agrees with Owen’s(1981) supposition that herbivores may benefit plants by selective removal of unproductive tissue.55The higher incidence of disease detected in the fertilizer treatment may indicate that pests preferto infect rapidly growing tissue, or tissue with higher nutrient concentrations. Alternatively, theycould be preferentially selecting tissue that has a higher turnover rate. Very little is currentlyknown about the role of disease in plant population dynamics, and it has not been clearlydemonstrated that disease is capable of affecting a planCs fitness, or what factors are important indetermining rates of infection in hosts (Alexander 1992). It has been suggested that disease onlybecomes an important mortality factor once the plant is afready under stress from other factors(i.e. herbivory, Crawley 1988). Wennstrom , Ericson (1991) have also shown that Pulsatillapratensis, a perennial dicot occurring in Sweden, has more vigorous vegetative growth andsurvivorship when it is infected by Puccinia pulsatillae, a sterilizing rust fungus. Diseased plantswere also reported to be less susceptible to grazing by cattle or leaf-eating insects.In 1991, total leaf mortality increased when neighbours were present, indicating thatcompetition is occurring in this environment. Fertilizer marginally increased leaf mortality in1991. This effect has been shown in other fertilizer studies (Noble et al. 1979; Shaver 1983) andhas frequently been attributed to higher probability of leaf death (turnover) in rapidly growingtissue (Harper 1977). More specifically, fertilizer may increase leaf mortality by hastening theonset of competition for light or other factors. Shaver (1983) also proposed that the cost ofmaintaining old leaves may outweigh their advantages for nutrient storage, and old leaves areallowed to die. Maillette (1992) in a greenhouse experiment using Potentilla anserina also foundthat increased nutrient levels increased leaf mortality as measured by leaf death ratio (number ofdead leaves/total number of leaves).In 1992, clipping again reduced the incidence of disease in the population, but no effectwas detected for fertilizer despite the increased application rate. As in 1991, no treatment effectswere detected for leaf mortality in 1992. This suggests that variability in other environmentalfactors may interact to determine the relative importance of the factors tested. For instance,Eriksson (1986), in a study of Potentilla anserina on a Baltic seashore, reported that population56dynamics were mostly driven by external factors such as poor weather that synchronized birth anddeath of ramets and modules. Climatic factors may be interacting with the experimentaltreatments to produce variability between years.In both 1991 and 1992, fertilizer and neighbour removal treatments were shown not tosignificantly affect the overall population density of leaves. This at first appears surprisingbecause both treatments should make more resources available to the target plants. However, L.arcticus living in this habitat experiences many of the characteristics described by Grime (1979) asbeing a high stress environment- low light below the canopy, low fertility soils, and low annualprecipitation (35 cm falling mostly as winter snow). Thus, we might predict that L. arcticus is astress tolerator (sensu Grime 1977) and should show little or no response to short term increasesin nutrients or reduced competition.The production of leaf buds which is an estimate of birth rate of new leaves was analyzeddifferently in 1991 than in 1992, because a correlation in time was detected in bud density in1992. In spite of the different analytical techniques, no significant treatment effect was detectedin the birth of new leaves (buds) in either year. Although clipping did reduce the overall leafpopulation, and presumably the leaf surface area in both 1991 and 1992, this was not shown tosignificantly influence the ability of L. arcticus to produce new leaves. That L. arcticuspopulations are able to compensate for clipping and maintain the same birth rate of new leaves.This may indicate that L. arcticus has evolved to withstand periodic episodes of intense herbivoryduring hare peak. Without this attribute, L. arcticus could be driven locally extinct. It isgenerally believed that some plants may be capable of compensating for tissue lost to herbivores ifthe tissue removed was relatively unproductive (i.e. respiration costs > photosynthesis), or if therewere sufficient reserves available to regenerate the lost surface area. Chazdon (1991), in a 3-yearclipping experiment on an understory clonal palm (Geonoma congesta) removed both rametsand/or leaves and reported little response to ramet removal or defoliation in terms of mortality,stem and height growth, rate of ramet production or reproduction. This study concluded that57Geonoma congesta’s resistance to repeated defoliation and ramet removal was likely due to themobilization of stored reserves to maintain normal patterns of growth.There is some question as to the reason why bud density in 1992 was found to have acorrelation in time when the other variables did not, and why this effect was not detected in 1991.The repeated measures MANOVA performed on bud density in 1992 showed a significant timetrend and a significant time by clipping interaction although no significant treatment effect wasdetected for clipping. This time trend was either linear or quadratic depending on whether theanalysis included the elevated clipping treatment. Although the biological significance of theseresults is uncertain, perhaps this is the first indication of a treatment effect of clipping on buddensity that was not previously detected due to some lag in the response of L. arcticuspopulations.STANDING CROP AVAILABLEIn 1991, live summer leaf days (C1), an index of the amount of leaf matter available toherbivores through the season, was greatest when populations were clipped, and there was asignificant interaction between fertilizer and neighbour removal such that the reduction in standingcrop available when fertilizer was applied was partly ameliorated if neighbours were removed.The result obtained for clipping was rather unexpected as clipping reduced overall leaf density inL. arcticus populations. The higher standing crop in this treatment suggests that lower turnoverof leaves is occurring when populations are clipped (Louda 1984). If this were true, it may be amechanism by which the plants maintain leaves longer to compensate for reduction in leaf surfacearea when clipped.The significant interaction between fertilizer and neighbour removal suggests that thepresence of neighbours is mediating the access of L. arcticus to the nutrients applied. Fastergrowing neighbouring species may use the nutrients before L. arcticus (Noble et al. 1979) so that58the removal of neighbours allowed L. arcticus greater access to the nutrients and resulted inincreased standing crop available when neighbours were absent.In 1992, no interaction between fertilizer and neighbours was detected, however fertilizerdid reduce standing crop relative to the control. If neighbours are limiting the access of L.arcticus to the nutrients applied, it is not known why an interaction did not occur in 1992 as in1991. However, the neighbour removal treatment only manipulated the abundance of herbaceousspecies, and does not account for competition from trees and shrubs. In 1992, trees and shrubswhich were slow to respond in 1991 may have usurped nutrients to the detriment of allherbaceous species including L. arcticus.Although it was not possible to assess the effect of treatments on ramet or genetpopulations of L. arcticus, it may be possible to speculate that factors affecting metapopulationsof leaves may be translated to ramet and genet dynamics if the treatments are maintained for asufficient period. Maillette (1992) reported that compound reiterated units (modules, stolons)that are composed of several smaller units (leaves, buds, roots) were ‘demographically buffered’and slower to respond to treatments. Therefore it might be expected that populations respondingto changes in environmental factors would respond initially with changes in the smallest units ofreiteration (leaves) and these changes would then be transmitted to higher levels in the reiterationhierarchy (sensu Maillette 1992). So changes in the birth rate or death rate of leaves shouldeventually appear as births or deaths in the ramet population.CHAPTER 3LEAF COHORT SURVIVORSHIPIn 1992, a leaf cohort survivorship study was initiated as it was not possible to assesspatterns of leaf or ramet survivorship in the main experiment. This experiment was designed todetermine if the treatments had different effects on patterns of leaf survivorship, and if thesurvivorship curves differed qualitatively or quantitatively. Quantitative differences insurvivorship would indicate that the pattern of mortality between treatments was the samewhereas the rate of mortality differed. Qualitative differences in leaf survivorship indicate that thepattern of risk of mortality over time is different between treatments (i.e. the shape of thesurvivorship curves is different).METHODSIn 1992, a single subpopulation of lupine leaves was chosen within the main experimentalquadrats. Each subpopulation was chosen to incorporate the highest density of lupine leavesthat could fit in a 25 cm X 25 cm (625 cm 2) area, to ensure an adequate population size. Thesubpopulation was permanently marked in each quadrat with a plastic-coated wire quadrat, 25 cmon a side. The first cohort was tagged on June 7, 1992, the second cohort on June 24, 1992.Leaf cohorts were identified by placing one drop of Testor’s model paint on a leaflet. Differentcoloured paints were used for two cohorts. These cohorts within the subpopulations of leaveswere successfully monitored over the course of the 1992 season. Attempts at marking latercohorts were not successful due to poor weather conditions. Prior to using the paint, field testswere performed in Vancouver, B.C. and there was no indication of damage or reduced leaf orplant survival with this method. Paint tags had a high degree of permanency under variable5960Vancouver weather conditions. Cohorts were surveyed five times through the season: June 7,June 24, July 10, July 28, and August 19, 1992.STATISTICAL ANALYSISBetween Treatment SurvivorshipComparison of survivorship between cohorts was performed to determine if survivorshipdiffered depending on birth date. Two methods were used to analyze the survivorship data in thisexperiment and showed different results. The first involved fitting a parametric model tosurvivorship data and then comparing model parameters between treatments and cohorts. Thesecond method was the traditional nonparametric rank sum procedure used for ecologicalsurvivorship experiments. Two analytical methods are incorporated for comparative purposes.Cohort life tables were not constructed as part of this analysis because the data did not meet lifetable requirements described in Lee (1980, 1992). According to Lee (1980, 1992), cohort lifetable analysis cannot contain right-censored data, so that all cohorts must be followed until 100%mortality occurs. Second, Lee (1980, 1992) states that a large number of censuses must beperformed so that data can be grouped into intervals. She suggests that a minimum of 10intervals is adequate for good analysis. Since this data set does not meet either of these criteriaanother method was sought. Life table analysis is also a non-parametric method and is not aspowerful as parametric methods that are available because it is not capable of accounting forvariance between replicates in the analysis. Most life table analyses are performed onunreplicated populations; this cohort study was replicated (n=4) so parametric methods werepreferred.PARAMETRIC MODEL FITTINGThe results from the leaf cohort survivorship study were analyzed using SAS Proc Liferegand Proc Nun (SAS 1979) to select and fit an appropriate theoretical survival distribution model61to the data. The Proc Lifereg module was used initially to select a survival distribution to fit thedata. Both the Weibull distribution and the log-logistic distribution had good fit, but the predictedvalues calculated by quadrat indicated that the Weibull model showed the better fit. The SASProc Nun module was then used to calculate the parameters for the Weibull model for eachquadrat and cohort individually. A complete description of the Weibull model is in Appendix A,and a brief description follows.WEIBULL MODEL DESCRIPTIONThe survival curve generated by the Weibull distribution is modelled with the followingequation:aS(t)=lOO * e .(*age) (1)S(t) is the probability of survivorship at time or age (t), lambda (2) is the scale parameter, andalpha (a) is the shape parameter.The values for ? and a that were calculated by the nonlinear model fitting module in SAS foreach cohort and quadrat were used as the test statistics to compare treatment effects on leafcohort survivorship. A one-way ANOVA of mean parameter estimates for each treatment andcohort was used to determine if the experimental treatments produced significantly (P<O.05)different patterns of leaf survivorship between the first (emerged June7) and the second (emergedJune 24) cohorts. A one-way ANOVA was used in place of a factorial design to accommodatethe additional clipping treatment in 1992. Both 2 and a were log transformed prior to analysis tomake the data homoscedastic.62NONPARAMETRIC STATISTICSThe nonparametric log rank procedure used for multiple comparisons of the leaf cohortsurvivorship is described in Pyke , Thompson (1986) and amended in Pyke , Thompson (1987).The null hypothesis tested in this analysis is that the survival rates of different populations (i.e.treatments) or cohorts do not differ (Hutchings et al. 1991). The log rank test statistic is basedupon a chi square distribution with degrees of freedom equal to one less than the number ofgroups (treatments) compared. The method used here tests for differences between unreplicatedpopulations. Survivorship data were collected on replicated subpopulations of leaves. Tocalculate the log rank test statistic, the subpopulation data for each cohort were pooled, and themeans of each treatment were used to calculate rank in accordance with Pyke (1993, personalcommunication).RESULTSWEIBULL MODEL ANALYSESThe shape (a) and scale () parameters generated by the Weibull model are summarized bytreatment for each cohort in Table 11. Analysis of variance for cohort 1 detected no significanttreatment effects for either 2. or a (Tables 13, 14). Analysis of the cohort 2 parameters showedthe scale and the shape parameters to be highly significant (P=O.OO1, Table 12, and P=O.028,Table 13), indicating that survivorship in the younger cohort was highly dependent on thetreatment imposed. A post hoc Tukey comparison of the scale parameter in cohort 2 detected asignificant difference between the intense clipping treatment (C+) and the control, fertilized,neighbour removal, fertilizer by neighbour removal and clipping by neighbour removal groups.Tukey comparison of the shape parameter in cohort 2 detected significant differences betweenclipping and fertilizer, and between clipping and fertilizer by neighbour removal treatments.Survivorship in the clipping treatments decreased more rapidly than other treatments (Fig. 3).63Table 11: Scale (2k) and shape (ct) parameter estimates for the Weibull Distribution (1992).± SEM ±SEM ± SEM I ± SEMTable 12: Combined source of variation for the one-way ANOVAs for cohort 1 and 2 comparingthe log transformed ? (scale) Weibull parameter by treatment.Cohort 1 Cohort 2TreatmentControlFertilizedRemovalClippingFert/ClipClip/RemFert/RemFert/Clip/RemExtra clip0.0 180.0200.0210.0200.0190.0200.0180.0250.02 10.0020.0020.0060.0060.0020.0020.0020.0030.0011.9371.9 172.1061.3691.3591.7511.7281.5761.9600.3340.5070.24 10.1120.1370.4020.2740.2200.0790.0130.0130.0140.0530.1850.0170.0130.0370.4390.0030.0030.0040.02 10.1520.0020.00 10.0070.2442.2552.1462.1460.3770.8461.1012.4451.4071.0590.7640.4580.6700.1310.3560.3220.5170.5680.577Cohort Source SS dF MS F ratio Probability1 Treatment 0.347 8 0.043 0.5 10 0.838Error 2.297 27 0.0852 Treatment 33.776 8 4.222 3.896 0.001Error 23.285 27 0.862641-.0C4-a)0a)0Time (days)• Control -I- FC * FNR CNR X FCNRFigure 3: Leaf survivorship curves for cohort 1 and cohort 2 by treatment in 1992.F=fertilized, C=clipped, NR=neighbours removed, C+=extra clipping.10.1 -0.01 -10 20 30 40 50 60 70 800.10.01 -110 20 30 40 50 60 70 80• Control • F A NR • C D C+11-Cohort210 20 30 40 50 60 70 80 0.1- 20 30 40 50 60 70 80Time (days)0.10.0165Theoretical Survival CurvesIn addition to determining the effect of treatment on survivorship, parameters generated byfitting the Weibull model to survivorship function can convey pertinent biological information byfitting the experimental data to a theoretical model of survivorship (Begon & Mortimer 1986).Certain combinations of and a can be used to generate Type I, II, and III survivorship curves(see Appendix A, Figure 9). When ? is constant, an a greater than 1 produces a Type III curve;a equal to 1 is a Type II curve; a less than 1 is a Type I curve. Examination of the mean and ccparameter estimates (Table 11) indicates that a Type III survivorship could describe leafsurvivorship for both cohorts. A Type III curve is characterized by a high rate of early mortalitythat declines as survivors age. Not all treatments show an obvious Type ifi curve. Mean a(shape) parameter estimates by treatment and by cohort show variability that could generate allthree of the survivorship models. In the first leaf cohort, all a estimates are greater than one,however in cohort 2, the values for different treatments range from 0.377 to 2.445 (Table 11).Between Cohort SurvivorshipAfter the analysis of survivorship curves by treatment within cohort detected no significantdifference (Table 12, 13) , an a posteriori decision was made to compare survivorship betweencohorts within treatment groups. Two-way ANOVA results of the log transformed shape andscale parameters indicated that only the shape of survivorship curves showed a significantbetween cohort effect (P=O.003, Table 14).NONPARAMETRIC ANALYSESIn contrast to the parametric model, the nonparametric analysis of differences insurvivorship between treatments for cohort 1 and cohort 2 detected a significant treatment effectfor both cohorts at P<O.000 (Table 15). The nonparametric log rank test does not permitpairwise comparisons to ascertain where the differences between treatments lie. However,66inspection of the data suggests that with the exception of the clipping treatments, survivorship ofcohort 2 may be slightly greater than cohort 1.Table 13: Combined source of variation table of one-way ANOVAs for cohort 1 and 2comparing the log transformed shape (ct) parameter by treatment.Cohort Source SS dF MS F ratio Probability1 Treatment 0.736 8 0.092 1.030 0.438Error 2.412 27 0.0892 Treatment 14.124 8 1.765 2.634 0.028Error 18.096 27 0.670Table 14: Combined source of variation table for two-way ANOVAs testing for the effect ofcohort within treatment on log transformed shape and scale parameters.Parameter Source SS dF MS F ratio ProbabilityLambda (?.,) Cohort 1.556 1 1.556 3.284 0.075Treatment 18.214 8 2.277 4.806 0.000Cohort*Treatment 15.909 8 1.989 4.198 0.001Error 25.583 54 0.474Alpha (x) Cohort 3.596 1 3.596 9.464 0.003Treatment 9.446 8 1.181 3.109 0.006Cohort*Treatment 5.414 8 0.677 1.782 0.101Error 20.508 54 0.380Table 15: The nonparametric log rank statistics testing for between treatment effects incohort 1 and cohort 2 leaf survivorship curves.Cohort Log rank test dF Chi square critical Probabilitystatistic value1 257.832 8 15.507 0.0002 701.716 8 15.507 0.00067DISCUSSIONLittle is known about the causes of mortality in plants in the field and why species andpopulations show different patterns of survivorship at different sites and environmental conditions(Hutchings et al. 1991). As a result, studies investigating how growth, reproduction, andmortality of populations respond to different environmental factors are still necessary until we areable to explain with some degree of certainty how and why populations change over time.Understanding how patterns and rates of mortality are influenced by time, age, and environmentalfactors allows us to predict population change when combined with birth, immigration andemigration statistics. Most studies of population survivorship focus on two questions. (1) Doesthe proportion of individuals reaching age x differ in one population relative to other populations?(2) Does the average lifespan of an individual differ between populations? (Pyke & Thompson1986). This study is primarily concerned with the question of longevity of leaves under differenttreatments which depends on the shape of the survivorship curve in a leaf population (Pyke,Thompson 1986). Ideally, we would follow cohorts of seedlings. However, in 1992, there wereinsufficient seedlings (genets) of L. arcticus established and consequently cohorts of leaves wereused.According to Harper (1977), “a leaf has a life history, a changing pattern of behaviour frombirth ... to death from senescence or some environmental hazard” (p. 23). As such, “for somepurposes the population dynamics of plant parts may be more useful than the dynamics of wholeplants in a community” (p. 21). This view is partly based on the presumption that predictingpopulation changes in plant communities may not come until we understand how plants respondto changes in their environment at all levels of organization (i.e. leaf, rarnet and genet) becausechanges at lower organizational levels (i.e. the leaf) ultimately translate to higher levels oforganization. From an ecological standpoint, studies of leaf lifespan are important because leaf68longevity determines the duration of soil coverage, shading of neighbours, rainfall interceptionlosses (i.e. from transpiration or evaporation before it hits ground), soil moisture depletion,microbial and herbivory life (diversity and abundance) in the canopy, and air pollution sensitivity(Diemers et al. 1992). With a modular framework of plant organization in mind, patterns ofsurvivorship between treatments were assessed at the level of the leaf by comparing survivorshipbetween leaf cohorts of different ages.EFFECTS OF CLIPPINGIn this study, clipping had the strongest effects on survivorship. This effect was greatest incohort 2, the youngest leaf cohort, with little detectable effect in cohort 1. Clipping at the normalor more intense level caused leaf death because leaves were removed as a part of the treatmentprotocol. However, all differences were not wholly due to clipping because not all treatmentsexperiencing clipping showed the same survivorship pattern. The more intense clipping treatmentcaused a more rapid rate of decline in survival in cohort 2. When clipping was combined withother treatments there was greater survival than treatments of clipping alone. Several factors maybe responsible for higher levels of mortality in the clipping treatments. This could be the result of(1) the increased likelihood or incidence of disease following tissue removal, however incidence ofdisease was shown to decline with clipping in chapter 2, (2) decreased competitive ability withneighbours due to subordination in the canopy, (3) the reduced ability to defend againstsubsequent herbivore attack, or (4) reduced ability to fix nitrogen because of reduced carbonfixing ability. Clipping in combination with neighbour removal or fertilizer may have mitigatedsome of the negative effects associated with the loss of leaf surface area or tissue damage.Louda (1984) studied the effect of herbivory on leaf dynamics and reported that reducedlevels of insect damage increased the survival of mature leaves and increased initiation of leaves.Similarly, clipping in my study did produce a notable decrease in leaf survival, but not in allclipped treatments. Chabot & Hicks (1982) in a review of leaf lifespans suggested that the impact69of leaf removal depends on the amount, timing, nature of damage and local environmentalconditions. Interaction of clipping with fertilizer and/or neighbour removal alleviated the effect ofclipping on leaf survivorship in my study. According to Chabot & Hicks (1982), vulnerability tothe effects of clipping may vary with the stage of development a leaf has reached when damageoccurs. In my experiment, clipping was performed on the same date for both cohorts, but atdifferent ages, and this difference may explain why clipping effects were not seen in both cohorts.EFFECT OF FERTILIZER AND REMOVING NEIGHBOURSFertilizer and removing neighbours did not have a significant effect on leaf survivorship.This lack of significant response was surprising because both of these treatments have thepotential to strongly influence leaf longevity. For instance, removing neighbours increases theamount of light available for photosynthesis and reduces shading. This could change a leaf from anet drain or a sink to a source of energy. Application of fertilizer in other studies changes growthrates which could in turn change patterns of leaf turnover and mortality. Several studies detail theability of light and nutrients to affect leaf properties including longevity, dynamics and nutrientcontent. Bazzaz & Harper (1977) reported that both light and density influenced leaf survivorshipin Linum usitatissimum L. They also showed that experimentally withholding nutrients, removalof flower buds, and shading lower leaves all influenced the onset of leaf mortality although thenutrient treatment had the strongest effect. Shaver (1983), reported leaf lifespan decreased inLedum palustre when fertilizer was applied. Fox & Morrow (1992) reported that fertilizer didchange leaf properties and increased foliage nitrogen and phosphorus content in Eucalyptus spp.although no change in growth was observed. Diemers et al. (1992) studied 29 species of forbsand grasses and detected a strong correlation between leaf duration and nitrogen content. Theyalso noted that in general leaf turnover increased with increasing competition for light, and withincreasing vigour of growth.70COMPARISON OF COHORTSA difference in survivorship between cohorts and within treatments was noted in this set ofexperiments. In the parametric analyses of survivorship, no treatment effects were observed in thefirst leaf cohort, including no observable clipping effect. This may be due to morphologicaldifferences in leaves between cohorts. Lupinus arcticus was observed to have small scale-likeleaves that never grew larger than 2- 3 cm long over the growing season. This heterophylly maybias the results of this analysis if the appearance of these leaves within cohorts is not random.These smaller leaves would never be at risk of clipping because the treatment is imposed using aheight criteria. This may be why no clipping response was seen in the first leaf cohort. A similartype of heterophylly has been reported for Pseudopanax crassfolius, a sub-tropical tree, wherethe oldest leaf cohort is scale-like and small relative to subsequent leaves. These leaves arebelieved to be formed in the previous growing season and remain dormant, only to appear at thebeginning of the next season (Clearwater & Gould 1994). Even if the differences betweencohorts in lupines is not due to these scale leaves, differences between cohorts in the size ofleaves, or position in the herbaceous canopy may produce different patterns of survivorship.Alternatively, the second leaf cohort may respond to treatments differently for otherreasons. Normally, if no differences in survivorship are observed between cohorts, patterns ofleaf mortality are presumed to be determined by leaf age or leaf stage. Sydes (1984), comparedbetween cohort survivorship among several species of herbs in lime grassland and reported thatearly leaf cohorts in many species grew rapidly and sustained higher rates of mortality than latercohorts. In Sydes (1984), the differences between cohort survivorship may be determined byother environmental factors or seasonal effects, rather than factors related to leaf age or stage. Ifthis is true, the higher mortality reported in the first cohort would be determined more by externalfactors such as timing rather than internal factors influenced by leaf age.Other authors studying leaf demography in herbs have also compared cohort survivorshipamong different species. Diemers et al. (1992) followed leaf cohorts of 29 herbs and showed for71many species that early leaf cohorts experienced higher mortality than later cohorts. This wasalso reported for Solidago spp. by Schmid et al. (1988). As mentioned earlier, Diemers et al.(1992) correlated leaf lifespan to nitrogen content in leaves. The authors surmised from theirstudy that leaf longevity was not determined by efficiency of light harvest, but ratherenvironmental factors such as mechanical strength, herbivory, and pathogens.Examination of patterns of variation present in the shape parameter (o) with respect to thetheoretical survival curves they generate produced enough variation present in cohort 2 toproduce Type I, II or III curves depending on the treatment. These different theoreticalsurvivorship models have radically different implications for the dynamics of populations. Thissuggests that further investigation is required to determine if these treatment effects on patterns ofleaf survivorship are real or merely the result of the highly variable nature of field experiments.In spite of the need to understand patterns and causes of mortality in ecology, differences insurvivorship between cohorts and populations are not often analyzed in ecological journals (Pyke& Thompson 1986). Therefore, most methods available to analyze survivorship data cannot copewell with ecological experiments. Rather they were developed for clinical studies, or often forfailure time studies for equipment manufacturers (Lee 1980; Hutchings et al. 1991). Until bettermethods become available for analyzing this type of data, we are left with rather crude methods toassess differences that may or may not be ecologically meaningful.In choosing the appropriate method a number of factors were considered including thenature of the data collected, right censoring of the data set, replication of data, etc. Thisexperiment was designed to generate life table data but was later analyzed by fitting a theoreticalsurvival distribution to the observed data and using the calculated parameter estimates as the teststatistics. Gehan (1969) suggested this as an appropriate method if some survival data areavailable for populations, but the true theoretical distribution is not known for various reasons.This method is not able to generate the same descriptive statistics as the life table method,however it is the only method available that can handle censored and replicated data.72Some of the drawbacks of this method with regards to this study in particular are that it isnot possible to determine with certainty that the theoretical model selected is the most accuratedescription of the population’s survival curve if the survivorship data are censored. Specifically,to say with certainty that the theoretical model chosen to describe the data is the best model, allindividuals or cohorts must be monitored to determine the exact time of mortality. That is, alltreatments must be monitored until 100% mortality occurs. That is relatively rare in biologicalstudies where the resources are not available to wait until all subjects have died. Thereforemethods have been developed to deal with censoring in data sets. With regards to analyses in thisstudy, it can oniy be said that the parameter estimates generated, and the model selected, areappropriate for the time frame over which this study occurred. Therefore, it can not be concludedthat these parameter values generated for the Weibull model are true population estimates.An additional problem in a study this large is that the number of survivorship curvesavailable for comparison is so large that it is not conducive for fitting a large number of differenttypes of theoretical distributions. Due to time and computing constraints, I was confined to those Cdistributions available in SAS Proc Lifereg. Also because the treatment groups in this study havethe potential to have widely different effects on survivorship patterns, using a single theoreticalmodel to describe all treatments and replicates may not have been the best method. Othertheoretical distributions may have fit some treatments better than others. However, it is notpossible to statistically compare parameters of different distributions. With regards to this study,this may or may not have been a problem because of the selection of the Weibull model in thisanalysis. Specifically, the Weibull model, as a two parameter model that allows the hazard rateh(t) to vary in simple to complex patterns, is quite flexible in fitting a variety of linear and nonlinear survival curves described previously.The results of some of these survival analyses were able to detect a clipping effect onsurvivorship as well as differences between cohorts within treatment in spite of the limitationsassociated with the methods used. However, further investigation on the effect of these73treatments as limiting factors on patterns of survivorship is warranted. In the future, it would beadvantageous to modify this study to incorporate additional leaf cohorts, as well as monitor leafcohort mortality to 100% mortality in all cohorts. More cohorts monitored would also help sortout the nature of differences in survivorship between cohorts that this study only alludes to, andwould eliminate the problems associated with analyzing survivorship in right censored data.Further studies comparing the effects of different types of limiting factors on patterns ofsurvivorship may wish to consider more sophisticated (and more complex) analyses where thetheoretical distribution that best fits individual treatments is determined for all treatment groups.Once a set of theoretical models have been selected, a comparison that explores biologicalmechanisms/causes that produce qualitatively different survivorship models may be in order. Thiswould likely involve assessing causes of mortality in the field (i.e. predation, disease,shading/competition) throughout the experiment.Traditional non-parametric log rank tests indicated significant treatment differences insurvivorship in both cohorts relative to the parametric analysis. The non-parametric methods arenot considered to be as powerful as parametric methods, however, this is the most commonmethod to analyze survivorship in ecological studies, and even, the parametric method used in thisanalysis is also not without problems.CHAPTER 4VEGETATIVE GROWTHIndividual plants may respond to changes in the environment by altering reproduction andgrowth. The variable used to measure plasticity in response to the experimental treatments wasthe length distribution (i.e. size structure) of petioles in the population; petiole length is a measureof vertical growth of petioles in the leaf population. Plasticity of growth was measured at apopulation level because individuals were not identifiable. An additional experiment was done tomeasure the relative effect of different treatments on horizontal space acquisition in L. arcticus.This experiment measured changes in horizontal spread in response to treatments by measuringchanges in percent cover of L. arcticus (and other species) from 1991 to 1992.METHODSSIZE STRUCTUREData were collected throughout the 199 1/1992 growing season using non-destructivemethods to examine the plasticity of size response of L. arcticus. Due to the inability to identifygenets or individual ramets without destructive sampling (excavation and/or harvest), the size ofL. arcticus individuals were measured collectively. This was done by measuring petiole length foreach leaf in all quadrats and then assigning all leaves to one of four size classes based on petiolelength: I) 0-5cm, II) 5-10cm, III) 10-15cm, and IV)> 15 cm. Data in size classes ifi and IV werelater pooled because of the infrequent number of petioles in the tallest class. Size class data wereanalyzed in 1991 and 1992 by constructing multi-dimensional contingency tables. Systat’s loglinear model fitting program was then used to generate Pearson’s chi square goodness of fit valuesfor the treatment factors and their interactions. The null hypothesis in this analysis was that theeffect of fertilizer, neighbour removal and clipping on size distribution were independent. This7475analysis was done once on mid-season data and once on late season data in each year to determineif distribution of petiole lengths changed through the growing season. In 1991, survey dates wereJuly 24 (survey 5), and August 17 (survey 7), in 1992 they were on July 14 (survey 4), andAugust 13 (survey 6). Petiole length is used as an indicator of plant size and has been shown tobe correlated to other plant characters in other species (Evans 1986), and is capable of respondingto environmental change (Birch & Hutchings 1992; Evans 1992).COVERVegetation composition in the experimental quadrats was measured in 1991 and 1992. Thispermitted an assessment of how the abundance of L. arcticus changed relative to changes in theabundance of other species in response to the treatments imposed. These supplementary datawere gathered to help interpret the results of the main experiments. Composition of the herbcommunity in the quadrats was determined using a percent cover estimate using a grid of 10 x 10regularly spaced points at intervals of 10 cm. A vertically placed pin was used as the samplingpoint. Each different species contacting this pin was recorded once at each point. Percent coverof each species was determined by summing the number of times a particular species contactedthe pin in the 100 points occurring within the quadrat. Total cover estimates exceed 100%because the sampling pin often contacted more than one species at a sampling point due tolayering in the vegetation. Mosses, lichens, and fungi were grouped into their own categories.Sampling was done on August 1, 1991 and July 211992.RESULTSSIZE STRUCTURE OF PETIOLES 1991In 1991, the middle and late season analyses (Table 16) indicated that the effect of fertilizer,neighbour removal, and clipping were not independent with respect to distributions of petiolelengths (P <0.000 in all cases). Therefore the null hypothesis was rejected. A source of variationtable summarizes the results of the test of independence for first order interactions of treatment by76size distribution as well as the higher order interactions (Table 16). All interactions from firstorder to third order were significant at P=O.000 for surveys 5 (middle season), and 7 (late season)(Table 16). According to this test, petiole lengths vary in a complex manner according to thetreatment a population receives.Typically, a chi square analysis of independence of categorical variables stops here.Further statistical methods to describe how different variables on a multidimensional table interactare not currently available. However, there are trends in the data (Table 17 a, b), the mostobvious of which, in both surveys, is the interaction between the fertilizer and the neighbourremoval treatment, Initial predictions in this set of experiments anticipated that fertilizer andneighbour removal would have similar effects on plant populations. This is not the case forpetiole length distribution and the pattern of response to these treatments changes between themiddle and the end of the growing season.For survey 5, fertilizer addition appears to produce proportionally greater numbers ofpetioles in the lower size classes when lupines are not clipped (Table 17 a). In survey 7, leafpopulations that have not been clipped differ within the fertilized treatment if neighbours areabsent such that when neighbours were removed, more petioles fall into the smallest size class. Inthe unfertilized, unclipped treatment, this difference between treatments in the presence andabsence of neighbours is not as distinct (Table 17 b).Comparison of clipped vs. unclipped populations in both surveys showed a pattern whereunclipped populations had a greater proportion of petioles in the tallest length class. Thecomplexity of higher order interactions on length distribution of leaves makes it difficult to makeany further conclusive observations without further experimentation. The highly plastic nature ofthe response of petiole length distribution with respect to the three factors is important to note.Similarly, the effect of fertilizer on petiole length distribution in 1991 was significant in spite ofthe low level of fertilizer applied.77Table 16 Chi square test for independence of fertilizer, clipping and neighbour removal on petiolelength distribution using log linear analysis. (year=1991, F=fertilized, C=cipped,N= neighbour removal, D=size distribution)Table 17 a): Length distribution of petioles by treatment reported as frequency and percentoccurring in each class (Survey 5, 1991).Length(cm)0-5%Clipped Not clipped130958.91Neh!hbours+104851.5110353+170648.21Fertilized Not fertilizedTotal 2222 2035 2081 3539 2133 1948. 2505 24945-10%>10%88339.74301.3596747.52200.9878837.871909.13158244.712517.08Clipped I Not clipped134663.1077536.38120.56Neighbours+120461.8173937.9450.2695538.12120147.9434913.93+144457.9085334.201977.90Surveys Survey 7Source dF x 2 Probability dF 2 ProbabilityF*C*N*D 2 38.66 0.000 2 97.58 0.000C*N*D 2 46.55 0.000 2 29.57 0.000F*N*D 2 126.34 0.000 2 150.23 0.000F*C*D 2 54.0 0.000 2 17.6 0.000F*D 2 30.3 0.000 2 40.8 0.000C*D 2 688.71 0.000 2 1082.4 0.000N*D 2 51.04 0.000 2 81.44 000078Table 17 b): Length distribution of petioles by treatment reported as frequency and percentoccurring in each class (Survey 7, 1991).Fertilized Not fertilizedClipped Not clipped Clipped I Not clippedNeighbours NeighboursLength- +- +- + - +(cm)0-5 1281 906 1155 731 1044 825 611 656% 55.17 50.39 51.79 28.16 55.21 52.35 29.45 38.665-10 999 874 887 1440 826 742 1092 741% 43.02 48.61 39.78 55.47 43.68 47.08 52.63 43.67>10 42 18 188 425 21 9 372 300% 1.81 1 8.43 16.37 1.11 0.57 17.93 17.68Total 2322 1798 2230 2596 1891 1576 2075 1697SIZE STRUCTURE OF PETIOLES 1992Length distribution of petioles measured in 1992 was analyzed as in 1991. The chi squaretest for independence of fertilizer addition, neighbour removal and clipping on size distribution ofpetioles for survey 4 (middle season - July 14, 1992) indicated that there was a significant(P<0.000) three-way interaction of the treatments on size distribution. The interaction of clippingand neighbour removal was not significant (P>0.05), however, this may be due to the presence ofhigher order interactions in the data. All other interactions were significant at P<0.0O1 (Table 18).As in 1991, the pattern of response did change during the growing season. Near the end of thegrowing season (survey 6 - August 13,1992), the interaction of clipping and neighbour removalbecame significant at P<0.05 (Table 18).79As in 1991, the complexity of the higher order interactions make it difficult to interpret howdifferent treatments act on distribution of petiole lengths. In 1992, clipping again reduced theproportion of petioles reaching the tallest class for both surveys (Table 19 a, b). The strongestclipping effect was seen in survey 6. The intensified clipping treatment added in 1992 isdesignated as “Extra clip” on the tables and was not incorporated into the chi square log linearmodel fitting because it unbalanced the analysis. The intensified clipping treatment showed theleast amount of growth into the tallest length class with the greatest proportion of petiolesoccurring in the middle class. The clipping regime for this treatment was at a height of 4 cm. Thelarge proportion of leaves growing above this clipping level indicates a rapid re-growth into thissize class following clipping.The interaction of fertilizer and neighbour removal on distribution of petioles is notable inboth surveys in 1992. When fertilizer is applied, petiole growth into the largest class is higherwith neighbours and without clipping than in all other treatment combinations. In contrast, theunfertilized treatment block does not show this pattern. Rather, the greatest amount of growthinto the upper size class occurs without clipping and without neighbours.Table 18: Chi square test for independence of fertilizer, clipping and neighbour removal onpetiole size distribution using log linear analysis. (year 1992, F=fertiized, C=clipped,N= neighbour removal, D=size distribution)Survey 4 Survey 6Source dF X 2 Probability dF I x 2 ProbabilityF*C*N*D 2 131.62 0.000 2 115.56 0.000C*N*D 2 0.7 2 6.22 0.05F*C*D 2 231.44 0.000 2 207.2 0.000F*N*D 2 155.4 0.000 2 27.96 0.000F*D 2 88.05 0.000 2 255.56 0.000C*D 2 518.43 0.000 2 604.14 0.000N*D 2 14.14 0.001 2 50.11 0.00080Table 19 a): Length distribution of petioles by treatment reported as frequency andpercent occurring in each class (Survey 4, 1992).Fertilized Not fertilizedClipped Not clipped Clipped Not clipped Extra________________clipNeighboursLength - +- + - +- + +(cm)0-5 485 377 157 299 260 253 248 161 425% 17.76 19.68 5.64 9.07 9.75 11.44 7.82 5.73 21.775-10 1227 778 1543 1204 1467 1194 1047 1358 1340% 44.93 39.72 55.46 36.53 55.03 54.00 33.02 48.29 68.65>10 1019 761 1082 1793 939 764 1876 1293 187% 37.31 39.72 38.89 54.40 35.22 35.55 59.16 45.98 9.58Total 2731 1916 2782 3296 2666 2211 3171 2812 195281Table 19 b): Length distribution of petioles by treatment reported as frequency andpercent occurring in each class (Survey 6, 1992).5-10%137257.6586850.73144954.4572031.73117661.7092760.0780833.9687647.12CHANGES IN PERCENT COVERThe effect of treatment on the change in percent cover in L. arcticus from 1991 to 1992 issummarized in Table 20. Analysis of these data as another estimate of size change indicates thatonly neighbour removal had a significant effect (P<0.025) on clonal spread in L. arcticus aftertwo seasons of treatment (Table 21), The increase in percent cover in the neighbour removaltreatment was almost five times greater than the control (Fig. 4). Treatments of neighbourremoval in combination with fertilizer or clipping also showed a mean increase in percent coveralmost as great as neighbour removal alone, however, the response was more variable. Cover inL. arcticus showed a nonsignificant increase when fertilized, or when clipped, but not to thedegree that neighbour removal stimulated. The increase in cover in these two treatments wasmuch greater when combined with neighbour removal. Only the fertilizer by clipping treatmentFertilized Not fertilizedClipped Not clipped Clipped Not clipped ExtraclipNeighboursLength- + - +- + - + +(cm)0-5 261 285 196 225 109 75 105 37 88% 10.97 16.66 7.37 9.92 5.72 4.86 4.41 1.99 8.2>10 747 558 1016 1324 621 543 1466 946 45% 31.39 32.61 38.18 58.35 32.58 35.19 61.62 50.89 4.19Total 2380 1711 2661 2269 1906 1543 2379 1859 107394087.682showed a net loss in percent cover of lupines over the study. All other treatments stimulated anincrease in cover in L. arcticus, although only the neighbour removal treatment was significant(Fig. 4).CHANGES IN VEGETATION COMPOSITIONChanges occur in vegetation composition along with changes in L. arcticus (Table 20).Species observed to have low percent cover estimates (<2%) within the quadrats are not reported,and no significant changes were observed for these species during the study period. The greatestchanges in cover were recorded for Festuca altaica, Achillea millefolium and Senecio lugens.The grass, F. altaica, showed a significant increase (29 %) when fertilized (Fig. 4). Neighbourremoval (NR), and the interaction between the fertilizer and neighbour removal (F-NR) also hadsignificant (P<0.0 1) effects on Festuca (Table 20). The significant reduction of vegetation coverfor some species is an artefact of the removal treatment itself and will not be further discussed.This increase may indicate that clipping of Lupinus arcticus resulted in the competitive release ofFestuca. In contrast with Festuca, Achillea millefolium showed reduced abundance underfertilizer. The interaction between fertilizer and neighbour removal seen for Achillea occurred forthe same reason as for Festuca.The abundance of Senecio lugens showed a significant interaction between fertilizer andclipping. The interaction between these treatments was such that when fertilizer was applied,Senecio declined in abundance. However, when lupines were clipped, or clipped and fertilized,Senecio showed a slight increase in abundance (P=0.063, Table 20). Festuca altaica also showeda large, nonsignificant increase (32%, P>0.05) when fertilized and clipped (Fig. 4). Total cover,which is the summation of cover values of all species, did not change significantly except whenneighbours were removed. Solidago canadensis showed a nonsignificant response (P<0.09,Table 20) to clipping and clipping by neighbour removal. Cover of Solidago was reduced relative83to control in the clipping treatment. The interaction between clipping and neighbour removal isan artefact of this species being removed as part of the treatment.Table 20: Summary of mean changes (± SEM) by treatment of percent cover of speciesoccurring in quadrats from August 1, 1991 to July 21, 1992. The significance values from a threeway ANOVA of arcsine transformed percent cover are indicated with * P<0.05,***P<0.001.Species Control Fertilizer Clipped Removal F/NR F/C C/NR F/C/NR(F) (C) (NR)Lupinus 2.5 7.0 4.5 12.25.* 11.5 -5.5 11.0 6.25arcticus ±3.75 ±7.94 ±4.47 ±2.78 ±6.12 ±0.87 ±5.87 ±1.89Festuca 16.0 29.75** 12.0 Ø75*** 75** 32.25 -0.75 -2.25altaica ±4.509 ±3.351 ±4.4 16 ±2.689 ±2.562 ±3.97 ±0.75 ±2.056Linnaea 0.75 0.5 0.50 0 0 -0.25 0 0.0borealis ±0.854 ±0.289 ±0.289 ±0.63 ±0.408Achillea 2.0 0.0 * 2.50 0 0 * -0.75 0 0.0millefolium ±0.913 ±0.408 ±1.555 ±0.48Solidago 1.75 -0.75 0.50 2.75 5.0 0.75 0.25 -6.25canadensis ±1.75 ±0.479 ±0.289 ±1.109 ±2.858 ±1.11 ±0.479 ±6.588Senecio 9.75 -2.5 1.75 0.250 ** -2.0 1.75 * -4.25 -3.25lugens ±3.568 ±1.555 ±3.25 ±3.473 ±0.816 ±1.32 ±2.323 ±1.601Epilobium 1.75 0.25 1.5 -2.5 0 0.5 -0.25 -0.25angustfolium ±1.436 ±0.25 ±2.217 ±0.25 ±0.5 ±0.25 ±0.25Cornis 2.5 4.0 5.75 -2.5 0.25 -2.25 0 0canadensis ±2.843 ±3.028 ±1.436 ±0.25 ±0.25 ±2.25Total Cover 22.79 27.5 -8.25 6.75 *** -8.25 31.5 -3.25 -8.0±11.821 ±18,799 ±7.653 ±6.575 ±8.38 ±2.02 ±7.653 ±4.60184Table 21: Summary of ANOVA results for arcsine transformed changes in percent coverin Lupinus arcticus.Source Sum of dF Mean F-ratio ProbabilitySquares SquareFertilizer 0.006 1 0.006 0.652 0.427Clipped 0.015 1 0.015 1.568 0.223Neighbours Removed 0.053 1 0.053 5.744 0.025Fert*Clip 0.017 1 0.017 1.875 0.184Fert*Rem 0.000 1 0.000 0.000 0.998Clip*Rem 0.001 1 0.001 0.085 0.773Fert*Clip*Rem 0.005 1 0.005 0.583 0.453Error 0.223 24 0.009ChangeinpercentcoverChangeinpercentcover0-r3()--r\)0000000010(110IICD-.0I— ‘DCD-I•CD“0 0-t-t-1 0) CD D -I.ControlFertClipRemovalFert-ClipFert-RemClip-RemFert-Cl-RemControlFertClipRemovalFert-ClipFert-RemClip-RemFert-Cl-RemIIIftI CD00 1186DISCUSSIONPETIOLE LENGTH DISTRIBUTIONCompetitive advantage in plants has been shown to be positively correlated with size (Grace& Wetzel 1981; Soibrig 1981; Meagher & Antonovics 1982; Wolfe 1983). In herbs such as L.arcticus, the advantage conferred by larger size may be due to vertical growth that results inshading and suppression of competitors, or by horizontal spread. In my experiment, patterns ofvertical growth within the population were measured by petiole length distributions, whereaschanges in percent cover of L. arcticus were used to estimate horizontal spread, or the acquisitionof space in response to the treatments imposed. In some species, the ability to penetrate theupper canopy and become a good competitor may be determined by petiole length. Black (1960)compared three varieties of Trifolium subterraneum that differed in petiole length and reportedthat in mixture, the long-petiole variety dominated. Petiole length could be particularly importantfor species such as L. arcticus which have no above-ground support structures with which toobtain canopy position in the herb community except through petiole extension. Alternatively,understory perennial herbs that live in shaded habitats may have little use for vertical growth togain a competitive edge with their neighbours. Instead, competitive ability may be determined byhorizontal space acquisition that usurps space and resources from other species (Hutchings &Slade 1988). In this case, changes in percent cover of L. arcticus may be a better indicator ofgrowth response to the treatments imposed in this experiment.It has long been known that plants respond to stress by varying their number of modules(Harper 1977). Harper (1977) has suggested that understanding the dynamics of plant organs(modules) may in fact be more useful than the dynamics of whole plants for interpreting what ishappening in a community, because subtle changes in module dynamics may appear long beforechanges in the abundance of individuals, or other variables are detectable. In spite of this, Birch& Hutchings (1992) have noted that “there has been little research directly determining the abilityof plant organs to respond to changes in environment during development”. Most studies of87module dynamics in plants concentrate on changes in module abundance, however, some speciesmay have the ability to alter other module characteristics such as size (i.e. petiole length) as aresponse to stress. Quantifying this type of morphological change at the individual or populationlevel may provide new methods by which to measure the response of a species to changes in it’slocal environment.The results of these experiments investigating length structure of petiole populationssuggest that L. arcticus does respond to changes in their environment by modifying the size ofplant organs such as petiole length. Although plasticity is often measured as an individualresponse to local environmental conditions, this experiment measured plasticity at the wholepopulation level because individuals could not be defined. Comparison of the size structure of thetotal petiole population under different combinations of clipping, fertilizer, and neighbour removalshowed that it was dependent on treatment. Contingency table methods can not describe howtreatments differed with respect to petiole length distribution, however, inspection of the datasuggested certain patterns, the most interesting of which were the fertilizer and neighbour removaleffects. The complex nature of the interactions between treatments make it difficult to interpretpotential patterns, however, there was evidence to suggest that petiole lengths may be longerwhen neighbours are present. This was seen to vary in some cases depending on the fertilizer andclipping treatment experienced. Fertilizer appeared to ameliorate the response to clipping andallow petioles to grow longer in some cases, although this was not conclusive. Although petiolelength distributions did respond to clipping treatments, this was largely due to the method ofremoval of leaves during treatment which selected leaves for clipping based on petiole length.Variation was observed in the response of petiole distribution to treatments from 1991 to1992 as well as from early in the season to later in the season. Further investigation is required todetermine if petiole length distribution will show a stable pattern after long term exposure to thesetreatments. Petiole length distribution may show an inherent variability in response due tovariability in other factors in the environment, or there may be a lag time in response to the88treatments imposed. Long term exposure to treatments of fertilizer, neighbour removal andclipping may eventually show a predictable petiole length distribution. The treatments in thisexperiment may eventually select clones with specific petiole distributions if different distributionsprovide an advantage to populations of L. arcticus against their neighbours. Second, certainpetiole length distributions may be short term plastic responses to environmental changes. This issupported by Birch & Hutchings (1992) who reported that petiole length in Glechoma hederaceaincreased over a longer period of time and was partly age dependent as well as sensitive tochanges in the environment.Petiole length distribution in this experiment was used as a measure of growth or sizedifferences between populations of L. arcticus experiencing different treatments. The ecologicalramifications for differences in petiole length distribution between populations are not yet known.More research is required to determine the implications of a population dominated by longer orshorter petioles Excessive extension may weaken petioles and make them more susceptible todamage in some situations. De Kroon et al. (1992) explain that plant size can increase without acorresponding increase in biomass due to etiolation. Therefore, taller plants may have increasedaccess to light, but be more susceptible to damage due to structural weakening during petioleextension.Other studies that used leaf characteristics such as size, number, and height of leaves tomeasure plasticity of plant growth have reported changes in leaf height or petiole length underdifferent environmental conditions. De Kroon et al. (1992) measured longest leaf (leaf sheathheight) as an indicator of shoot size, and showed that leaf height increased with density for Carexspp. Smaller (shorter) shoots were also showed to have higher mortality at all ranges of density.A study of plasticity of plant size and architecture measured various leaf traits (number ofleaves, shoot height, and leaf length) as size indicators and reported that these growth charactersvaried with the environment in Aster spp., and to a degree in Solidago spp (Schmid & Bazzaz891990). Of the three traits measured, only number of leaves in Solidago spp. did not show aplastic response to changes in environmental conditions.Grace (1985), in an investigation of the trade-off between growth and reproduction in twospecies of cattails (Typha spp.) examined the relationship between tallest leaf and reproductiveoutput. He reported that the taller species tended to have higher reproductive output and moreramets than a shorter, closely related species. Although leaf length may not have been the reasonthat one species had greater reproductive output than the other, a comparison of the correlationbetween reproductive output and size within a species may determine if a trade-off betweenvegetative growth and reproductive allocation is occurring.Most studies examining the plasticity of plant growth as a response to changes in theenvironment concentrate on measuring size in terms of biomass accumulated. While thisinformation is useful, a functional perspective that gives us information on how plants allocatebiomass is also important. To understand this plasticity, we need to know if allocation of biomassto structures changes because number of modules changes, or if module size changes, as in thecase of petiole extension in L. arcticus. Therefore, studies measuring plasticity through biomasschanges alone are not sufficient to describe the nature of a plastic response to environmentalchanges. A plastic response may occur with little or no detectable change in plant biomass due toa re-allocation of existing biomass to new or different structures. A combined approach istherefore recommended that would measure both biomass changes and structural changes within apopulation under different treatments.PERCENT COVERFertilizer and neighbour removalIn the second set of experiments comparing plant growth in different treatments in terms ofhorizontal space acquisition (percent cover), abundance of L. arcticus showed the strongestresponse to neighbour removal. Hutchings & Slade (1988) studied clonal spread in Glechoma90hederacea and showed that shading had a greater affect on architecture than nutrients. Under lowlight conditions there was greater allocation of biomass to petioles.Although the increase in percent cover of L. arcticus in the absence of neighbours indicatedneighbours interfered with the growth of L. arcticus, increasing nutrient availability by applicationof NPK fertilizer did not have the same effect. Fertilizer treatment did show a nonsignfficantincrease in cover of L. arcticus from 1991 to 1992. There are several possibilities for thesediffering responses. First, the presence of neighbours in this habitat did not interfere with L.arcticus’s access to the nutrients applied (NPK). Therefore, the limiting factor in this habitat maybe something else such as various micronutrients, water, or light. Second, the addition ofnutrients in the fertilizer may have been usurped by other more rapidly growing species before L.arcticus could utilize them. Senecio lugens increased when fertilized and clipped, but decreasedwhen only fertilizer was applied. This suggests that L. arcticus was competing with this speciesas Senecio was only able to increase when fertilized if lupines were clipped. Evidence supportingthe hypothesis that rapidly growing species are usurping fertilizer at the expense of L. arcticus isdemonstrated in the rapid increase in cover of Festuca altaica when fertilized, a response alsodemonstrated by other work at these sites (Nams et al. 1993; John & Turkington unpublished).Festuca altaica also increased (nonsignificantly) when lupines were clipped and fertilizer wasapplied. Although this difference was not significant, it was quite large and may indicate thatclipping lupines released of Festuca from competition with L. arcticus. Nams et al. (1993) alsoreported that fertilizer increased growth in other species at these sites as well including:Calamagrostis lapponica, Epilobium angustifolium, and Achillea millefolium. A similarresponse was observed by Davy & Bishop (1984) where application of fertilizer produced a rapidincrease in the graminoids Festuca ovina and Koeleria macrantha to the detriment of Hieraceu,npilosella. A nitrogen fixing legume, Astragulus danicus, was not reported to respond to fertilizertreatments in these same experiments. Similarly, although L. arcticus did increase in cover thiswas not significant in the fertilizer treatment. Third, the rate of fertilizer applied was not sufficient91to produce a strong increase in cover relative to that made available through the removal ofneighbours. Fox & Morrow (1992) also reported that application of fertilizer did not change leafgrowth, but did change leaf quality.ClippingThe effect of removal of leaf tissue in the clipping treatment did not significantly changepercent cover of L. arcticus in this experiment although a nonsignificant increase in cover wasobserved. Several explanations are possible for this unexpected result. First, L. arcticus maycompensate for loss of biomass due to herbivores by stimulating new leaf or ramet production.Crawley (1983) suggests that compensation in plant growth following herbivory may occur at anindividual or a population level. At the population level, tissue loss in some individuals may becompensated by increased growth in individuals not attacked or attacked to a lesser degree due toa competitive release. On an individual level, loss of shaded or unproductive tissue may result inincreased growth or productivity of remaining leaves.Second, insufficient biomass may have been removed to significantly effect growth. In thishabitat, L. arcticus is subjected to periodic high levels of herbivory from fluctuating snowshoehare populations. Because of this they may have developed compensatory mechanisms thatobscure the effect of lost tissue at low to moderate levels of herbivory. It is possible that onlyhigh levels of damage are capable of inducing mortality in ramets or clones. Integration amongramets may expedite this recovery. Jonsdottir & Callaghan (1989) demonstrated that only atcontinuous, high levels of grazing of tillers of Carex bigelowii, was support from physiologicallyintegrated neighbours cut off. In addition, the procedure for simulating herbivory was based on aheight criteria. This allowed smaller leaves to escape clipping. This “refugi&’ for small leavesmay have permitted a rapid replacement of lost leaves and compensation for lost photosyntheticarea. Lupinus arcticus, like many grazed species, has underground meristems that are protected92from mammalian herbivory. This may also have permitted quick recovery of lost biomass and leafsurface area.InteractionsAlthough neighbour removal was the only treatment to show a significant change in percentcover of L. arcticus, some trends in the interaction between neighbour removal, and clipping, andfertilizer were shown. When these two treatments were combined with neighbour removal, theincrease (nonsignificant) in percent cover was greater than either treatment alone. More data isrequired to determine if these trends are true responses or not. Other studies have investigatedinteractions between competition, nutrient availability, and herbivory on plants. In a study ofgrowth and reproduction in Ipomea hederacea it was shown that fertilizer produced a significantincrease in biomass of leaves, stems, seeds, roots and fruit. Negative effects due to thecompetition treatment were partly ameliorated by the addition of nitrogen to the plants (Whigham1984). In my investigation, the effect of fertilizer on growth of L. arcticus was not as great asremoving neighbours.Davy & Bishop (1984b) reported that the response of Hieraceumpilosella to rabbit grazingvaried with application of fertilizer. Application of nutrients (NPK in various combinations)reduced the effect of rabbit grazing on these plants. In contrast, application of fertilizer to clippedpopulations of L. arcticus aggravated the effects of clipping. This treatment was the only one toshow a reduction in cover in L. arcticus relative to control. Although nonsignificant, theameliorating effect of neighbour removal when combined with clipping in my experiment wascomparable to the effect of fertilizer on clipped Hieraceum pilosella reported in Davy & Bishop(1984).Both experiments considered the growth response of L. arcticus under different treatmentsfrom two different perspectives (petiole length distribution and percent cover) and both detectedsignificant treatment effects on growth. Although the fact that petiole length distribution in L.93arcticus is dependent on treatment is interesting, more conclusive results are required todetermine how these distributions vary with treatment. If the trend to grow longer petioles in thepresence of neighbours is a real pattern, this would suggest a trade-off between vertical growthand horizontal spread because the increase in percent cover of L. arcticus was highest whenneighbours were absent. A more detailed experiment of the pattern of response of petiole lengthto the treatments imposed would allow a comparison of horizontal (percent cover) to vertical(petiole length) growth by L. arcticus. This type of study may give us greater information aboutthe nature of L. arcticus interactions with its neighbours in the field, and how these interactionschange when L. arcticus experiences herbivory or changes in nutrient availability. Questions suchas ‘does percent cover increase when petiole populations are taller or shorter?’ and ‘are petioleslonger when neighbours are present?’ could be addressed.CHAPTER 5REPRODUCTIONMany components of reproduction respond to changes in a plant’s local environment, all ofwhich must be measured to get an accurate picture of how plant reproduction varies with theenvironment. These components are (1) reproductive investment (also known as effort), (2)reproductive output, and (3) reproductive efficiency. Reproductive investment is the amount ofresources a plant allocates to a reproductive episode under certain conditions. It includes theamount of energy or resources a plant allocates to the production of reproductive structures andtheir support tissues including reproductive meristems, inflorescences, fruit, flowers, peduncles,pollen, nectar, and seeds. The second component, reproductive output, is limited by the amountof resources a plant invests in reproduction in one season. It is usually considered to be thenumber of seeds or propagules a plant or population is capable of maturing under a certainenvironment, and the number of those seeds that successfully recruit and grow to reproductivematurity. This estimate is often used as a fitness correlate. The third measure is reproductiveefficiency, and it is a ratio of the amount of effort invested in reproduction relative to the numberof offspring (or propagules) produced as a consequence of that investment. The reproductivedata that was used to estimate or calculate these three measures of reproduction for L. arcticuswere collected as part of the main population experiment described in Chapter Two, and adetailed description of methods is contained in that chapter. A brief overview of methods used tocollect data on reproduction follows.9495METHODSDuring the bi-weekly population surveys in of the quadrats in 1991 and 1992, thereproductive variables measured included: number of racemes produced / m 2, number of flowerbuds initiated /m2, number of pods produced /m 2, and the length of individual racemes andpeduncles. At the end of each growing season, this information was used to calculate (pertreatment, and per m2) the mean number of racemes produced, mean number of flowers initiatedand mean number of fruit matured (pods), and mean size of reproductive structures. Based onthis data, reproductive effort in L. arcticus was estimated as the mean number of racemes andflowers initiated /m 2 Reproductive output was estimated as the mean number of mature fruit(pods) produced /m 2 The reproductive efficiency of L. arcticus at converting flowers to maturefruit under different treatments was calculated as a ratio of the number of mature fruit (pods) setrelative to the number of flower buds initiated. This estimate of reproductive efficiency does nottranslate into a fitness correlate because the proportion of fruits set by a population under certainenvironmental conditions is not a measure of the ability, or number of the seeds to recruit andgrow to maturity. However, this measure of reproductive efficiency does give an indication ofhow plants respond to different environments (or treatments). Measurements of long-termrecruiting success were not in the scope of these experiments.STATISTICAL ANALYSISData in 1991 and 1992 were analyzed separately for two reasons. First, the fertilizertreatment was increased in 1992 because of negligible response in the first year. Second, anadditional treatment group was added in 1992 to investigate how a more intense clippingtreatment would affect L. arcticus dynamics. The addition of a new treatment unbalanced theexperimental design in 1992. To accommodate this change in the design, the effect of theintensified clipping treatment added in 1992 was determined by analyzing the 1992 data with aone-way ANOVA by treatment for all reproductive variables measured in addition to thefollowing statistical analyses of the main factorial design. Yearly variation in reproductive96characters was analyzed using a three-way MANOVA of number of racemes, flowers, and podsproduced per m2 by year. The three variables were transformed using the square root functionfor Poisson data (x + 0.5) 1/2 (Zar 1984). Reproductive effort between treatments was analyzedusing a three-way MANOVA of raceme and flower production in 1991 and 1992. Poissonvariables in all analyses were transformed with the square root function mentioned previouslyunless otherwise specified. Reproductive output between treatments was compared using athree-way ANOVA of pod production per m2. Reproductive efficiency in each treatment groupwas calculated as (pods produced per m2 / flower buds initiated per m2). This was analyzedusing the same methods as for reproductive output except that reproductive success values weretransformed using an arcsine transformation for percent data. The effect of the additional clippingtreatment in 1992 was determined by re-analyzing the data set using a one-way MANOVA bytreatment.If data were found to violate assumptions of normality or homoscedasticity, transformationswere performed on the data prior to analysis. If following transformations data still did notconform to the assumptions required for parametric analysis, a non-parametric Kruskal-Wallisanalysis was performed. The results were reported and compared to parametric statistics whennecessary.The effect of treatments on mean size of reproductive structures (raceme and pedunclelength) was analyzed in 1992 to determine if there was plasticity in the size of the structures inresponse to treatments. Low numbers of reproductive structures formed in 1991 precludedanalysis of 1991 data.RESULTSYEARLY VARIATIONThere was significant (P=O.OO1) yearly variation in both reproductive investment and outputacross all treatments (Tables 22 a, b). The trend observed in this experiment was relatively lowreproductive effort and output in 1991 with only 107 racemes and 399 flowers initiated (Fig. 5).97The total number of fruits matured across all treatments was 31(7.8% of the number of flowersinitiated). Peak densities of snowshoe hares occurred in the study area in 1990 and were indecline by 1991 (Fig. 6). Reproductive output in L. arcticus increased substantially in 1992following the start of the decline in hare numbers at Kluane. Reproductive investment increasedto 281 racemes and 2858 flower buds produced in the 1992 season, this was an increase of 262%and 716% respectively over the previous year. The number of fruits reaching maturity increasedto 625 pods, an increase of 2 1.9% (Fig. 5). In 1991, five experimental quadrats in four differenttreatments showed no reproductive investment at all. In the following year, all quadrats notreproducing in the previous year showed some reproductive investment ranging from 1 to 13racemes. In 1992, two quadrats from two treatments showed no reproductive investment. Onehad no prior reproductive investment in 1991 and the other had produced 1 raceme.REPRODUCTION 1991InvestmentAnalysis of reproductive investment by treatment in 1991 indicated that there were nosignificant differences in the amount of investment made to the formation of reproductivestructures (racemes and flower buds) between the treatment groups (Tables 23 a, b). Alltreatment groups showed some degree of allocation to reproduction in 1991, but investmentwithin treatments was highly variable (Fig. 7 (raceme production), Figure 8 a (flower budinitiation))OutputReproductive output in 1991 was low. Parametric tests of the effect of treatments onsquare root transformed pod production ,1m 2 indicated no significant treatment effects. Nonparametric Kruskal-Wallis ANOVA was performed due to violations in the assumptions of98normality and homoscedasticity. No significant effects were found in the parametric (Table 24 a),or the non-parametric tests (Table 24 b).Reproductive efficiencyIn addition to comparing the effect of treatments on total reproductive output I m 2, therelative efficiency of reproduction in the experimental populations was tested to determine if theproportion of fruits ripening differed between treatments. This was a comparison of reproductiveefficiency, or the ability of treated populations to translate the number of flowers initiated intomature fruit. Reproductive success or efficiency was estimated and compared using the arcsinetransformed function of total number of pods / m 2 relative to the total number of flower budsinitiated / m 2 A three-way ANOVA of reproductive efficiency in 1991 indicated a significantclipping effect (P<O.003) on that reduced the proportion of fruits maturing (Table 25, Figure 8).Non-parametric Kruskal-Wallis was performed due to violations in the assumptions of normalityand homoscedasticity. The non-parametric test indicated no significant treatment effect onreproductive efficiency (P<O.098, Table 24 b).RecruitmentSeedling recruitment in the first year of the study was almost nonexistent across alltreatment populations as well as in the natural population. Only three seedlings recruited into theexperimental quadrats in 1991, and they all recruited in different quadrats and in differenttreatment groups. All seedlings over-wintered and re-emerged in 1992. Growth was limited inthe first two seasons of growth with a maximum of two leaves appearing in the first summer andthree in the second.992500-2000-0z 1500-11000-500 -0-7__________________________6CuFigure 6: Snowshoe hare spring densities on CSP treatment grids from 1987-1994. The controltreatment shows the beginning of the crash in 1991.3000RacemesFlowers• Fruit1991 1992Yearly differencesFigure 5: A comparison of total number of racemes, flowers and podsproduced across all treatments in 1991 and 1992.LEI ControlFoodFertilizerPredator ExciosurePredator Exclosure + Foodft,Ii1987 1988 1989 1990ii1991 1992 1993 1994100Table 22: a) Univariate source of variation for a one-way MANOVA comparing total raceme,flower bud and pod production/rn2in 1991 and 1992.Variable SS dF MS F ratio ProbabilityRacemes 10.503 1 10.503 6.681 0.001Error 97.479 62 1.572Flower buds 325.543 1 325.543 16.373 0.000Error 1270.582 62 20.493Pods 69.670 1 69.670 13.886 0.000Error 311.063 62 5.071b) Multivariate test statistics of racemes, flower buds, and pods /m2 by year.Test Statistic Statistic dF ProbabilityWilks’ Lambda 0.58 1 3, 60 0.000F statistic 14.4451014030•ü racemes-19922011’k-E; - - - . E E •1) r iTreatmentFigure 7: Reproductive effort in 1991 and 1992 as measured bymean (±SEM) raceme production per m2 for each treatment.FrequencyofflowerbudsorpodsFrequencyofflowerbudsorpodst’J‘3)-CDu.0uocuc.0000000C00V0cMCcMCM0IIIIII—IIIIIIICD____________________________________________ControliControl-—FertilizedFertilized---ClippedClippedICDRemovali‘Removal-tJzHCMH-CDFert-Clip--‘—Fert-Clipa.Fert-Rem.-_______Fert-Rem.___oQ_______Clip-Rem±:j—‘0-IFert-CI-RemClip-Rem___IFert-Ci-RemCMExtra-Clip-________________________________________________________________________________CD-________________________C) DCD:L1•L1•00CD00CDCD -Cl)CM I0-L’J\0C103Table 23: a) Univariate statistics from a MANOVA testing the effect of treatment onsquareroot transformed raceme production and flower bud initiation in 1991.Effect Variable SS dF MS F ProbabilityFertilizer Racemes 1.421 1 1.421 1.781 0.195error 19. 149 24 0.798Flowers 1.379 1 1.379 0.246 0.625error 134.731 24 5.614Clipping Racemes 0.061 1 0.061 0.076 0.784error 19. 149 24 0.798Flowers 6.275 1 6.275 1.118 0.301error 134.731 24 5.614Neighbours Removed Racemes 1.009 1 1.009 1.265 0.272error 19. 149 24 0.798Flowers 16.618 1 16.618 2.960 0.098error 134.731 24 5.614b) Wilks’ Lambda and F statistics from MANOVA of 1991 estimates of raceme and flowerbud production in 1991.Effect Statistic dF ProbabilityFertilizer Wilks LambdaF-statistic 2, 23 0.375Clipping Wilks’ LambdaF-statistic 2, 23 0.494Neighbours Removed Wilks’ LambdaF-statistic 2, 23 0.261104Table 24 a: Three way ANOVA test of square root transformed pod production / m2 in 1991.Source SS dF MS F ratio ProbabilityFertilizer 1.392 1 1.392 2.229 0.150Clipping 1.692 1 1.692 2.710 0.114Removal 0.026 1 0.026 0.042 0.840Fert* Clip 1.008 1 1.008 1.614 0.217Fert* Rem 0.018 1 0.018 0.029 8.650Clip*Rem 0.114 1 0.114 0.182 0.674Fert*Clip*Rem 0.002 1 0.002 0.003 0.960Error 13.728 22 0.624Table 24 b: Summary of Kruskal-Wallis non-paramethc test results for some reproductionvariables.Variable Year Kruskal-Wallis statistic dF ProbabilityReproductive output 1991 5.591 7 0.588Reproductive efficiency 1991 12.068 7 0.098Reproductive output 1992 6.066 8 0.640Reproductive efficiency 1992 17.5 8 0.025Raceme length 1992 8.625 7 0.28 1Peduncle length 1992 5.180 7 0.638105Table 25: Three-way ANOVA test of arcsine transformed reproductive efficiency in 1991.Source dF MS F ratio ProbabilityFertilizer 1 0.000 0.009 0.928Clipping 1 0.252 15.436 0.003Removal 1 0.007 0.43 1 0.526Fert* Clip 1 0.001 0.055 0.8 19Fert* Rem 1 0.00 1 0.039 0.847Clip*Rem 1 0.004 0.265 0.618Fert*Clip*Rem 1 0.002 0.116 0.741Error 10 0.016REPRODUCTION 1992InvestmentIn spite of higher overall levels of reproduction in 1992, no significant treatment effectswere observed on the amount of effort invested in reproduction measured by the number ofracemes produced and the number of flower buds initiated / m 2 Tests for treatment effects onreproductive investment were performed using a three-way MANOVA to assess the effect of theoriginal factorial experimental design. The effect of the intensified clipping treatment added in1992 indicated no significant treatment effects (Tables 26 a - c). Variables used to measurereproductive investment in 1992 violated assumptions of bivariate normality. Transformationsused on variables to make them more homoscedastic were only moderately successful. Varioustransformations did improve the data in that although the assumptions were still violated, it was toa lesser degree. Reproductive investment variables were measured as ‘count’ or discrete numbersand normally fall into a Poisson distribution. When sample sizes are large enough, the Poissondistribution approximates a normal distribution. The highly variable level of reproductive106investment observed in the field (Figs. 6, 7 b), suggest that samples sizes may not have been largeenough to meet these criteria.OutputIn 1992, parametric analysis of the log transformed data on reproductive output detected nosignificant treatment effects (Tables 27 a, b). Data in this analysis was not normally distributed.Although ANOVA is robust to departures from normality and homoscedasticity, a non-parametricKruskal-Wallis was done and also failed to detect significant treatment effects (P<0.640, Table 24b). Two treatment groups were marginally insignificant in their effect on total pod production/m2: fertilizer (P<0.094), and neighbour removal (P<0.080). Further examination of the fertilizerand neighbour removal treatments indicates that they have opposing effects on reproductiveoutput such that neighbour removal increased pod production and fertilizer decreased it (Fig. 8 b).Application of fertilizer reduced mean (±SEM) pod production! m2 to 2.00! m2 ±1.683 relative tothe control of 23.75±20.126, whereas removal of neighbouring species increased pod productionto 61.75 pods /m2 ±33.908.Reproductive efficiencyThe reproductive efficiency of L. arcticus in terms of fruit maturation changed from 1991to 1992. In 1992, a significant fertilizer effect (P<0.001), and a moderately nonsignificantneighbour removal effect on reproductive efficiency (P<0.084) were detected (Table 28 a). Nosignificant clipping effects were observed in the three-way analysis of the original experimentaldesign. One-way ANOVA of all nine treatments in 1992 found a significant treatment effect atP<0.002 (Table 28 b). A post hoc Tukey grouping (Table 28 c) indicated a strong fertilizer effectsuch that fertilizer, the interaction between fertilizer and neighbour removal, and the fertilizer byclipping treatment were significantly different from neighbour removal and the extra clippingtreatment in 1992. A significant difference in clipping on reproductive efficiency was only seen107for the intense clipping treatment in 1992. A Kruskal-Wallis analysis by treatment was performedon arcsine transformed reproductive efficiency was significant (P<0.025, Table 24 b). Fertilizer,fertilizer by clipping and fertilizer by neighbour removal have lower mean reproductive efficiencythan the intense clipping treatment and neighbour removal (Fig. 8 b).Size of Reproductive StructuresMultivariate analysis of treatment effects on the length of racemes and peduncles (Table 29)indicated no significant treatment effects (Table 30a - d). A Kruskal-Wallis analysis confirmedthe non-significant results (Table 24 b).Table 26: a) Univariate statistics from a MANOVA testing the effect of treatment ontransformed raceme production and flower bud initiation in 1992.Effect Variable SS dF MS F ratio ProbabilityFertilizer Racemes 0.008 1 0.008 0.040 0.84Error 4.988 24 0.208Flowers 62.286 1 62.286 1.729 0.20Error 864.470 24 36.020Clipping Racemes 0.427 1 0.427 2.055 0.16Error 4.988 24 0.208Flowers 6.960 1 6.960 0.193 0.66Error 864.470 24 36.020Neighbours removed Racemes 0.250 1 0.250 1.204 0.28Error 4.988 24 0.208Flowers 106.164 1 106.16 2.947 0.09Error 864.470 24 36.020108b) Wilks’ Lambda and F statistics from MANOVA of 1992 estimates of log transformed racemeand flower bud production in 1992.Effect Test Statistic dF ProbabilityFertilizer Wilks’ Lambda 0.93 1F-statistic 0.858 2, 23 0.210Clipping Wilks Lambda 0.9 16F-statistic 1.057 2, 23 0.164Neighbours Removed Wilks’ Lambda 0.848F-statistic 2.056 2, 23 0.474Table 27: a) Three-way ANOVA of log transformed pod production I m 2jn 1992.Source dF MS F ratio ProbabilityFertilizer 1 1.341 3.036 0.094Clipping 1 0.000 0.001 0.976Removal 1 1.476 3.342 0.080Fert* Clip 1 0.228 0.515 0.480Fert* Rem 1 0.035 0.079 0.78 1Clip*Rem 1 0.014 0.032 0.859Fert*CIip*Rem 1 0.045 0.102 0.752Error 24 0.442b) One-way ANOVA of pod production I m2 in 1992 to determine the effect of theintensified clipping treatment on reproductive output.Source SS dF MS F ratio ProbabilityTreat 3.140 8 0.393 0.912 0.522Error 11.623 27 0.430109Table 28: a) Three-way ANOVA of arcsine transformed reproductive efficiency in 1992.b) One-way ANOVA of arcsine transformed reproductive efficiency in 1992.Source SS dF MS F ratio ProbabilityTreat 10.210 8 1.276 4.258 0.002Error 8.093 27 0.300c) Post hoc Tukey comparison of reproductive efficiency by treatment in 1992. Note: C+ isthe intensive clipping treatment added in 1992.C+ NR C/NR C Cont F/C/NR F F/NR F/C.V V V V V V VTable 29: Mean size (±SEM) of reproductive structures by treatment for 1992.Length Cont Fert Clip Rem F/C F/N C/N F/C/N Extra(cm) (F) (C) (N) ClipRaceme 4.46 8.05 4.49 6.75 2.62 3.81 7.17 4.29 6.23±SEM 0.70 0.39 0.70 1.13 0.69 1.02 0.37 0.36 0.92Peduncle 4.60 6.1 3.98 5.66 2.77 2.69 5.57 4.99 4.3±SEM 0.70 0.02 0.61 0.84 0.74 0.54 0.27 0.32 0.62Source dF MS F ratio ProbabilityFertilizer 1 4.441 13. 170 0.001Clipping 1 0.152 0.452 0.508Removal 1 1.095 3.246 0.084Fert*Clip 1 0.054 0.161 0.692Fert*Rem 1 0.268 0.794 0.382Clip*Rem 1 0.933 2.767 0.109Fert*Clip*Rem 1 0.021 0.062 0.806Error 24 0.337110Table 30: a) Univariate (three-way) test statistics of length of racemes, and length of peduncles in1992.Effect Variable SS dF MS F ratio ProbabilityFertilizer Raceme length 3.503 1 3.503 0.293 0.59Error 286.899 24 11.954Peduncle length 0.853 1 0.853 0.009 0.75Error 207.613 24 8.651Clipping Racerne length 8.527 1 8.527 0.7 13 0.40Error 286.899 24 11.954Peduncle length 0.172 1 0.172 0.02 0.88Error 207.613 24 8.651Removal Raceme length 11.843 1 11.843 0.991 0.33Error 286.899 24 11.954Peduncle length 4.309 1 4.309 0.498 0.48Error 207.613 24 8.651b) Multivariate test statistics from three-way MANOVA of length of racemes, and length ofpeduncles in 1992.Effect Test statistic Statistic cIF ProbabilityFertilizer Wilk& Lambda 0.98 8 2, 23 0.870F statistic 0.140Clipping Wilks’ Lambda 0.946 2, 23 0.53 1F statistic 0.652Neighbours Removed Wilks’ Lambda 0.960 2, 23 0.662F statistic 0.484c) Univariate (one-way) test statistics of length of racemes and length of peduncles by treatmentin 1992.Variable SS dF MS F ratio ProbabilityRaceme length 93.939 8 11.742 0.929 0.50Error 341.414 27 12.645Peduncle length 52.925 8 6.6 16 0.747 0.65Error 239.143 27 8.857111d) Multivariate test statistics from one-way MANOVA of length of racemes and length ofpeduncles by treatment in 1992Test statistic Statistic dF ProbabilityWilks’ Lambda 0.695 16, 52 0.829F statistic 0.648RecruitmentIn 1992, seedling recruitment into the experimental quadrats remained low with only fivenew seedlings appearing in four different treatment groups. All seedlings survived until August24, 1992.DISCUSSIONReproductive investment in L. arcticus in the Kluane region was very low in 1991 relativeto investment in the following year irrespective of treatment applied. Yearly variation inreproduction is not uncommon in many species and in some cases has been shown to becorrelated to growth in the previous year (Barkham 1980). Although many environmental factorsvary yearly, patterns of reproductive investment in L. arcticus may be correlated with the 10-yearcycle of abundance of snowshoe hares, a keystone herbivore in the area that is known to feed onL. arcticus. The first year of the experiment followed the year of peak hare density. Lupinusarcticus probably experienced high levels of herbivory at the study site in 1990, and this may haveresulted in low investment to reproduction in 1991 in the treatment quadrats. In 1991, thesnowshoe hare population was declining, and experimental quadrats were protected frommammalian herbivores by fencing. This reduced level of herbivory across all treatments may haveresulted in higher levels of investment in 1992. Investment to reproductive structures in L.arcticus occurs very early in the growing season immediately following snow melt and prior toleaf opening. Racemes break ground concurrently with leaf buds. As a result, it was unlikely112that any treatment effect on reproductive investment would be observed in 1991, particularlybecause clipping and neighbour removal did not occur until plants in the experimental quadratswere established. No treatment effect was observed on other measures of reproduction in 1991.Several explanations are possible. The slow growth rate of L. arcticus in the northern borealforest may produce a lag in response time to treatments. If this is the case, long term experimentswould be required to detect treatment effects. Alternatively, Grime (1977) would argue that L.arcticus in this environment is a stress tolerator and is not adapted to respond to short-termenvironmental changes. Rather, it is adapted to endure adverse conditions and should notrespond to any treatments in this experiment unless they represent long term environmentalchange. If this were the case, no treatment effect should be expected in 1992.Although there were no significant differences in reproductive investment, reproductiveoutput, or size of reproductive structures, reproductive efficiency did show some complextreatment effects in 1992. Three-way analysis of reproductive efficiency indicated that onlyfertilizer had a significant treatment effect such that reproductive efficiency in maturing fruit wassignificantly lower when L. arcticus was fertilized than in all other treatments and combinations.A comparison of treatments showed that intense clipping and neighbour removal were differentfrom fertilizer and all two-factor fertilizer interactions. The fertilizer effect on reproduction hasbeen observed with other species and may be due to suppression of L. arcticus by more rapidlygrowing neighbours that usurp nutrients. This is a reasonable explanation because when fertilizerwas applied in the absence of neighbours no significant effect was detected for reproductiveefficiency of reproduction. Also, experiments on growth of L. arcticus found that percent coverof a graminoid Festuca altaica increased significantly when fertilized. Removal of neighboursmay have reduced interference and allowed L. arcticus to mature a greater proportion of fruitsthan in other treatments. The rate of fertilized application was low in 1991 and may not havebeen sufficient to produce a detectable response. Rate of application was doubled in 1992 to113ensuse that the level of nutrient applied to the experimental quadrats was sufficiently high enoughto constitute a treatment capable of producing detectable responses.It is unclear why the intense clipping treatment had a higher proportion of fruit mature thanother treatments. It is possible that this is not a true treatment effect if there is a lag in theresponse time of L. arcticus to environmental changes. These populations may be responding toenvironmental changes occurring in the previous season, rather than to treatment effects imposedin 1992. More data are required to test this hypothesis.Although reproductive output in 1992, as measured by pod production /m2, showed nosignificant treatment effects, two treatments were only marginally nonsignificant: fertilizer andneighbour removal. No conclusions can be drawn due to lack of significant statistical results, butthis may indicate a trend that would become significant if the experiment were maintained longer.Reproductive output was lower in the fertilizer treatment than in the control, and much lower thanmean output when neighbours were removed. This suggests that the presence of neighbours maybe mediating the response of L. arcticus to changes in the environment. With the high level ofvariability in reproductive measures within treatment, it is not possible to determine if theobservations regarding the fertilizer and neighbour effect are spurious. Alternatively, the highlevel of variability in response variables is commonly observed in field experiments and may bemasking treatment effects in this experiment. This variability may be due to uncontrolled factorssuch as spatial heterogeneity in the growing environment, heterogeneity in the past histories ofexperimental populations (i.e. past flowering history, grazing, neighbour differences, availabilityof resources, etc.). These uncontrolled variables may be more important in determining theresponse of L. arcticus populations than the experimental treatments imposed. Therefore, theinterpretation of nonsignificant results must be undertaken with caution and must be verified withfurther experimentation before any conclusions are drawn.Seedling recruitment into experimental quadrats was low in 1991 and 1992. Little is knownabout patterns of recruitment of L. arcticus in this environment, however patterns of recruitment114may also be correlated to the snowshoe hare cycle. No data are available at this time to commenton this possibility. However, field observations at this study site in 1993, one year after the crashin hare populations, recorded increased levels of L. arcticus recruitment not seen in previousyears.115CHAPTER 6SUMMARY AND CONCLUSIONSThe aim of this study was to investigate the relative importance of clipping, neighbours, andsoil fertility level as potential limiting agents on field populations of Lupinus arcticus anddetermine how they interact to produce observed patterns of abundance and dynamics. Althoughpopulations did respond to the treatments imposed in this experiment, none of the main effectsstrongly influenced their dynamics, but rather had other effects (Table 31). Lupinus arcticuspopulations were influenced by the interactions of those main effects, in particular, the distributionof petiole lengths was strongly influenced by an interaction between treatments.Table 31: Summary of significant main effects detected in 1991 and 1992.Fertilizer effects Clipping effects Neighbour removal effectsT incidence of disease 1991 incidence of disease 1991/2L standing crop available 1992 1’ standing crop available 1991reproductive efficiency 1992 reproductive efficiency 1992I- total leaf density 199 1/21- leaf survivorship 1992L total leaf mortality 1991IL. arcticus cover 1991/2The most striking effects detected were the increased percent cover and the reduced leafmortality in L. arcticus when neighbours were removed, indicating that competitive release didoccur. Second, there was some evidence to suggest that L. arcticus was able to compensate forclipping as percent cover of lupines was not reduced by clipping, rather it showed an116nonsignificant increase. The standing crop of lupine leaves was also greater when populationswere clipped in spite of a reduction in overall leaf density, which suggests that a compensatorymechanism that reduces the rate of leaf turnover when populations were clipped. In addition, theproduction of new leaves (buds) did not decline in response to clipping although total leaf densitywas reduced. This suggests that L. arcticus is able to maintain leaf production in spite of thedisturbance. Third, the consequences of increased nutrient availability on L. arcticus populationswere primarily negative as fertilization was shown to increase the incidence of disease on leaves,as well as to reduce both the reproductive efficiency and standing crop. These negative effectsmay be attributed to the inhibitory effect of nutrient addition (NPK) on nitrogen fixation in L.arcticus. However, there was also evidence to suggest that nutrients may have been usurped bymore rapidly growing species consequently leading to reduced growth of L. arcticus.Although L. arcticus populations did show some significant responses in this study, theywere comparatively weak considering the intensity of the treatments imposed. These unexpectedresults necessitate the question as to why L. arcticus did not respond more strongly to thesefactors. To properly address this question, we must take a ‘plant’s-eye-view’ of the habitat that L.arcticus lives in - the boreal forest understory. It is generally considered that plant’s growing inthis habitat have to contend with two major factors. First, they must endure a relatively constant,stressful (sensu Grime 1979) habitat characterized by reduced levels of light filtering through thecanopy, prolonged periods of low winter temperatures, brief, cool growing seasons, low soilfertility, and low moisture (Elliott-Fisk 1993). Second, they are subjected to periodic increases inherbivory when snowshoe hare abundance is at its peak during their 10-year cycle. From Grime’s(1979) point of view, this is a stressful environment subject to short, periodic increases lowintensity disturbance (herbivory). In this type of environment, where plants have evolvedcharacteristic which permit prolonged, slow vegetative growth it might be expected that plantswould not respond, or would respond very slowly to short term increases in nutrients or reducedinterference from neighbours. The abiotic conditions required to sustain elevated levels of plantgrowth to take advantage of short term increases in nutrients or competitive release (neighbour117removal) are not available in this environment. As conditions for plant growth in the boreal forestare relatively poor, competition between neighbours may be weak (Grime 1977), or strong(Tilman 1988). Lupinus arcticus populations persist during the peak densities of snowshoe hares,so it is evident that they must be able to resist, or compensate for the periodic increases in theamount of herbivory they experience. This may explain why clipping, even severe clipping, didnot affect L. arcticus dynamics to any great degree. The low level of response of L. arcticuspopulations to these treatments supports Grime’s (1979) premise that plants living in stressfulenvironments have evolved strategies of tolerance that do not permit them to respond to shortterm increases in local resource abundance. Lupinus arcticus does have some of the propertiescharacteristic of a stress-tolerant species described by Grime (1979) (e.g. long-lived perennial,nitrogen fixer, low rate of flowering, slow growth rate). However, other common attributes arelacking (e.g. evergreenness, lack of morphogenetic plasticity), and therefore it probably falls intoGrime’s (1979) stress-tolerant competitor category. This may explain why a low level of responsewas detected to the treatments imposed.As for how this study fits into the debate regarding top-down and bottom-up regulation ofcommunities, the lack of strong response to short term treatments of clipping, fertilizer, andneighbour removal suggests that neither top-down or bottom-up factors were strongly limitingthese natural populations at the time of the survey. It may be that evolutionary characteristics thathave permitted L. arcticus to persist in this stressful (sensu Grime 1979) habitat are moreinfluential in determining the patterns of growth and dynamics than short-term changes inecological factors. In this sense, the long-term adaptations to the abiotic conditions of the borealforest may have pre-determined the limited responses to short-term changes in herbivory,interference from neighbours, or soil nutrient availability.An alternative explanation to the low level of response in these experiments may be theresult of high levels of variation within treatments masking the response to treatments. Theoriginal experimental design tried to deal with this problem by replicating and randomly assigningtreatments to the experimental quadrats. In addition, attempts were made to choose quadrat sites118that were visually similar in terms of density of lupines and vegetation cover. In spite of this, thelarge error sums of squares in some of these analyses indicated that high amounts of variationwere still present in the data. This thesis did not attempt to partition the variation in this errorterm to test this possibility due to the lack of pre-treatment data, or appropriate variables to use ascovariates in the analysis. Future experiments in this area would be advised to address thisproblem in the experimental design by (i) doing one pre-treatment survey in order to have somemeasure of pre-treatment variation in the experiment. (ii) Use a randomized block design, (iii)collect additional data to be used as covariates during the analyses.APPENDIX ACHARACTERISTICS OF THE WEIBULL DISTRIBUTIONThe Weibull distribution (Weibull 1931, 1959) is a “generalization of the exponentialdistribution,” but it does not assume, as does the exponential model, a constant hazard rate overtime or age (Lee 1980). This model of survivorship is described by 4 equations:cxS(t)= e _Q,*t) (1)h(t)= ?. * a * (*t)(a -1) (2)aF(t)= 1 -e *t) (3)af(t)=?* a * (2L*t)(a l)*eO*t) (4)S(t) is the probability of survivorship at time or age (t), lambda () is the scale parameter,and alpha (a) is the shape parameter, h(t) is the hazard function that describes the instantaneousmortality rate or age-specific mortality rate, F(t) and f(t) are the cumulative distribution functionand the probability density function respectively (Lee 1980). The relationship between theseequations is described by Lee (1980):S(t) = 1 - F(t) (5)Therefore F(t), the cumulative distribution function is the probability that an individual diesbefore time (t). The probability density function f(t), is the limit of the probability that anindividual dies or fails in a short time interval of t + z\ t per unit width A t. Briefly, the probabilityof dying in a very short time window. These two functions are used to determine theinstantaneous mortality rate:h(t)=f(t)/(1-F(t)) (7)119120In the Weibull model of survivorship, the instantaneous mortality rate can remain constant,increase, decrease, or vary nonlinearly.According to Lee (1980), the use of theoretical survival distributions to describe observedsurvivorship data has occurred since the 1940’s beginning with the exponential decay curve.Theoretical distributions are useful because they allow a mathematical description of survival timebetween groups even when the factors causing this distribution are not known or available formeasure (Lee 1980). This is the case with many ecological studies. The parameters generated bythe fitting a theoretical model to an observed data set are not necessarily known biologicalparameters, however they can be used as the basis for comparing different treatments. In thisanalysis two Weibull model parameters form the basis of the statistical comparison: the scaleparameter (?), and the shape parameter (cc). The two Weibull equation parameters . and cxdetermine both quantitative and qualitative aspects of the predicted curve generated. When (X=1, the hazard rate is constant (exponential decay). The hazard rate increases (over time) when cx>1, and decreases as o< 1. Changes in 2 (scale parameter) produce quantitative changes in thesurvival curve (Lee 1980, 1992). Varying cc (shape parameter) results in qualitative changes tothe survival curve that allow equations to be fit to different models of survivorship including TypeI, II , III hypothetical survivorship models (Fig. 9).(J) -(I——4Figure 9: Survival curves generated by the Weibull distribution when ?=1 and a is allowed tovary. Excerptedfrom Lee (1992).iti:lype ]I4 tBIBLIOGRAPHYAarssen, L.W. , G.A. Epp. 1990. Neighbour manipulations in natural vegetation: a review. J.Veg. Sci. 1:13-30.Aarssen, L.W. , D.L. Irwin. 1991. What selection: herbivory or competition. Oikos 60(2):261-262.Aarssen, L.W. , D.R. Taylor. 1992. Fecundity allocation in plants. Oikos 65:225-232.Abul Fatih, H.A. , F.A. Bazzaz. 1984. The influence of defoliation on the performance ofAmbrosia trifida plants. J.Coll.Sci. King Saud Univ. 15(1):55-62.Alexander, H. 1992. Evolution of disease resistance in natural plant populations. In ed.R.S. Fritzand E.L. Simms Plant Resistance to Herbivores and Pathogens. University of Chicago Press,ChicagoAndersson, M. , S. Jonasson. 1986. Rodent cycles in relation to food resources on an alpineheath. Oikos 46:93-103.Antonovics, J. 1980. Concepts of resource allocation and partitioning in plants. Pages 1-25 inLimits to Action. The allocation of individual behaviour. J.E.R. Staddon, editor. AcademicPress, New York.Bakker, J.P. , J.C. Ruyter. 1981. Effects of five years of grazing on a salt-marsh vegetation.Vegetatio 44:81-100.Baldwin, I. 1988. The alkaloidal responses of wild tobacco to real and simulated herbivory.Oecologia 77:378-38 1.Baldwin, I. 1990. Herbivory simulations in ecological research. TREE 5(3): 91-93.Banyikwa, F.F. 1988. The growth response of two East African perennial grasses to defoliation,nitrogen fertilizer and competition. Oikos 51: 25-30.Barkham, J.P. 1980. Population dynamics of the wild daffodil (Narcissus pseudonarcissus). II.Changes in number of shorts and flowers, and the effect of bulb depth on growth andreproduction. J. Ecol. 68:635-664.Wilson, J. B. 1989. Root competition between three upland grasses. Func. Ecol.3:447-451.Bazely, D.R. , R.L. Jefferies. 1986. Changes in the composition and standing crop of salt-marshcommunities in response to the removal of a grazer. J. Ecol. 74:693-706.121122Bazzaz, F.A. , J.L. Harper. 1977. Demographic analysis of the growth of Linum usitatissimum.New. Phytol. 78:198-209.Begon, M. , M. Mortimer. 1986. Population Eco1ov: A Unified Study of Animals and Plants..Sinuaer Associates Inc. Publishers. Sunderland Massachusetts. 220 pp.Beisky, A.J. 1986. Does herbivory benefit plants? A review of the evidence. Am. Nat.127(6)870-892.Beisky, A.J. 1987. The effects of grazing confounding ecosystem, community, and organismscales. Am. Nat. 129(5):777-783.Berendse, F. 1985. The effect of grazing on the outcome of competition between plant speceieswith different nutrient requirements. Oilcos 44: 35-39.Binkley, D. 1986. Forest Nutrition Management. John Wiley, Sons. New York.Birch, C.P.D., M.J. Hutchings. 1992. Analysis of ramet development in the stoloniferous herbGlechoma hederacea using a plastochron index. Oikos 63:387-394.Bishop, G.F. , A.J. Davy. 1984. Significance of rabbits for population regulation of Hieraceumpilosella in Brecidand. J. Ecol. 72:273-284.Black, J.N. 1960. The significance of petiole length, leaf area, and light interception incompetition between strains of subterranean clover. Aust. J. Agric. Res. 11:277-291.Bradshaw, A.D. 1959. Population differentiation in Agrostis tenuis Sibth. II. The incidence andsignficance of infection by Epichloe typhina. New Phytol. 58:310-3 15.Bryant, J.P., F.S. Chapin III., , D.R. Klein. 1983. Carbon/nuthent balance of boreal plants inrelation to vertebrate herbivory. Oikos 40:357-368.Bryant, J.P., T.P. Clausen, P.B. Reichart, M.C. McCarthy, R.A. Werner. 1987a. effect ofnitrogen fertilization upon the secondary chemistry and nutritional value of quaking aspen(Populus tremuloides) leaves for the large aspen tortix (Choristoneura conflictanan (Walker)).Oecologia 73:513-517.Bryant, J.P., F.S. Chapin III, P.B. Reichart, , T.P. Clausen. 1987b. Response of winter chemicaldefence in Alaska paper birch and green alder to manipulations of plant carbon/nutrient balance.Oecologia 72:510-5 14.Callaghan, T.V. , U. Emanuelsson. 1985. Population structure and processes of tundra plants andvegetation. The population structure of vegetation. In White, J. (ed.) Junk, Dordrecht. pp. 399-439.123Callaghan, T.V. 1987. Growth and population dynamics of Carex bigelowii in an alpineenvironment. Oilcos. 27:402-413.Campbell, B.D., J.P. Grime, J.M.L. Mackey, A. Jalili. 1991. The quest for mechanisticunderstanding of resource competition in plant communities: the role of experiments. Funct. Ecol.5:241-253.Campbell, B.D. , J. P. Grime. 1992. An experimental test of plant strategy theory. Ecology73(1): 15-29.Cavers, P.B. , J.L Harper 1967. Studies in the dynamics of plant populations. I. The fate of seedand transplants introduced into various habitats. J. Ecol. 55:59-7 1.Chabot, B.F. , D.J. Hicks. 1982. The ecology of leaf life spans. Ann. Rev. Ecol. Sys. 13:229-259.Chapin, D.M. , L.C. Bliss. 1989. Seedling growth, physiology and survivorship in a subalpine,volcanic environment. Ecology 70(5): 1325-1334.Chazdon, R.L. 1991. Effects of leaf and ramet removal on growth and reproduction of Geonomacongesta, a clonal understorey palm. J. Ecol. 79(4): 1137-1146.Chesson, P.L. , R.R. Warner. 1981. Environmental variability promotes coexistence in lotterycompetitive systems. Am. Nat. 1 17(6):923-943.Clay, K. 1988. Fungal endophytes of grasses: a defensive mutualism between plants and fungi.Ecology 69: 10-16.Clearwater, M.J. , K.S. Gould, 1993. Comparative leaf development in juvenile and adultPseudopanax crassifolius. In press. C.J.B.Cody, M.L. 1966. A general theory of clutch size. Evolution 20:174-184.Coley, P.D., J.P. Bryant, , F.S. Chapinlll. 1985. Resource availability and plant antiherbivoredefence. Science 230(4728) :895-899.Connell, J.H. 1983. On the prevalence and relative importance of interspecific competition:evidence from field experiments. Am. Nat. 122(5):661-696.Coppock, D.L., J.K. Detling, J. Ellis, , M.I. Dyer. 1983. Plant-herbivore interations in a NorthAmerican mixed grass prairie. I. Effects of Black-tailed Prairie Dogs on intra-seasonalaboveground plant biomass and nutrient dynamics and plant species diversity. Oecologia 56: 1-9.Crawley, M.J. 1983. Herbivorv: The Dynamics of Animal-Plant Interactions. Studies in Ecologyvol. 10. University of California Press, Berkeley. 437 pp.124Crawley, M.J. 1988. Herbivores and plant population dynamics. Chapter 18 In Plant PopulationEcology: The 28th Symposium of the British Ecological Society. Sussex 1988. Ed. A.J. Davy,M.J.Crawley, M.J. 1990. Rabbit grazing, plant competition, and seedling recruitment in acidgrassland. J. App. Ecol. 27:803-820.Cristie, P., E. Newman, and R. Campbell. 1978. The influence of neighbouring grassland plantson each others’ endomycorrhizas and root-surface microorganisms. Soil Biol. Biochem. 10:521-527.Davy, A.J. , G.F. Bishop. 1984. Response of Hieraceum pilosella in Breckland grass heath toinorganic nutrients. 3. Ecol. 72:319-330.Dc Kroon, H., Hara, T. , R. Kwant. 1992. Size hierarchies of shoots and clones in clonal andmonocultures:do clonal and non-clonal plants compete differently? Oikos 63 :4 10-419.Dennis, W.D. , J. Woledge. 1985. The effect of nitrogenous fertilizer on the photosynthesis andgrowth of white clover! perennial ryegrass swards. Ann. Bot. 85:171-178.Detling, J.K. , E.L. Painter. 1983. Defoliation responses of western wheatgrass populations withdiverse histories of prairie dog grazing. Oecologia:57:65-7 1.Diemer, M., Korner, C., S. Procks. 1992. Leaf lifespans in wild perennial herbaceous plants: asurvey and attempts at functional interpretation. Oecologia 89:10-16.Doak, D.F. 1992. Lifetime impacts of herbivory for a perennial plant. Ecol 73(6):2186-2099.Dolinger, P., P. Ehrlich, W. Fitch, D. Breedlove. 1973. Alkaloid and predation patterns inColorado Lupine populations. Oecologia 13:191-204.Donald, C.M. 1963. Competition among crop and pasture plants. Adv. Agron. 15:1-118.Donald, C.M. 1989. Invasive root growth into disturbed soil of two tussock grasses that differ incompetitive effectiveness. Funct. Ecol. 3:345-353.Dunn, D.B. , J.M. Gillet. 1966. The Lupines of Canada and Alaska. Can. Dept. Agric. Res.Branch Monograph 2. Queen’s Printer, Ottawa.Elliot-Fisk, D. 1993. The boreal forest. Chapter Two in North American Terrestrial Vegetation.ed. M.G. Barbour, and W.D. Billings. Cambridge University Press, New York.Eriksson, 0. 1986. Survivorship, reproduction and dynamics of ramets of Potentilla anserina ona Baltic seashore meadow. Oecologia 47:378-380.125Evans, J.P. 1992. The effect of local resource availability and clonal integration on rametfuncitonal morphology in Hydrocotyle bonariensis. Oecologia 89:265-276.Evans, R.C. 1986. Morphological variation in a biotically patchy environment: evidence from apasture population of TriXolium repens. M.Sc. thesis. University of British Columbia.Fetcher, N. 1985. Effects of removal of neighbouring species on growth, nutrients andmicroclimate of Eriophorum vaginatum. Arctic and Alpine Research 17(1):7-17.Firbank, L.G. , A.R.Watkisnon. 1985. On the analysis of competition within 2 species mixtures ofplants. J. App. Ecol. 22:503-5 17.Ford, H. 1981. The demography of three populations of dandelion. Biol. J. Linn. Soc. 15:1-11.Fowler, N. 1981. Competition and coexistence in a North Carolina grassland. II. The effects ofexperimental removal of species. J.Ecol. 69:843-854.Fowler, N. 1982. Competition and coexistence in a North Carolina grassland. Ill. Mixtures ofcomponent species. J.Ecol. 70:77-92.Fowler, N.L. 1984. The role of germination date, spatial arrangement and neighbourhood effectin competitive interactions in Linum perenne. J. Ecol. 72:307-318.Fowler, N.L. , M.D. Rausher. 1985. Joint effects of competitors and herbivores on growth, andreproduction in Aristolochia reticulata. Ecology 66:1580-1587.Fox, L.R. , P.A. Morrow. 1992. Eucalypt responses to fertilization and reduced herbivory.Oecologia 89:217-222.Gehan, E. 1969. Estimating survival functions from the life table. J. Chronic Diseases 21:629-644.Goldberg, D.E. , A.M. Barton. 1992. Patterns and consequences of interspecific competition innatural communities: a review of field experiments with plants. Am. Nat. 139(4) :771-801.Gosz, J. R. 1984. Biological factors influencing nutrient supply in forest soils. Chapter 5 inNjitrition of Plantation Forests. Academic Press, London.Grace, J.B. 1985. Juvenile vs. adult competitive abilities in plants: size dependence in cattails.Ecology 66(5): 1630-1638.Grace, J.B. , R.G. Wetzel. 1981. Effects of size and growth rate on vegetative reproduction inTypha. Oecologia 50:158-161.126Grace, J.B. , R.G. Wetzel. 1982. Variations in growth and reproduction within populations oftwo rhizomatous plant species: Typha latifolia and Typha angustifolia. Oecologia 53:258-263.Grime, J.P. 1977. Evidence for the existence of three primary strategies in plants and its relevanceto ecological and evolutionary theory. Am. Nat. 111:1169-1194.Grime, J.P. 1979. Plant Strategies and Vegetation Processes. John Wiley, Sons, London.Grubb, P.J. 1985. Plant populations and vegetation in relation to habitat, disturbance andcompetition: problems in generalizations. pp. 595-621 in J. White and J. Beeftink, ed. Thepopulation structure of vegetaton. Dr. W. Junk, Dordrecht, The Netherlands.Gurevitch, J. 1986. Competition and the local distributionof the grass Stipa neomexicana.Ecology 67(1 ):46-57.Gurevtich, J. , S.T. Chester. 1986. Analysis of repeated measures experiments. Ecology67(1):251-255.Gurevitch, J. , R.S. Unnasch. 1989. Experimental removal of a dominant species at two levels ofsoil fertility. Can. J. Bot. 67:3470-3477.Gurevitch, J., P. Wilson, J.L. Stone, P. Teese, , R.J. Stoutenburgh. 1990. Competition among oldfield perennials at different levels of soil fertility and available space. J. Ecol.78:727-744.Halvorsen, J.J., J.L. Smith, , E.H. Franz. 1991. Lupine influence on soil carbon, nitrogen, andmicrobial activity in developing ecosystems at Mt. St. Helens. Oecologia 87:162-170.Halvorson, J.J., E.H. Franz, J.L. Smith, , R.A. Black. 1992. Nitrogenase activity, nitrogenfixation and nitrogen imputs by lupines at Mt. St. Helens. Ecology 73(1):87-98.Harper, J.L. 1967. A Darwinian approach to plant ecology. J. Ecol. 55:247-270.Harper, J.L. 1977. Population Biology of Plants. Academic Press, New York. 891 pp.Harper, J.L. 1989. The value of a leaf. Oecologia 80:53-55.Hartnett, D.C. , F.A. Bazzaz. 1985. The genet and ramet population dynamics of Solidagocanadensis. J. Ecol. 73:407-413.Hawthorn, W.R. , P.B. Cavers. 1976. Population dynamics of the perennial herbs PlantagomajorL. and P. rueglii. J. Ecol. 64:511-527.Henry, G.H.R., D.C. Glenn-Lewin , J. Svoboda. 1986. Effects of fertilization on three tundraplant communities of a polar desert oasis. Can. J. Bot. 64:2502-2507.127Hobbs, RJ., S.L. Gulmon, V.J. Hobbs , H.A. Mooney. 1988. Effects of fertiliser addition andsubsequent gopher disturbance on a serpentine annual grassland community. Oecologia 75:291-295.Holland, E., W.J. Parton, J.K. Detling, , D.L. Coppock. 1992. Physiological responses of plantpopulations to herbivory and their consequences for ecosystem nutrient flow. Am. Nat.140(4):685-706.Holmes, R. , K. Jepson-Innes 1989. A neighbourhood analysis of herbivory in Bouteloua gracilis.Ecology 70(4)971-976.Horvitz, C.C. , D.W. Schemske. 1988. Demographic cost of reproduction in a neotropical herb:an experimental study. Ecology 69(6): 1741-1745.Huffaker, C.B. , C.E. Kennet. 1959. A 10 year study of vegetational changes associated withbiological control of Klamath weed species. J. Range. Man. 12:69-82.Hunter, M.D. , L.W. Aarssen. 1988. Plants helping plants. BioScience 38:34-40.Hunter, M.D. , P.W. Price. 1992. Playing chutes and ladders:heterogeneity and the relative rolesof bottom up and top down forces in natural communities. Ecology 73:1135Hutchings, M. , A.J. Slade. 1988. Morphological plasticity, foraging and integration in clonalperennial herbs. Chpt 6 in Plant Population Ecology: The 28th Symposium of the BritishEcological Society. Sussex 1987. Davy, AJ., M. Hutchings, , A.R. Watkinson, editors. BlackwellScientific Publications, Oxford. pp367-392.Hutchings, M.J., Booth, K. , S. Waite. 1991. Comparison of survivorship by the LogrankTest:criticisms and alternatives. Ecology 72(6):2290-2293.Jaramillo, V.J. , J.K. Detling. 1988. Grazing history, defoliation and competition on short grassproduction and nitrogen accumulation. Ecology 69(5): 1599-1608.Johnson, N.D. , B.L. Bentley. 1988. Effects of dietary protein and lupine alkaloids on growth andsurvivorship of Spodoptera eridania. J. Chem. Ecol. 14(5): 1391-1403.Johnson, N.D., B.Lin, , B.L. Bentley. 1987. The effects of nitrogen fixation, soil nitrate, anddefoliation on the growth, alkaloids and nitrogen levels of Lupinus succulentus. Oecologia74:425-431.Jon sdottir, I.S. 1991. Effects of tiller size and population dynamics in a clonal sedge Carexbigelowii. Oikos 62: 177-188.Jonasson, S. 1992. Plant response to fertilization and species removal in tundra related tocommunity structure and clonality. Oikos 63:420-429.128Jonsdottir, I.S. , T.V. Callaghan. 1989. Localized defoliation stress and translocation of 14C-photoassimilates between tillers of Carex bigelowii. Oikos 52:211-219.Karban, R. , S. Strauss. 1993. Effects of herbivores on growth and reproduction of theirperennial host, Erigeron glaucus. Ecology 74(1):39-46.Keddy, P.A. 1989. Effects of competition from shrubs on herbaceous wetland plants: a four yearfield experiment. Can. 3. Bot 67:708-7 16.Keddy, P.A. 1989. Competition. Population and community biology series. Chapman and Hall,London, England.Kotanen, P. , R.L. Jefferies. 1987. The leaf and shoot demography of grazed and ungrazed plantsof Carex subspathacea. J. Ecol. 75:961-975.Kotanen, P. , R.L. Jefferies. 1989. Responses of arctic sedges to release from grazing: leafdemography of Carex Xflavicans. Can. J. Bot. 67:1408-1413.Krebs, C.J. 1986. Population biology of snowshoe hares. I. Demography of food supplementedpopulations in southern Yukon 1976-1984. J. Anim. Ecol. 55:963-982.Larsson, S., A. Wiren, L.Lundgren, , T. Ericsson. 1986. Effects of light and nutrient stress on leafphenolic chemistry in Salix dasyclados and susceptibility to Galerucella lineola (Coleoptera).Oikos 47:205-2 10.Lee, E.T. 1980. Statistical methods for survival data analysis. Lifetime Learning Publications,Belmont, California.Lee, E.T. 1992. Statistical methods for survival data analysis. Second Edition. Wiley , Sons,Inc. New York.Lee, T.D. 1988. Patterns of fruit and seed production. Chpt. 9 in Plant Reproductive Ecology:Patterns and Strategy. Lovett-Doust, J , L. Lovett-Doust, editors. Oxford University Press, NewYork.Lee, T.D. , F.A. Bazzaz. 1980. Effects of defoliation, and competition on growth andreproduction in the annual plant Abutilon theophrasti. J. Ecol. 68:813-821.Lee, T.D. , F.A. Bazzaz. 1982. Regulation of fruit and seed production in an annual legumeCassiafasciculata. Ecology 63(5): 1363-1372.Lee, T.D. , F.A. Bazzaz. 1986. Maternal regulation of fecundity: non-random ovule abortion inCassiafasciculata. Ecology 68:459-465.129Lieffers, V.J. , S.J. Titus. 1989. The effects of stem density and nutrient status on size inequalityand resource allocation in lodgepole pine and white spruce seedlings. Can. J. Bot. 67:2900-2903.Loader, C. , H. Damman. 1991. Nitrogen content of food plants and vulnerability of Pieris rapaeto natural enemies. Ecology 72(5): 1586-1590.Louda, S. M. 1984. Herbivore effect on stature, fruiting and leaf dynamics of a native crucifer.J.Ecol. 59:767-783.Mack, R.N. 1976. Survivorship of Cerastium atrovirens at Aberffraw, Anglesey. J. Ecol. 64:309-312.Mack, R.N. , J.L. Harper. 1977. Interference in dune annuals: spatial pattern and neighbourhoodeffects. J. Ecol. 65:345-363.Maillette, L. 1992. Plasticity of modular reiteration in Potentilla anserina. J. Ecol. 80(2):231-239.Mauricio, R., M.D. Bowers, , F.A. Bazzaz. 1993. Pattern of leaf damage affects fitness of annualplant Raphanus sativus L.. Ecology 74(7) :2066-2071.McClure, M.S. 1980. Foliar nitrogen: a basis for host suitability for elongate hemlock scale,Fiorina externa (Homoptera: Diaspididae). Ecology 61(l):72-79.Meagher, T. R. , 3. Antonovics. 1982. Life history variation in dioecious plant populations: acase study of Chamaelirium luteum. pages 139-154 in H. Dingle , J.P. Hegmann eds. Evolutionand Genetics of Life Histories. Springer Verlag, New York.Mellinger, M.V. , S.J. McNaughton. 1975. Structure and function of successional vascular plantcommunities in central New York. Ecol. Monogr. 45:161-182.Mihaliak, C.A. , D.E. Lincoln. 1985. Growth pattern and carbon allocation to volatile leafterpenes under nitrogen-limiting conditions in Heterotheca subaxillaris (Asteraceae). Oecologia66:423-426.Mihaliak, C.A. , D.E. Lincoln. 1989. Plant biomass partitioning and chemical defence: response todefoliation and nitrate limitation. Oecologia 80:122-126.Muenchow, G. 1986. Ecological use of failure time analysis. Ecology 67(1):246-250Nams, V., Foilcard, , J.N.M. Smith. 1993. Effects of nitrogen fertilization on several woody andnonwoody boreal forest species. Can. J. Bot. 7 1:93-97.Newman, E.I. 1973. Competition and diversity in herbaceous vegetation. Nature 244:310.130Noble, J.C., A.D. Bell, , J.L. Harper. 1979. The population biology of plants with clonal growth.I. The morphology and structural demography of Carex arenaria. J. Ecol. 67:983-1008.Onuf, C.P., J.M. Teal, , I.Valiela. 1977. Interactions of nutrients: plant growth and herbivory in amangrove ecosystem. Ecology 58:514-526.Owen, D.F. 1980. How plants may benefit from the animals that eat them. Oikos 35:230-235.Pacala, S.W. , M.J. Crawley. 1992. Herbivores and plant diversity. Am. Nat. 140(2):243-260.Paimblad, G.S. 1968. Competition in experimental populations of weeds with emphasis onregulation of population size. Ecology 49:26-34.Perry, D.A., M.P. Amaranthus, J.G. Borchers, S.L. Borchers, , R.E. Brainerd. 1989.Bootstrapping in ecosystems. Bioscience 39(4):230-236.Pitelka, L.F. , J.W. Ashmun. 1985. Physiology and integration of ramets in clonal plants. Pages379-435 in Population biology and evolution of clonal organisms. J.B.C. Jackson, L.W. Buss,R.E. Cook, editors. Yale University Press, New Haven.Polley, H.W. , J.K. Detling. 1989. Defoliation, nitrogen and competition: effects on plant growthand nitrogen nutrition. Ecology 790(3):721-727.Potvin, C., M.J. Lechowicz, , S. Tardif. 1990. The statistical analysis of ecophysiologicalresponse curves obtained from experiments involving repeated measures. Ecology 71(4):1389-1400.Price, P.W. 1991, The plant vigor hypothesis and herbivore attack. Oikos 62:244-25 1.Pyke, D.A. 1986. Demographic responses of Bromus tectorum and seedlings of Agropyronspicatum to grazing by small mammals: occurrence and severity of grazing. J. Ecol. 74:739-754.Pyke, D.A. , J.N. Thompson. 1986. Statistical analysis of survival and removal rate experiments.Ecology 67(1) :240-245.Pyke, D.A. , J.N. Thompson. 1987. Erratum. Ecology 68:232.Rausher, M.D. 1981. Host plant selection by Battus philenor butterflies: the roles of predation,nutrition, and plant chemistry. Ecological Monographs 51:1-20.Rausher, M.D. , P. Feeny. 1980. Herbivory, plant density and plant reproductive success: theeffect of Battus philenor on Aristolochia reticulata. Ecology 6 1(4): 905-917.Reader, R. 1992. Herbivory, competition, plant mortality and reproduction on a topographicgradient. Oikos 65:414-418.131Reekie, E.G. , F.A. Bazzaz 1987a. Carbon allocation to reproduction. Am. Nat. 129(6):876-896.Reekie, E.G. , F.A. Bazzaz 1987b. Reproductive effort in plants. U. Does carbon reflectallocation of other resources? Am. Nat. 129(6):897-906.Reekie, E.G. , F.A. Bazzaz 1987c. Reproductive effort in plants. Ill. Effect of reproduction onvegetative activity. Am. Nat. 129(6) :907-917.Reekie, E.G. , F.A. Bazzaz. Cost of reproduction in genotypes of two congeneric plant specieswith contrasting life histories, unpublished manuscript.Rees, M. , V.K. Brown 1992. Interactions between invertebrate herbivores and plantcompetition.Rousi, M., J. Tahvanainen, H. Hentonen, , I. Votila. 1993. Effects of shading and fertilization onresistance of winter dormant birch (Betula pendula) to voles and hares. Ecology 74(1):30-38.Ruess, R.W, S.V. McNaughton, , M.B. Coughenour. 1983. The effects of clipping, nitrogensource and nitrogen concentration on growth responses and nitrogen uptake of an East Africansedge. Oecologia 59:253-261.Sarukhan, J. 1974. Studies on plant demography: Ranunculus repens L., R. bulbosus L., andR..acris L. 3. Ecol. 62:151-177.Sarukhan, J. , J.L. Harper. 1973. Studies on plant demography: Ranunculus repens L., R.bulbosa L. and R. acris. I. Population flux and survivorship. 3. Ecol. 61:675-7 16.SAS Institute. 1979. SAS Procedures Manual. SAS Institute. Raleigh, N.C.Schmid, B. , F.A. Bazzaz. 1987. Clonal integration and population structure in perennial plants:effects of severing rhizome connections. Ecology 68:20 16-2022.Schmid, B. , F.A. Bazzaz. 1990. Plasticity in plant size and architecture in rhizome derived vs.seed derived Solidago and Aster. Ecology 71(2):523-535.Schmid, B. , F.A. Bazzaz. 1992. Growth responses of rhizomatous plants to fertilizer applicationand interference. Oikos 65:13-24.Schmid, B., Puttick, G.M., Burgess, K.H. , F.A. Bazzaz. 1988. Correlation between genetarchitecture and some life history features in 3 species of Solidago. Oecologia 75:459-464.Schmid, B., G.M. Puttick, K.H. Burgess, , F.A. Bazzaz. 1988. Clonal integration and the effectof simulated herbivory in old field perennials. Oecologia 75:465-47 1.132Schmid, B., Puttick, G.M., Burgess, K.H. , F.A. Bazzaz. 1988. Correlation between genetarchitecture and some life history features in 3 species of Solidago. Oecologia 75:459-464.Shaver, G.R. , F.S. Chapin. 1980. Response to fertilization by various plant growth forms in anAlaska tundra: nutrient accumulation and growth. Ecology 61(3):662-675.Shaver, G.R. 1983. Mineral nutrition and leaf longevity in Ledum palustre: the role of individualnutrients and the timing of leaf mortality. Oecologia 56:160-165.Silander, J.A. , S.W. Pacala. 1985. Neighbourhood predictors of plant performance.Smith, J.N.M., C.J. Krebs, A.R.E. Sinclair, R. Boonstra. 1988. Population biology of snowshoehares. II. Interactions with winter food plants. J. Anim. Ecol. 57 :269-286.Snow, A.A. , D.F. Whigham. 1989. Cost of flower and fruit production in Tipularia discolor(Orchidaea). Ecology 70(5): 1286-1293.Soibrig, O.T. 1981. Studies on the population biology of the genus Viola II. The effect of plantsize on fitness in Viola soloria. Evolution 35:1080-1093.Stephenson, A.G. 1980. Fruit set, herbivory, fruit reduction and the fruiting strategy of Catalpaspeciosa. Ecology 61(1):57-64.Stephenson, A.G. 1984. The regulation of maternal investment in an indeterminant floweringplant Lotus corniculatus. Ecology 65(1): 113-121.Swank, S. , Oechel, W.C. 1991. Interactions among the effects of herbivory, competition andresource limitaiton on chaparral herbs. Ecology 72(1): 104-115.Sydes, C.L. 1984. A comparative study of leaf demography in limestone grassland. J. Ecol.72:33 1-345.Tanner, E.V.J., V. Kapos, , W. Franco. 1992. Nitrogen and phosphorus fertilizer effects onVenezuelan montane forest trunk growth and litterfall. Ecology 73:78-86.Tansley, A.G. , R.S. Adamson. 1925. Studies of the vegetation of the English chalk. III. Thechalk grasslands of the Hampshire-Sussex border. J. Ecol. 13:177-223.Thurston, J.M. 1968. The effect of liming and fertilizers on the botanical composition ofpermanent grassland, and on the yield of hay. In Rorison, I.H. ed. Ecological aspects of themineral nutrition of plants. Blackwell Scientific Pubi., Oxford.Tilman, D. 1987. On the meaning of competition and the mechanisms of competitive superiority.Funct. Ecol. 1:304-3 15.133Tilman, D. 1988. Plant strategies and the dynamics and structure of plant communities.Monographs in population biology. Princeton University Press, Princeton, New Jersey.Traczyck, T., Taczyk, H. , D. Pasternak-Kusmierska. 1984. Reaction of meadow vegetation afterseven years of intense inorganic fertilization. Ekol. Pol. 32:581-596.Trudgill, S. T. 1979. Soil and Vegetation Systems: Contemporary Problems in Geography.Oxford University Press, Oxford.Underwood, A.J. 1981. Techniques of analysis of variance in experimental marine biology andecology. Oceanogr. Mar. Biol. Ann. Rev. 19:513-605.Wailer, D.M. 1981. Neighbourhood Competition in several violet populations. Oecologia51:116-122.Watkinson, A.R., Huiskes, A.H.L. , J.C. Noble. 1979. The demography of sand dune species withcontrasting life cycle. In Ecological processes in Coastal Environments. Ed. by R.L. Jefferies,,A.J.Watson, M.A.1984. Developmental constraints: effect on population growth and patterns ofresource allocation in clonal plants. Am. Nat. 123:411-426.Weibull, W. 1931. A statistical theory of the strength of materials. Ingeniors VetenskapsAkademien Handlingar: the phenomenon of rupture in solids. No. 151: pp 28 1-298.Weibull, W. 1959. A statistical distribution of wide applicability. J. Appl. Mech. 18:293-197.Weiner, J. 1982. A neighbourhood model of annual plant interference. Ecology 63(5):1237-1241.Weiner, J. 1984. Neighbourhood interference amongst Pinus rigida individuals. J. Ecol. 72:183-185.Weiner, J. 1985. Size hierarchies in experimental populations of annual plants. Ecology66(3) :743-752.Wiens, J.A. 1977. On competition and variable environments. Am. Sci. 65:590-597.Wiens, l.A. 1984. On understanding a non-equilibrium world: myth and reality in communitypatterns and processes. In D.R. Strong, D. Simberloff, L.G. Abele, and A.B. Thistle, ed.,Ecological Communities: Conceptual Issues and the Evidence. Princeton University Press,Princeton, N.J.Wennstron, A. , L. Ericson. 1991. Variation in disease incidence in grazed and ungrazed sites forthe system Pulsatilla pratensis-Puccinia pulsatillae. Oikos 60:35-39.134Whigham, D.F. 1984. The effect of competition and nutrient availability on the growth andreproduction of Ipomea hederacea in an abandoned old field. J. Ecol. 72:721-730.White, J. 1979. The plant as a metapopulation. Ann. Rev. Ecol. Syst. 10:109-145.White, J. 1985. The census of plants in vegetation. In The Population Structure of Vegetation.ed. J. White. pp. 34-88. Dr W. Junk Publishers, Dordrecht.White, J. , J.L. Harper. 1970. Correlated changes in plant size and number in plant populations. J.Ecol. 58:467-477.Wit, C.T. de. 1960. On competition. Versi. Landbouwk. Onderz. 66. 1-82.Wilcox, A., , M.J. Crawley. 1988. The effects of host plant defoliation and fertilizer application onlarval growth and oviposition behaviour in cinnabar moth. Oecologia 76:283-287.Wilson, S.D. , P.A. Keddy. 1986. Diffuse competition along an environmental gradient: resultsfrom a shoreline plant community. Am. Nat. 127(6):862-869.Wilson, K.G. , R.E. Stinner. 1984. A potential influence of rhizobium activity on the availabilityof nitrogen to legum herbivores. Oecologia 61:337-341.Wolfe, L.M. 1983. The effect of plant size on reproductive characteristics in Erythroniumarnericanum (Liliaceae). Can. J. Bot. 61:3489-3493.Yoda,K., T. Kira, H. Ogawa, K. Hozimu. 1963. Self-thinning in overcrowded pure stands undercultivated and natural conditions. J. Biol. Osaka Cy University.Zar, J.H. 1984. Biostatistical Analysis. Second Edition. Prentice-Hall, Inc., Englewood Cliffs,NJ.Zarzycki, K. 1983. The competitive performance of grassland plants on acid soils as influencedby fertilization and with different plant competitors in semi-natural meadows in the PieninyNational Park, Poland. Verhandi. Ges. Okol. 11:505-509.Zimmerman, J.K., I.M. Weis. 1984. Factors affecting survivorship, growth, and fruitproduction in a beach population of Xanthium stromarium. Can. J. Bot. 62:2122-2127.

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.831.1-0087483/manifest

Comment

Related Items