Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Gap-phase community dynamics in a sub-alpine old growth forest Lertzman, Kenneth Peter 1989

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

Item Metadata

Download

Media
831-UBC_1989_A1 L47_5.pdf [ 8.39MB ]
Metadata
JSON: 831-1.0098316.json
JSON-LD: 831-1.0098316-ld.json
RDF/XML (Pretty): 831-1.0098316-rdf.xml
RDF/JSON: 831-1.0098316-rdf.json
Turtle: 831-1.0098316-turtle.txt
N-Triples: 831-1.0098316-rdf-ntriples.txt
Original Record: 831-1.0098316-source.json
Full Text
831-1.0098316-fulltext.txt
Citation
831-1.0098316.ris

Full Text

GAP-PHASE COMMUNITY DYNAMICS IN A SUB-ALPINE OLD GROWTH FOREST By KENNETH PETER LERTZMAN B.Sc. (hons.), The University of Manitoba, 1978 M.Sc, The University of British Columbia, 1981 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Zoology)  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA, September 1989 © Kenneth Peter Lertzman, 1989  *  h  <  In  presenting  degree freely  at  this  the  available  copying  of  department publication  of  in  partial  fulfilment  of  the  University  of  British  Columbia,  I  agree  for  this or  thesis  reference  thesis by  this  for  his thesis  and  scholarly  or for  her  Department  Date  DE-6 (2/88)  Columbia  I  further  purposes  gain  that  agree  may  be  It  is  representatives.  financial  permission.  T h e U n i v e r s i t y o f British Vancouver, Canada  study.  requirements  shall  not  that  the  Library  permission  granted  by  understood be  for  allowed  an  advanced  shall for  the that  without  make  it  extensive  head  of  my  copying  or  my  written  ii ABSTRACT Small-scale natural disturbances involving the death of one to a few trees and creating gaps in the forest canopy are critical to the population and community ecology of many forest types. I studied the role of canopy gaps in the structure and dynamics of a high-elevation, old growth forest in coastal British Columbia. The research was conducted in four stands at Cypress Provincial Park (~1,100 m) occupying the transition from the upper montane ecosystems of the Coastal Western Hemlock Zone to the lower elevations of the Mountain Hemlock Zone. They contained four tree species: Pacific silver fir (Abies amabilis). western hemlock (Tsuqa heterophylla). mountain hemlock (Tsuga mertensiana). and Alaska yellow-cedar (Chamaecyparis nootkatensis). Though the stands varied in the proportions of each species, all had a similar distribution of area underdosed canopy (29%) and in gaps (52% expanded gap; 18% canopy gap). The overstory of two stands was dominated by Pacific silver fir and western hemlock, and two had a more equitable distribution of species. Pacific silver fir was much more dominant in the sapling layer (81%) than the canopy layer (43%) in all stands. Western hemlock was the next most frequent species among saplings (13%). Growth rate among Pacific silver fir saplings was greater in both classes of gap than it was under closed canopy. The distribution of saplings was independent of canopy classes, but there was a significant interaction between sapling species and rooting substrate. Western hemlock saplings were largely restricted to stumps, whereas firs occurred on stumps and on the forest floor. Most fir saplings (81 %) showed evidence of suppression, and 23% had experienced multiple periods of suppression and release. Most gaps had more than one gapmaker (90%). Half of all gapmakers died standing, and only 13% were windthrown. Pacific silver fir was represented among gapmakers in a much higher proportion than among canopy trees in general (64% vs. 45%). Median canopy and expanded gap areas were 41 and 203 m2, respectively. The estimated forest turnover time varied from 280-1000 years depending on  iii assumptions about the time taken for gaps to be filled. The most likely range is 600700 years. I examined methods for calculating turnover time by comparing estimates from the literature and from simulations. The two common methods for estimating forest turnover time produce different estimates for stands where both can be calculated. The inverse of the rate of creation of new gap area (TT1) is consistently lower than the estimate based on total gap area and the time taken for gaps to fill (TT2). Tropical forests appear to have faster turnover times than do temperate ones, but comparisons are confounded by differences in estimation method. I found both TT1 and TT2 to be sensitive to assumptions about the equilibrium structure of the forest. TT2 reaches its steady state value later in stand development, but TT1 is more sensitive to short term fluctuations in gap creation rate. I examined the species composition of gapfillers with respect to two sets of hypotheses. The first set related the species of each gapfiller to the species of the gapmaker it is replacing. I did not find evidence that self-replacement or reciprocalreplacement act to maintain the current community composition. Gapmaker-gapfiller comparisons indicated preferential replacement of all species by Pacific silver fir, suggesting that the community is in a state of change. The second set of hypotheses related the species composition of gapfillers to features of the gap environment. I did not find evidence to support the ideas that gap size, location within a gap, or local canopy composition exert a strong influence on the species composition of regeneration within the gap. The only circumstance where Pacific silver fir was not overwhelmingly dominant among gapfillers was on stumps, where almost all successful western hemlock gapfillers were located. These patterns suggest that neither species-specific interactions between gapmakers and gapfillers nor variability in gap environments is adequate to maintain the current composition of the forest canopy.  (  iv I used the matrix of species-specific transitions between gapmakers and gapfillers as a basis for modelling the longer term consequences of these replacement patterns. The models incorporated species-specific mortality rates to express differential longevity and variation in the transition probabilities to represent climatic fluctuations. I found that differential longevity among species exerted a strong influence on the equilibrium species composition, on the rate of community change, and the time to equilibrium. When the differential in mortality rates among species is proportional to their representation among gapmakers, the equilibrium composition is close to the current canopy composition. Because of its higher representation among gapmakers, Pacific silver fir may not be increasing in the canopy, despite its dominance among gapfillers. The simulations with varying climate produced 4 main results. 1) Unless the duration of climate fluctuations is either very long or very short, forest composition is in a continual state of disequilibrium. 2) Because of differential longevity, species have different response times to changes in climate. 3) Because of the difference in response times, the mean abundance of each species under a varying climate scenario is different than the composition expected from the mean climate state. 4) The rare, long-lived species, Alaska yellow-cedar, was favored by climatic fluctuations at the expense of the more common shorter lived species, Pacific silver fir. In this system, current canopy composition could be maintained by a combination of differential longevity among species and climatic fluctuations allowing periodic recruitment of the less common species.  V TABLE OF CONTENTS ABSTRACT  ii  TABLE OF CONTENTS  v  LIST OF TABLES  vii  LIST OF FIGURES  ix  ACKNOWLEDGEMENTS  xi  1. GENERAL INTRODUCTION TO THE IDEAS AND THE STUDY SYSTEM . . 1 1.1 Introduction 1 1.2 Gap-Phase Processes 2 1.3 Study Sites and Species 6 Species Autecology 7 Research Stands 9 1.4 Synopsis of Chapters 10 2. GAP-PHASE STRUCTURE OF A SUB-ALPINE OLD GROWTH F O R E S T . 2.1 Introduction 2.2 Methods.. . Sampling for canopy structure Sampling For Saplings Description of Individual Gaps Stand vs. forest level analyses.. 2.3 Results Percent of Forest in Each Canopy and Substrate Category Composition and Structure of the Canopy Layer Composition, Structure and Distribution of the Sapling Layer Diversity of the understory layer Number of Gapmakers Per Gap and Mode of Mortality Gap Size and Geometry 2.4. Discussion Percent of the forest in each canopy category Size and age structure of the canopy trees The sapling layer Species diversity of the understory layer Similarities and differences among stands Gapmakers: mode of mortality and number per gap Size of gaps Forest Turnover Time  14 14 .16 16 19 19 22 22 22 23 26 .28 29 32 33 .33 34 35 37 37 38 39 40  3. ON ESTIMATING FOREST TURNOVER TIMES 3.1 Introduction 3.2 One Parameter Method: The Inverse of Gap Birth Rate 3.3 Two Parameter Method: Time to Fill Plus Residence Time 3.4 Comparing the Two Methods: Data Comparing TT1 and TT2 Comparing Estimates From Tropical and Temperate Forests 3.5 Comparing The Two Methods: Simulations The Model Model Results 3.6 What is the "Real" Turnover Time?  .70 .70 72 73 74 74 75 76 76 -78 80  vi 4. PATTERNS OF GAP-PHASE REPLACEMENT IN A SUB-ALPINE, OLD GROWTH FOREST 4.1 Introduction Hypotheses about the identity of gapfillers: gapmaker-gapfiller comparisons Hypotheses about the identity of gapfillers: gap-envrironment - gapfiller interactions 4.2 Methods General gapfiller data and definitions Constructing the species-specific transition matrix Gapfiller species and gap characteristics 4.3 Results Composition of the overall gapfiller population Gapmaker-gapfiller transition matrix: Hypotheses 1-4 Gapfiller species and gap size Gapfiller species and local canopy composition Gapfiller species and location in gap Gapfiller species and substrate 4.4 Discussion Hypotheses relating gapfiller species to gapmaker species Predictions for gapfillers and gap environment at Cypress Provincial Park Hypotheses relating gapfiller species to gap environment Gap-phase mediated coexistence?  92 .92 93 93 .94 94 95 96 97 97 97 99 102 103 104 105 105 105 106 110  5. MODELLING LONG-TERM FOREST COMMUNITY CHANGE BASED ON GAP-PHASE TRANSITIONS 124 5.1 Introduction 124 5.2 Methods and Description of the Models 126 General gapfiller data and definitions 126 The Markov Model 127 Calculating Differential Mortality Rates 127 Simulations with non-stationary matrix Changing environmental conditions 129 5.3 Results 131 Equilibria and time to equilibrium with and without differential mortality 131 Equilibria produced by warm climate and cold climate matrices . . .134 The effects of changing climate and climate-differential mortality interactions 134 5.4 Discussion 137 Long-term dynamics of the Cypress Park system Assessing Change in the Field 138 Applicability of Results to Other Systems 139 Factors Not Included in the Model 141 6. CONCLUSIONS  152  7. LITERATURE CITED  155  vii LIST OF TABLES 1.1 Characteristics of the four focal stands  12  2.1 Decay classes for gapmakers  44  2.2 Frequency of canopy gap, expanded gap, and closed canopy among the four stands  .45  2.3 Species composition of canopy trees  46  2.4 Comparison of the mean diameters (at breast height) of canopy trees of all species among the four stands 2.5 Species composition of the sapling layer  .48 49  2.6 Densities, heights, and growth rates of the sapling layer under each canopy category  51  2.7 Frequency of different growth patterns in the sapling population  52  2.8 Distribution of saplings of each species among canopy classes  53  2.9 Distribution of saplings of each species among substrates  54  2.10 Number of gapmakers of each species, by type of mortality  55  2.11 Number of gapmakers in each age/decay class, by type of mortality  56  2.12 Summary statistics for gap size measures  57  2.13 Forest turnover times calculated according to equation 3.2, given 18% canopy gap in the forest 3.1 Parameters and turnover times from several tropical  58  and temperate forests 3.2 Summary statistics for Turnover parameters 3.3 Steady state turnover times for simulations with varying • input parameters 4.1 Number of gaps and gapmakers from each stand used to construct the matrix of transition frequencies 4.2 Matrix of transition frequencies between species of gapmakers and gapfillers  83 84 85 112 113  5.1 Annual mortality rates for each species in varying models  143  5.2 Proportions of each species among primary and secondary gapmakers combined, canopy trees, and definitive gapfillers 5.3 Matrices of transitions between species of gapmakers and gapfillers for 3 alternative climate states  144 145  VIII  5.4 Equilibrium species compositions for several versions of the Markov model  146  ix LIST OF FIGURES 1.1 Location of Cypress Provincial Park and the focal study stands  13  2.1 Definitions of canopy gap, expanded gap, and closed canopy  59  2.2 Size distributions of canopy trees  60  2.3 Ages of canopy trees  61  2.4 Decadal increments of basal area for 6 representative canopy trees  62  2.5 Height vs. age for dead, suppressed Pacific silver firs  63  2.6 Changes in vascular plant species diversity of the understory layer among canopy classes  64  2.7 Changes in the diversity of understory layer among stands  65  2.8 Frequency distribution of the number of primary and total gapmakers per gap 2.9 Ratio of Pacific silver fir to other species among gapmakers of different age-classes  66 67  2.10 Size frequency distributions for canopy gaps, expanded gaps, and gap aperture  68  2.11 Area-perimeter relationships for expanded gaps and canopy gaps  69  3.1 Plot of equation 3.2: TT2  86  3.2 Output from baseline simulation  •  3.3 Change in TT1 and TT2 over time with varying durations of Tfill  87 88  3.4 Change in TT1 and TT2 over time with varying durations of Tfill, fast turnover time scenarios 3.5 Change in TT1 and TT2 over time, showing true variation in TT1 3.6 Change in TT1 and TT2 over time with two 20 year fluctuations in mortality rate 4.1 Proportion of each species among definitive gapfillers (DGFS) and the overall gapfiller population (GFS) 4.2 Log number of gapfillers of Pacific silver fir (PSF) and western hemlock (WH), and proportion of Pacific silver fir among gapfillers vs. log expanded gap area 4.3 Log density of gapfillers of Pacific silver fir (PSF) and western hemlock (WH), a., and of Alaska yellow-cedar (AYC) and mountain hemlock (MH), b., on log expanded gap area  89 90 91 114 115  116  X 4.4 Log numbers of gapfillers vs. expanded gap area for Alaska yellow-cedar and mountain hemlock  117  4.5 Log numbers of definitive gapfillers vs. log expanded gap area for Pacific silver fir and western hemlock, and proportion of Pacific silver fir among definitive gapfillers  118  4.6 Log numbers of definitive gapfillers vs. expanded gap area for Alaska yellow-cedar and mountain hemlock  119  4.7 Boxplots of the size distribution of gaps containing at least one definitive gapfiller of each species  120  4.8 Percent of each species among definitive gapfillers each gap vs its representation in the canopy surrounding the gap  121  4.9 Species composition among definitive gapfillers in canopy and expanded gaps  122  4.10 Species composition among definitive gapfillers on the forest floor and on stumps  123  5.1 Total community difference from equilibrium composition: models with empirical transition matrix  147  5.2 Total community difference from equilibrium composition: cold climate and warm climate transition matrices  148  5.3 Forest change with equal mortality rates and cyclic climate  149  5.4 Forest change with differential mortality rates among species and cyclic climate 5.5 Change in coefficient of variation, by species, as the duration of each climate state increases  150 151  xi ACKNOWLEDGEMENTS I would like to thank my advisors C.J. Krebs and D. Ludwig for their continuing support, advice, and enthusiasm. J.P. Kimmins was indispensable; this work would have been much more difficult without him. I would also like to thank my other committee members: W.E. Neill and A. Black for helpful suggestions and stimulating discussions. I was ably assisted in the field by C. Trethewey, B. Walters, E. Kellerhals, and P. Freile. I would especially like to thank B. Walters for the summer he worked as a volunteer. Many other people helped in the field or provided thoughtful suggestions at critical times, and I am grateful to them all. In particular, D. Marmorek and W. Kurz have been constant sounding boards and sources of advice. D. Sprugel read Chapter 3 and provided critical and appreciated suggestions. Reviews by F. Bunnell, D. Lavender and T. Veblen were helpful in preparing the final draft. Others who have helped in the field are: G. Sutherland, J. Sutherland, N. Butler, P. Morrison, J. Lertzman, M. Lertzman, and J.E. Lertzman. I am grateful to the B.C. Ministry of Parks for permission to work at Cypress Park, for their interest and enthusiasm, and to Cypress Bowl Recreations Inc. for their interest and cooperation. MacMillan Bloedel Woodland Services Division, provided access to high elevation sites on Vancouver Island, which are not referred to in the thesis, but were useful in the development of my thinking. I am grateful to them for sponsoring my G.R.E.A.T. Fellowship. In particular, I thank G. Dunsworth for his help and thoughtful discussions. I was supported by a Killam Predoctoral Fellowship, an NSERC Postgraduate Scholarship, an H.R. MacMillan Family Fellowship, a B.C. Science Council G.R.E.A.T. Fellowship, and Joy and Morley Lertzman. Without the generous support of these agencies, I could not have conducted this research.  xii Finally, I would like to acknowledge the unending help of D. Lepofsky, in the field and in the city, without whom this would have been much less interesting, much less fun, and much more difficult.  1 1. GENERAL INTRODUCTION TO THE IDEAS AND THE STUDY SYSTEM  1.1  Introduction. We can contrast two perspectives through which vegetation dynamics are  viewed: that of succession, focussing on directional change in community composition, and that of co-existence, focussing on the processes maintaining community composition. The study of successional change in vegetation has a long history and lies at the core of much ecological thinking (Thoreau 1860; Clements 1916, 1936; Gleason 1917, 1926; Watt 1947; Whittaker 1953; Odum 1969; Drury and Nisbet 1973; West, Shugart and Botkin 1981). Recent studies of succession have focussed on individual species' life histories and population dynamics as explanatory variables for patterns of successional community change (Drury and Nisbet 1973; Connell and Slatyer 1977; Peet and Christenson 1980; Pickett 1982). There has also been an increasing emphasis on the role of disturbances in changing the context in which species' life histories are played out, resulting in the maintenance of early successional species in communities from which they would otherwise be lost (Connell 1978; Grubb 1977; Denslow 1980, 1987; Sousa 1984). The varying roles of disturbance in different communities have come to play a central role linking the successional and co-existence perspectives. For instance, Glitzenstein et al. (1986) found that small-scale disturbance in an east Texas forest accelerated succession towards later serai stages under some conditions, but reinitiated earlier successional stages under others. This resulted in long-term coexistence when averaged overtime and space. Similarly, long-term coexistence of Englemann spruce and sub-alpine fir in Colorado may result from the interaction between disturbances at several spatial scales and intensities with idiosyncratic patterns of stand history and demography (Veblen 1986; Aplet et al. 1988). While early explanations for co-existence were based on equilibrium assumptions about tightly linked competitive communities, the theoretical basis for co-existence now explicitly  2 recognizes the role of environmental variability and the non-equilibrium maintenance of communities (see reviews by Chesson and Case 1986 and Chesson 1986). In many cases, processes traditionally associated with succession, dispersed patchily in space and time, appear fundamental to coexistence. In forest communities where large scale, stand destroying disturbances are infrequent, small disturbances to the forest canopy caused by the mortality of one to a few trees provide the main opportunity for recruitment of new individuals (Runkle 1981, 1982, 1985; Brokaw 1985a,b; Whitmore 1978, 1989). Such openings are referred to as gaps, and the replacement processes in them as gap-phase succession (Jones 1945; Watt 1947; Bray 1956). Gap-phase processes have become a prime focus of research on the dynamics of both tropical and temperate forests (for example, see the papers comprising the special feature in Ecology 70:535-576). In this thesis, I will examine the role of small-scale gap-forming disturbances in the structure and dynamics of an old growth, sub-alpine forest in coastal British Columbia.  1.2 Gap-Phase Processes. The importance of gap-phase processes has long been recognized in a variety of forest types (Aubreville 1938; Jones 1945; Watt 1947; Richards 1952; Bray 1956). However, in the last 15 years they have become central to our understanding of the population and community ecology of many forests, and of the evolution of tree life histories (Hartshorn 1978; Whitmore 1978, 1989; Denslow 1980, 1985, 1987; Runkle 1981, 1982, 1985; Brokaw 1985a,b; Hubbell and Foster 1986a,b,c; Schupp et al. 1989). The direct effects of gap creation are 1) the physical removal of the tree crown(s) from the forest canopy, 2) changes in microclimate at ground level (Bazazz and Pickett 1980; Chazdon and Fletcher 1984; Denslow 1987), 3) physical disturbance to the soil and understory layer by falling bark, bole and branches (Putz et al. 1983; Orians 1982), and 4) disturbance to the understory layer and changes in soil structure and chemistry caused by windthrown stumps (Lutz 1940; Beatty and Stone 1986; Mladenhoff 1987).  3 Besides the effects on vegetation dynamics discussed below, the indirect effects of gap creation on animal communities are only just being recognized (Schemske and Brokaw 1981; Crome and Richards 1988; Feinsinger et al. 1988; Levey 1988; Shelly 1988; Walters and Lertzman in prep.). In general, the creation of a gap results in an increase in the levels of a variety of resources; gaps are brighter, warmer, and wetter than surrounding areas underneath a closed canopy. Competition may be reduced immediately after gap creation, but can quickly become intense among the seedlings and saplings competing for space and resources within gaps. Survival is greatest among individuals present shortly after a gap is created, and declines rapidly among later colonists (Brokaw 1985b). While there is good experimental evidence for intense belowground competition within intact stands (Tourney and Keinholz 1931; Korstian and Coile 1938; Christy 1986), and it seems reasonable that a canopy opening aboveground will be reflected in the distribution of resources belowground, the belowground status of gaps remains unclear (Vitousek and Denslow 1986; Mladenhoff 1987; Uhl et al. 1988). Gaps may be filled by five basic processes, reflecting the different origins of trees filling the gaps (Feinsinger 1989; Schupp et al. 1989): 1) release of previously established, suppressed individuals (Uhl et al. 1988), 2) germination of new individuals from the soil seed bank or seed rain (Garwood 1983; Putz 1983; Lawton and Putz 1988), 3) sprouting of the broken stems of the trees which created the gap {gapmakers) (Putz and Brokaw 1989), 4) lateral growth of adjacent canopy trees (Runkle 1982; Runkle and Yetter 1987), 5) or, in the tropics, growth of individuals that were previously epiphytes in the canopy of the gapmaker (Lawton and Putz 1988). In most cases, most of the new recruits filling gaps (gapfillers) will originate from the first or second of these categories. Gaps vary substantially in the degree of disturbance to the canopy, forest floor, and soil, and in the degree of modification to the ground-level microclimate (Denslow 1987). The source of variation among gaps that has received the most attention is gap  4  size. Gap size is a function of crown geometry, the number of gapmakers, the modes of mortality of gapmakers, and gap age (Runkle 1982; Putz et al. 1983; Brokaw 1985 a,b). Because various aspects of microclimate are strongly affected by gap size, even within an otherwise homogeneous stand, variation in gap sizes can provide conditions for the establishment of a variety of species. Space within gaps may be partitioned among species further by the restriction of some species to particular substrates or locations within gaps (i.e. exposed mineral soil or nurse logs, center vs. periphery of the gap, etc.; Orians 1982; Putz 1983; Putz et al. 1983; Lawton and Putz 1988; Brokaw 1985a). Most of the theories of the maintenance of species diversity by gap-phase processes are based on such differences in the regeneration niches of gapfilling species (life-history characters and requirements for reproduction and the early stages of establishment and recruitment; Grubb 1977; though see Hubbell and Foster 1986a,b for an alternate opinion). Species with similar regeneration niches have been classed as members of the same regeneration guild (Denslow 1980, 1987; Bazazz 1983; Hubbell and Foster 1986b). Several regeneration guilds have been identified for gapfilling species in the tropics, spanning the range from understory and small-gap specialists to large gap specialists and ruderal species (Denslow 1980, 1987; Bazazz 1983; Hubbell and Foster 1986b; Whitmore 1989). These can be portrayed dichotomously as small-gap specialists and large-gap specialists: Small-gap specialists ("climax" or "primary" tree species) may establish under a closed canopy, but require a gap to recruit to the canopy. They are shade-tolerant, have low growth rates, variable modes of dispersal, large seeds and relatively low seed productivity. Recruits in gaps have often spent a substantial period of time in a suppressed state prior to release. Because they can establish prior to gap formation, these trees may often take advantage of small gaps that would be closed laterally by adjacent canopy trees before a new seedling could recruit. They can also take  5 advantage of the intermediate light levels in small gaps or at the edges of larger ones (Canham 1988b, 1989). Large gap specialists (pioneer tree species) are "shade-intolerant", well dispersed, fast growing, and depend on reaching the canopy during the lifetime of a single gap. Of these, "early pioneer" species tend to be short in stature, small-seeded (with longer viability in the soil seed bank), and short-lived. "Late pioneer" species tend to be large in stature, large seeded (with shorter viability in the soil seed bank), and long-lived. Truly impressive growth rates have been recorded for early pioneer species growing in large gaps (eg. up to 7 m yeaH by Trema micrantha in Panama; Brokaw 1987). In some forests, only a fraction of this variation in life-history characters is represented. For instance, though there are substantial differences in life histories, all the tree species present in the stands discussed in this thesis could be classed in the small-gap specialist guild. For more detailed discussions of the role of gaps in maintaining species diversity, see reviews by Denslow (1980, 1987). Because gaps are loci of individual transitions of space from one individual tree to another, another focus of research has been to use patterns in these transitions to assess the successional status of forests (Barden 1979, 1980; Runkle 1981; Veblen 1985; White et al. 1985; Taylor and Zisheng 1988b). The constancy of the forest community in time is the net result of all the individual mortality and replacement events. Examination of such transitions gives a "snapshot" of the instantaneous rate and direction of change in the forest. Species-specific transitions can be used to build Markov models to project the longer-term consequences of current replacement patterns as well (above references; Horn 1971, 1975), given certain assumptions about the nature of replacement processes. Markov models have been used extensively to examine the dynamics of both forest (Horn 1971,1975; Barden 1979, 1980; Runkle 1981; Veblen 1985; White et al. 1985; Taylor and Zisheng 1988b) and heathland communities (Hobbs and Legg 1983;  6  Lippe et al. 1985). Though they have been criticized as unrealistic (van Hulst 1979a,b; Hobbs 1983; Lippe et al. 1985), Markov models of succession have been quite useful as benchmarks against which to judge field data (Usher 1981). They can be improved substantially with some simple modifications (Acevedo 1981; White et al. 1985).  1.3 Study Sites and Species. This research was conducted at Cypress Provincial Park in the North Shore Mountains of southwestern British Columbia (Figure 1.1; Latitude=49o25'N, Longitude=123°l2'W). The forests of Cypress Park are largely in the Mountain Hemlock Zone of British Columbia's Biogeoclimatic Classification System (Brook et al. 1970; Green et al. 1984; Pojar et al. 1987). The Mountain Hemlock Zone occurs between the lower elevation Coastal Western Hemlock Zone, and the higher Alpine Zone on the western side of the Coast Mountains of British Columbia (-900-1,700 m in the south, 300-900 m in the north). The stands I examined are at the lower range of the Mountain Hemlock Zone where it grades into montane stands of the Wetter Maritime Subzone of the Coastal Western Hemlock Zone. The Mountain Hemlock Zone can be divided into a Parkland Subzone at higher elevations, and a Forested Subzone at lower elevations. This research was conducted entirely in the Maritime Forested Subzone (Green et al. 1988). Plant associations and vegetation-environment relationships of the Mountain Hemlock Zone have been described extensively by Brooke et al. (1970). These forest classes correspond roughly to the upper range of the Abies amabilis Zone and the lower range of the Tsuga mertensiana Zone of the Cascade Range to the south (Franklin and Dyrness 1973), with which they have many similarities. The Mountain Hemlock Zone is characterized by a cool, wet climate, with significant precipitation every month of the year, and a substantial amount of the precipitation in the form of snow (Brooke et al. 1970). When rain or snow is not actually  7 falling, heavy fog is common. There is a strong maritime influence which is greatest nearest the coast and decreases inland as the climate becomes more continental in character. Late winter snow depths of > 3m are not uncommon, and even within the Forested Subzone, the snow-free period is usually less than 150 days (Brooke et al. 1970). The deep winter snowpack generally prevents soils from freezing. Small changes in slope position, slope angle, aspect, and elevation can have a pronounced influence on microclimate and vegetation. Despite its moderate coastal climate, vegetation zones are elevationally depressed at Cypress Park relative to other areas of similar latitude (i.e. Diamond Head in Garibaldi Park), because of exceptionally heavy snowfall and persistent spring snowpack (Brooke et al. 1970). A compacted or cemented glacial till with varying mixtures of rock types is the predominant parent materia! for soils. Podzolization and gleization are the predominant soil forming processes, and Humic and Humus Podzols are the zonal soils (Brooke et al. 1970; Ugolini 1982).  Species Autecology Dominant tree species at Cypress Provincial Park are Pacific silver fir (Abies amabilis (Dougl.) Forbes, Pinet.), western hemlock (Tsuga heterophylla (Raf.) Sarg.), mountain hemlock (Tsuga mertensiana (Bong.) Carr.), and Alaska yellow-cedar (Chamaecyparis nootkatensis (D. Don) Spach). No other species of trees are common in the forested areas, though western white pine (Pinus monticola Dougl.) occurs sporadically. These stands are above the elevational limit for Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco var. menziesii). All four dominant species are classed as shade tolerant (Fowells 1965; Krajina 1970; Franklin and Dyrness 1973; Minore 1979), and three are commonly considered "climax" species within particular elevational bounds. Western hemlock is the climatic climax of the adjacent lower elevation Coastal Western Hemlock Zone, and is exceeded regionally in its shade tolerance only by Pacific silver fir. Pacific silver fir is potentially  8 the climax species in an intermediate belt of montane to lower sub-alpine forests, and mountain hemlock is considered the climax within the Mountain Hemlock Zone proper. However, all three can also play pioneer roles in certain circumstances, emphasizing the context-dependent nature of the successional status of many western conifers (Franklin and Hemstrom 1981). Despite their broad similarity in shade tolerance, when compared with intolerant species, the four species do differ substantially in the life history characters and physiological parameters used to distinguish regeneration guilds. Though intolerant of drought on an absolute scale (for instance, in comparison to its lower elevation associate Douglas-fir (Pseudotsuga menziesii). western hemlock is substantially more tolerant of moisture stress than Pacific silver fir, and can sustain substantially greater growth rates in well lit environments (Kotar 1972; Long 1976; Grant 1980; Lassoie et al. 1986). Western hemlock is also more tolerant of drought stress than mountain hemlock or yellow-cedar (Lassoie et al. 1986). Both Pacific silver fir and western hemlock can persist as suppressed saplings and experience releases in growth when the canopy is opened above them (Herring and Etheridge 1976; Oliver 1976; Packee et al. 1982; Tucker et al. 1987). However, while Western hemlock acts as a "climax" type species at lower elevations, at higher elevations it role becomes increasingly that of a pioneer. Its upper elevational limit is set by a complex of factors related to the duration and characteristics of snowpack (see extensive discussions in Thornburgh 1969; Kotar 1972; Long 1976). Mountain hemlock and yellow-cedar are high elevation species, both with high moisture requirements, and great tolerance of short growing seasons and heavy snow loads (Fowells 1965; Krajina 1970). Yellow-cedar has been considered almost a ruderal species; tolerant of a variety of stresses, common in open areas, and rarely dominant in closed-canopy forest (Antos and Zobel 1986). However, because of its longevity (see below) its status within stands can be difficult to ascertain (Franklin et al. 1988).  9 Both Hemlock species produce large numbers of small, well dispersed seeds with high viability, and produce some seed nearly every year (reproductive and lifehistory information for all species from Fowells 1965; USDA 1974; Arno and Hammerly 1977; Franklin and Waring 1980; Waring and Franklin 1979). Yellow-cedar seeds are small, and well dispersed, but of low viability, and though some may be produced regularly, good seed years are infrequent. Pacific silver fir produces small numbers of large, heavy, poorly dispersed seed, and does so infrequently. The four species differ substantially in maximum longevity (in order of increasing longevity, firs < western hemlock < mountain hemlock < yellow-cedar). The firs can reach 500-600 years old, and both hemlocks can reach 800 years. Yellow-cedars commonly reach 1000 years, and much older ages have been reported.  Research Stands The locations of the four stands are indicated in Figure 1.1. Three stands are located along the Howe Sound Crest trail on the western slopes of Mt. Strachan. Two, HSC1 and HSC2 (Howe Sound Crest 1 and 2), are adjacent to each other. A third, NST1 (North Strachan 1), is several hundred metres farther north on the trail. HSC1 is above the trail, and HSC2 below the trail in the Howe Sound Overlook area. NST1 is below the trail. Stand HLY1 (Hollyburn 1) is on the western slope of Hollyburn Mountain, above the trail between the cross-country and downhill skiing facilities (about 2 km from the other stands). The stands are all near 1,000 m elevation (950-1145m), and vary from SSW to NW in aspect (Table 1.1). HSC1, NST1, and HLY1 are on slopes (20-300), and HSC2 is on a nearly level bench. In this area, cold air pools on flat areas, creating a microclimate more similar to higher elevation forests where the Forested Subzone begins to grade into the Parkland Subzone. This can cause substantial small-scale heterogeneity in vegetation. HSC2 is similar in structure and composition to stands on  10 benches above HSC1, NST1, and HLY1. Details of the structure and composition of each stand will be presented in Chapter 2. Soils in these stands are humic podzols with 40-60 cm deep organic layers. The humic soil layers remain wet even at the end of the summer dry season, so that pollen grains within the soil column are preserved. Lertzman and Brubaker (in prep.) studied the long-term history of Cypress Park by examination of the pollen and charcoal fragments in the soils within stands HSC1 and HSC2. Analysis of a 40 cm profile from HSC2 has been completed. Charcoal at the base of the profile was radiocarbon dated at 4,700±500 yr BP. The lower half of the profile contained two additional charcoal peaks, indicating fires, but there is no charcoal in the upper half of the column. Though species fluctuate in abundance substantially, the species present in the stand today were present throughout the last 4,700±500 yr, and there is no indication of other species that have been lost from the community. It is tenuous to assume constancy in sediment deposition rate, but if we do, this suggests that it has been conservatively 1,500-2,000 years since the last major fire. This conclusion is supported by the presence of large canopy trees (probably > 600 years in age) on the uncharred stumps of other equally large trees. In general, the analysis of Lertzman and Brubaker indicates that the study stands are a system where we can expect gap-phase processes to have played an important role in shaping the patterns in structure and composition apparent today. Further, we are looking at a community that has persisted through substantial changes in climate largely by internal accommodations, with no losses or additions of canopy species.  1.4 Synopsis of Chapters. Chapter 2 presents the basic description of the structure and composition of the Cypress Park forest. It focuses on the species composition of the four stands (canopy and understory layers), on how much of the forest can be classed as gap, and on gap  11  size, geometry and modes of formation. Based on these data,'. I estimate the time taken for gaps to "fill" and the consequent rate of turnover of space in the forest. Chapter 3 examines the two common methods for calculating forest turnover time by comparing data from a variety of forests (from the literature), and from simulation results. Chapter 4 examines the populations of saplings in gaps (gapfillers) in relation to the species of tree that died to create the gap (gapmaker) and the gap environment. These patterns are used to assess whether gap-phase processes act to promote coexistence in this forest. Chapter 5 uses the matrix of gapmaker-gapfiller transitions generated in Chapter 4 as the basis for Markov models of forest change. These models are used to assess the equilibrium status of community composition, and the potential roles of differential longevity and climatic fluctuations in modifying community dynamics. Chapter 6 is a brief discussion of the main results and conclusions.  Table 1.1 Characteristics of the four focal stands.  Stand  Elevation  Aspect  Slope  Soil Moisture  (m)  (degrees)  (degrees)  Regime  HSC1  1085-1145  221-260  21-31  moist  HSC2  1040-1050  180-240  0-17  fresh  NST1  1000-1030  225-315  21-30  moist  HLY1  950-1000  242-282  18-24  moist  13 Figure 1.1 Location of Cypress Provincial Park and the focal study stands.  CYPRESS PROVINCIAL PARK  Vancouver  Hiking Trails Chalrllft R o a d and Parking Lot Summit  14 2. GAP-PHASE STRUCTURE OF A SUB-ALPINE OLD GROWTH FOREST  2.1  Introduction. In forests where large, stand-destroying disturbances are infrequent, small-scale  disturbances associated with the mortality and replacement of individual trees are a primary source of heterogeneity in forest structure and composition. When a canopy tree dies in such forests, the removal of its crown from the canopy creates a gap (Jones 1945; Watt 1947; Bray 1956) in which species are able to establish or recruit that could not under a closed canopy (Runkle 1981,1982,1985; Brokaw 1985a,b; Denslow 1980, 1985; Whitmore 1989). In this way, gaps may contribute to the maintenance of species diversity in a variety of communities (Grubb 1977; Connell 1978; Denslow 1980, 1985; 1987). Much of the response to gaps is associated with changes in the light, temperature and water regimes on the forest floor (Bazazz and Pickett 1980; Chazdon and Fletcher 1984; Collins et al. 1985; Denslow 1987), but the belowground status of gaps is poorly understood (Vitousek and Denslow 1986; Mladenhoff 1987). Fallen trees create a variety of substrates not otherwise available (Orians 1982; Bazazz 1983), and windthrow events can exert a long term and substantial influence on both soil structure and chemistry (Lutz 1940; Stephens 1956; Beattie and Stone 1986). Gap-phase processes have been described most extensively for the mixed hardwood forests of eastern North America (Runkle 1981, 1982, 1985; Romme and Martin 1982) and the moist to wet neo-tropics (Hartshorn 1978; Brokaw 1985a,b; Hubbell and Foster 1986a,b). The old-world tropics have also received attention (Aubreville 1938; Poore 1968), as have some southern hemisphere temperate forests (Veblen 1985, 1989; Cullen 1987; Armesto and Fuentes 1988), and Asian coniferous and mixed forests (e.g. Naka 1982; Nakashizuka 1983; Taylor and Zisheng 1988a,b). Gap forming disturbances have received considerably less attention in the coniferous forests of western North America. This is probably due to the clear and often overwhelming impact of the catastrophic disturbances, mainly fires, to which many  15 stands owe their origin (Franklin and Hemstrom 1981; Hemstrom and Franklin 1982; Spies and Franklin 1989a). However, even in a matrix of younger, even-aged stands, there are frequently patches of many-aged old growth persisting on north facing slopes, in protected hollows, or wet pockets, that predate the surrounding forest (Eis 1962; Hemstrom and Franklin 1982). In these stands, and other areas which have escaped catastrophic disturbance for long periods, we might expect processes occurring at the spatial scale of one to a few trees to play an important role in forest structure and composition. In mid-elevation old growth Douglas-fir (Pseudotsuga menzesii) forests in the western Cascades of Oregon, canopy gaps are required for the replacement of the early serai Douglas-firs by western hemlock (Tsuga heterophylla) and Pacific silver fir (Abies amabilis) (Stewart 1986a,b; Spies and Franklin 1989ab). Differences in canopy structure between species, stand age, and disturbance history all affect both the development of advanced regeneration under intact canopies, and the response of vegetation to the creation of a gap. The presence of Douglas-fir, with its large dimensions and long tenure, exerts a profound influence on the vertical and horizontal structure of these stands, and on the time dynamics of succession.(Stewart 1986a,b; Spies and Franklin 1989b). At higher elevations in the Cascades, above the range of Douglas-fir, the autecology of the dominant species (true firs, hemlocks and cedars), combined with a cool, wet climate and infrequent fires suggests the potential importance of gap-phase processes. Though patterns of succession and stand development have been well described (Thornburgh 1969; Kotar 1972; Long 1976), gap-phase processes have not yet been examined explicitly. Where regeneration in canopy openings has been observed in higher elevation forests (Kotar 1972; Long 1976), the abundance and growth rates of Pacific silver fir and western hemlock were greater in the openings than under a closed canopy, and gaps in the canopy were considered necessary for recruitment.  16 The montane to sub-alpine forests of coastal British Columbia are similar floristically and climatically to those of the western Cascades (Krajina 1970; Brooke et al. 1970; Franklin and Dyrness 1973; Franklin et al. 1988). However, the importance of gap-phase processes has not been examined there. In this chapter I will describe stand composition and gap-phase structure of an old growth, sub-alpine forest just north of Vancouver, British Columbia. I will focus on the overall structure of the forest, and on the patterns of mortality that give rise to gaps, gap size, and geometry. Subsequent chapters will examine the patterns of tree replacement in these gaps, and the longer-term consequences of replacement patterns for forest dynamics.  2.2 Methods.  Sampling for canopy structure The methods I used to estimate the proportion of the forest represented by gaps, and to locate a population of gaps for more detailed description, are similar to those of Runkle (1982) and Veblen (1985). This sampling was done in the summers of 1985, 1986 and 1987. The sampling in 1987 followed very closely the protocol described in Runkle (1987, MS.). In each of the 4 stands (HSC1, HSC2, NST1, HLY1), I chose a starting point at the center of a gap. From this point, a 50 or 100 m transect (depending on topographic or other discontinuities in the stand) following a constant elevational contour was laid down with meter tape. The end was marked permanently with painted PVC pipe. Additional parallel transects were established at 30 m intervals (ground distance) either upslope or downslope from the first. I found 30 m to be enough distance between transects that all but the biggest gaps were not contacted by more than one transect, but few, if any, small gaps were missed between transects. If streambeds, rocky outcrops or other major exogenous discontinuities in forest structure were encountered, the transect was terminated or continued on the other side of them. These accounted  17 for a small proportion of the land surface area. A total of 350 m of transect were sampled in HSC1, and 300 m each in HSC2, NST1 and HLY1, for a total of 1,250m. At 1 meter intervals along each transect, I classified canopy status as canopy gap, expanded gap, or closed canopy. A canopy gap is the vertical projection onto the ground of the opening in the forest canopy caused by the mortality of a tree (Figure 2.1); for an opening to be considered a gap, the remains of a gapmaker (see below) must be present. I considered a point on the transect to be in canopy gap if, in sighting vertically above it (using a clinometer to insure a vertical sighting), no foliage of canopy trees was encountered. For this and all subsequent discussion, a canopy tree was defined as 22 cm dbh or greater and at least 10 m in height. The boundary of the expanded gap (Runkle 1982; Veblen 1985) is defined by the boles of the trees whose canopies define the canopy gap. A point was considered in expanded gap if it was underneath the foliage of these trees (Figure 2.1). With increasing latitude, and especially on slopes, an increasing amount of direct beam radiation enters a gap at angles other than 90° relative to the forest canopy. This results in light from a gap passing beneath the canopies of the trees bordering the canopy gap, suggesting that the expanded gap may be a more appropriate measure of the effective opening than the canopy gap (Canham1988b). All other points were under more or less continuous canopy and were considered closed canopy. There were occasionally small openings in the canopy where the crowns of adjacent trees did not quite meet, or where a very old gap had not quite filled. If these were, by visual estimate, less than half the diameter of the average canopy tree crown in that area, then they were still considered closed canopy. Although interstitial spaces between adjacent crowns were frequent, I never had any difficulty distiguishing these from canopy gaps as defined above. Other than those around stream courses or rocky outcrops (see above), there were no openings exceeding this size criterion that were not associated with canopy gaps.  18 I calculated the proportion of surface area in the forest represented by each canopy category as the proportion of the total meters of transect sampled that were classified as that category. At each meter along the transect I also recorded substrate, species occupying the understory layer, and species in the overstory. The species present in the understory at each sampling point, combined with the canopy category, was used to assess the species diversity of canopy gap, expanded gap, and closed canopy locations. At 25m intervals on the transects in HSC1, HSC2 and HLY1, I used the point quarter method to collect data for the estimation of size distributions and density of canopy trees. Canopy tree diameters were often too large, and the patterns of annual growth rings too asymmetric and complex, to obtain satisfactory age estimates through the use of tree cores. To estimate ages of canopy trees, I cut disks (whole diameters or wedge shaped radii) from all stumps that were not too decayed in clear cuts near the edges of HSC1 and HSC2. These were sanded until ring patterns were clear, and annual growth rings counted under a dissecting microscope. Many showed asymmetric growth, with multiple bands of reaction wood, many partially missing rings, and rings that were hard to resolve even under high magnification. I selected radii to maximize the number of rings in each count. Where possible, more than one radius was counted and the oldest age estimate used. These age estimates are minimum estimates, probably accurate to within 30 years. At several points along each transect, I selected codominant trees that appeared to represent the typical canopy height for that area, and obtained height estimates with a clinometer and trigonometry.  19  Sampling For Saplings Kohyama (1983) differentiates saplings from seedlings by the presence of a "fully expanded lateral branch system", representing decreased dependence of the tree on the current year's assimilate. For Abies mariesii and A. veitchii in Japan, he estimated the transition from seedling to sapling to occur at 20 cm in height. I defined the lower size limit for saplings to be 30 cm and the upper limit to be 6 m. I wanted to focus on "well established" members of the sapling bank that were candidates for future gapfillers. To describe the population of saplings, I surveyed ten 50m transects; six in HSC1 and 4 in HSC2. These were in the same area but not along identical lines as the canopy census transects described above. All were separated by 30m. At 10m intervals on each transect the distances to the 4 nearest trees between 0.3m and 6m tall were measured for point-quarter estimates of density. For each tree I also recorded height, rooting substrate, canopy status, and overstory species. Suppressed saplings that had died occurred sparsely on the forest floor. I collected basal disks from any such individuals encountered near sampling points on the sapling transects. These were prepared and aged in the same way as the samples of canopy trees. For each sample, at least 2 radii were counted, and the oldest estimate used. These estimates are probably substantially more accurate than those for the older trees.  Description of Individual Gaps All gaps with canopy gap or expanded gap intersecting the transects were enumerated and marked for further study. Line transects do not sample gaps of different sizes without bias - larger gaps are more likely to be sampled relative to their abundance (Runkle 1982). The inclusion of gaps whose expanded gap intersected the transects, rather than only those with canopy gaps intersecting the transects, partially corrects for this bias (Runkle 1987, MS.). To further correct for this, I enumerated any  20 very small gaps that I encountered which did not intersect any transect and would otherwise have been missed. This sampling scheme achieved at least close to a complete census within the sampled area. A gapmaker \s a tree whose mortality contributed to the creation of a gap. A gapfill&r is a tree growing in a gap. A definitive gapfiller\s a gapfiller which, by virtue of its size, location, and growth rate, was clearly dominant to the other gapfillers and likely to recruit into the canopy in that gap. Some gaps had no definitive gapfillers. All gaps were described in detail. I recorded aspect, slope, orientation of the long axis of the gap, canopy height, mode of mortality for each gapmaker (windthrown, standing dead, etc.) and, where possible, species and dbh of the gapmaker, species of the trees defining the gap boundaries, the species and sizes of gapfillers (and where possible, which gapmaker they were associated with), and expanded gap size. For all gaps in HSC1 and NST1, and most in HSC2, I also measured canopy gap sizes, gap aperture (after Lawton and Putz 1988), the number of years since the last release in height growth on gapfillers showing a distinct release pattern, diameter at breast height for the surrounding canopy trees, and whether they showed indication of having germinated on logs or stumps (e.g. the presence of stilt roots). The proportions of each species and the diameter distributions for these canopy trees were combined with the point-quarter data for the overall estimates of species composition and structure of the canopy (Table 2.3a). I also sampled vegetation and substrate along line transects on the major and minor axes of each gap. I classified the "age" (time since mortality) of gapmakers using a decay class system based on morphological characters observable in the field. This system differs from that used by Triska and Cromack (1979), Sollins (1982), Graham and Cromack (1982), and Means et al. (1985) because of the predominance of snags rather than fallen logs, and the need to include stumps that persist after the logs associated with them are not detectable. Table 2.1 shows the criteria used in classifying gapmaker age.  21 Young gapmakers are recently dead trees with all or most of their canopy branch structure intact and little or no evidence of decomposition in the bole or stump. Medium gapmakers have lost smaller branches, but may retain larger ones, have sloughing bark and show substantial signs of decay in heartwood (for standing dead) or sapwood (for logs on the ground), but are able to support their own weight. I classified old and very old gapmakers solely on characters of the stumps; logs from old gapmakers are mostly incorporated into the forest floor and logs from very old gapmakers are indistinguishable. Thus, for these latter two categories I could rarely assign a mortality class; they appear as snapped off stumps irrespective of the original mode of mortality. Old gapmakers had some sound wood protruding above the surface of the forest floor. This sound wood varied from a substantially intact stump representing the full basal diameter of the tree, to a narrow central core of sound heartwood. Very old gapmakers are visible as a mound in the forest floor mostly or entirely covered by leaf litter and moss, and usually supporting a substantial growth of conifers and shrubs. Young and medium gapmakers correspond roughly to Triska and Cromack (1979) and Sollins' (1982) classes 1-3. Old and very old gapmakers correspond to Triska and Cromack (1979) and Sollins' (1982) classes 4 and 5, and a stage beyond their class 5 where stumps are present but logs are no longer detectable. Gapmakers intermediate in character were classified as young-medium or medium-old. During analysis, I found those I had classified in the field as old-very old had the same distribution of species and modes of mortality as very old gapmakers (mostly unidentifiable in both cases) and therefore I combined the two classes. To measure canopy and expanded gap size, I laid perpendicular 50m tapes along the two main axes of a gap. These two tapes intersected as close to the visual center of the gap as possible. For each of these four radii, and four additional radii at angles equidistant between them, I measured the distances from the gap centre to the edges of the canopy gap and the extended gap (a line drawn between the centers of the boles of trees whose canopies defined the canopy gap edge). In structurally  22 complex gaps, I used additional measurements as needed to represent a gap accurately. I drew scale maps of each gap and obtained gap area and perimeter with a digitizer. For gap size, the expanded gap includes the canopy gap, whereas in the previous estimation of how much space in the forest is represented by each category, canopy gap and expanded gap are mutually exclusive categories. Gap aperture may be a better proxy for estimating the light available in a gap than gap size (Lawton and Putz 1988). Gap aperture is the average of the four angles from the zenith to the edge of the canopy, along the two main axes of a gap. I measured gap aperture at eye level in the center of each gap.  Stand vs. forest level analyses. I will present the basic descriptions of forest composition (canopy and understory layers) and gap-phase structure (percent canopy gap, expanded gap, and closed canopy) on both a stand-by-stand basis and averaged over the four study stands within Cypress Provincial Park. All discussion of the characteristics of gapmakers and gaps as populations (e.g. gapmaker species and mortality patterns, gap size and geometry) will be presented as one data set combining data from the four stands.  2.3 Results.  Percent of Forest in Each Canopy and Substrate Category. Of 1250m sampled on 14 transects in the 4 stands, 18% was in canopy gap, 52% in expanded gap and 29% in closed canopy (Table 2.2). Depending on whether expanded gap is considered a gap or closed canopy environment, this results in a total of 73 or 18 percent gap, respectively. The four stands did not differ significantly in the amount of each canopy class (Table 2.2). In fact, the mean amounts of closed canopy and total gap (canopy + expanded gap) were remarkably similar between stands. Though its total gap was similar to the others, more of NST1 was in expanded gap than  23 the other stands. This is probably because a greater proportion of the gaps in this stand were older than in the other stands, and, as gaps age, canopy gap is converted into expanded gap. Leaf litter was the predominant substrate in the forest, accounting for 67% of the sampled points. Logs represented 8.6% of the sampled surface, root mounds 3.5%, and stumps 3%. Miscellaneous dead woody debris (e.g. branches, highly decomposed old logs incorporated into the forest floor) made up an additional 17%, and 1% was boles, rocks, roots and bark. The relative paucity of large woody substrates (-15%) is significant because of their importance as a rooting substrate for western hemlock (see below).  Composition and Structure of the Canopy Layer When the data from all stands are combined, Pacific silver fir is the most abundant tree in the forest canopy, though it is most common among the smaller canopy size classes (Table 2.3 and Figure 2.2). Western hemlock is the next most abundant species overall; together these two dominate the forest with 79% of the individuals. Western hemlock occurs more frequently than the other species as very large individuals (Figure 2.2) and thus dominates the forest in basal area (i.e. relative dominance, Table 2.3). Mountain hemlock and yellow-cedar are less common in the forest as a whole, and tend to be smaller in diameter than western hemlock but larger than Pacific silver fir (Figure 2.2, Table 2.3). Yellow-cedar has the most skewed size distribution with a few scattered very large trees and a larger number of smaller ones. There is substantial variation between the stands in overall density of canopy trees and in species composition. Of the three stands for which I obtained estimates of densities, HSC1 was the least dense and HSC2 the most dense, though they are the closest in space (Table 2.3). Stand HSC2 was also the most distinct compositionally. It was the only one where mountain hemlock was more abundant than western hemlock and it had a nearly equitable mix of species. HSC1 and NST1 had the least equitable  24 mix of species, being largely dominated by Pacific silver fir and western hemlock, with very little yellow-cedar. HSC2 also differs from the other stands in the overall size distribution of canopy trees. Mean diameter is significantly smaller in HSC2 than for the other three stands, which, despite their differences in density and species composition, do not differ (Table 2.4) HSC2 is more similar in character to the forests at higher elevations than it is to the adjacent stands HSC1 and NST1. It has a higher proportion of the high-elevation tree species and a higher density of smaller trees in the canopy layer. Snowmelt in the spring is delayed up to 2 weeks relative to the other stands and higher elevation subalpine shrubs (e.g. Menziesia ferruaina and Rhododendron albiflorum) are much more abundant. These features are characteristic of flat benches in this area, which tend to grade from the Forested into the Parkland Subzone of the Mountain Hemlock Zone. The tallest trees in the canopy were western hemlocks, which averaged slightly taller than Pacific silver firs (45.8m vs. 41.5m). Mean canopy height for the four stands was 44.5m (N=22) and the tallest tree observed was a 61 m western hemlock in NST1. Within an area, the height of canopy dominants tended to be similar. Ages of individuals of Pacific silver fir and both hemlock species are among the highest reported for these species (Figure 2.3.; Fowells 1965; Waring and Franklin 1979). The data in Figure 2.3 are insufficient to characterize fully the distribution of tree ages, but illustrate several important points. First, the forest is multi-aged; canopy tree ages are spread, more or less continuously, from 200-1,000 years. Second, the hemlocks tend to be older than the firs, reflecting either their greater longevity (Fowells 1965; Waring and Franklin 1979) or greater recruitment of the firs over the last 400 years. Third, the absence of firs less than 220 years of age and of hemlocks less than 350 suggests that approximately 200-400 years is required on average to grow from stump height to canopy size.  25 Inspection of individual growth curves for representative canopy trees supports this estimate of recruitment time. Figure 2.4 shows decadal increments of basal area for six representative canopy trees from the clear cut adjacent to stand HSC2 (basal area calculated from diameter increment). Diameters of these trees varied from 62 cm to 1.05 m. Most show a substantial period of suppression during the first part of their lives, some for as much as 400 years. Two other patterns are superimposed on this: short peaks of a few years to a few decades duration, and a general increase in growth during the last 100 years. In general, these growth curves support the idea that the current canopy was established through a succession of gap forming and filling events similar to what will be described for the modern forest. This conclusion receives further support from the number of canopy trees showing evidence of having germinated on the (uncharred) stumps of older dead trees: 65% of all canopy trees (56% of Pacific silver firs, 88% of western hemlocks, 58% of mountain hemlocks, and 19% of yellow-cedars). The age estimates in general, and the length of the periods of suppression in particular, are likely underestimates because it may take 50-100 years to reach stump height (see below). Estimating ages for yellow-cedars was problematic. I found only a few yellowcedar stumps in the clear cut. They were mostly in a single cluster, represented only the smaller size classes, and were all within 50 years of 375 years in age. Because of the tendency of trees in the larger size classes to have heart rot, it was impossible to obtain age estimates for large living cedars. However, one yellow-cedar near the upper end of the size distribution for these stands and ca. 1 km from stand HLY1 has been aged at 1,100 years at 2-3m off the ground (L. Josza, Forintek Canada Corp., pers. comm.). 1,100 + 300 is probably a reasonable estimate of the age range for large diameter yellow-cedars within most of Cypress Park.  26  Composition, Structure and Distribution of the Sapling Layer. With 81 % of the individuals overall, Pacific silver fir dominated the sapling layer to a much greater extent than it did the canopy (Table 2.5). The other species decreased proportionately, especially mountain hemlock and yellow-cedar, which were represented by only a few individual saplings on the transects, even in HSC2. HSC1 and HSC2 differed primarily in the decreased abundance of western hemlock in HSC2, and the greater overall density of saplings in HSC2. There were differences among the three canopy status classes in sapling growth rates and density, but not in the distribution of sapling heights (Table 2.6). Mean sapling density of all species was greatest in expanded gaps, less in closed canopy, and least in canopy gaps. Height growth rates (measured for Pacific silver fir only) were indistinguishable between canopy and expanded gap, and mean growth rate in each of these cases was greater than in the closed canopy (Table 2.6). The distribution of sapling heights did not differ between the closed canopy areas and the two gap categories, or between saplings in canopy and expanded gap. The distributions of both heights and growth rates were skewed and highly leptokurtic, with many trees growing very little and a few growing at much greater rates. This was true even for the populations in gaps. Thus, gaps may provide increased opportunity for establishment and persistence for many seedlings and saplings, but apparently, even in gaps, the chance to achieve high growth rates is limited to a few trees. Many saplings in all locations reflect a long and variable growth history: 40% of all Pacific silver fir saplings showed evidence of having experienced more than one growth regime (suppression-release and multiple suppression-release; Table 2.7). Of the Pacific silver firs showing an open growth form in Table 2.7, 44% were in canopy gap and 50% in expanded gap. The high proportion of open grown trees in canopy gap is especially significant considering that there was nearly three times as much area in expanded gap as there was in canopy gap.  27 Saplings of all tree species occurred in greater numbers in gaps than under a closed canopy. However, they were in close to the same proportions as each canopy class occurred in the forest (Table 2.8). Pacific silver fir and western hemlock are distributed among canopy classes in proportions that do not differ significantly from the relative abundances of each canopy class in the forest (goodness of fit X2, p=.25 for both species). Mountain hemlock and yellow-cedar do not appear in sufficient numbers to test Table 2.8 as a whole for interaction between species and canopy class, but when these species are excluded, the hypothesis of no interaction between species and canopy class for western hemlock and Pacific silver fir cannot be rejected (likelihood ratio X2 for the contingency table, p=.07). There is a strong interaction between sapling species and substrate, and, within species, individuals are not distributed randomly among substrates (Table 2.9). Because of small sample sizes for mountain hemlock and yellow-cedar, Table 2.9 as a whole cannot be tested for species-substrate interaction, but the two-species table with Pacific silver fir and western hemlock shows a significant interaction (likelihood ratio X2, p<.001). Testing each of the two species individually against the distribution expected from the relative abundance of each substrate shows significant deviations (likelihood ratio X2, p<.001 for Pacific silver fir and western hemlock). Both species occur much more frequently than expected on stumps, and to a lesser extent on other woody substrates. Western hemlock occurs much less frequently than expected on the forest floor. How old are members of the suppressed sapling population? I used basal discs from dead, suppressed understory trees to estimate ages. This estimates age at death, not age in general, and presumably represents the slower growing part of the population. The mean age and height of dead suppressed Pacific silver firs (±1 SD) was 146±63 years and 1.7+1 .Om respectively. This distribution of heights has a slightly higher mean than the overall sapling population (t-test; p=0.044) and approaches normality more closely. The regression of height on age (Figure 2.5) gives a mean  28 growth rate of 0.9 cm/year, which is well within one standard deviation of the mean for the general understory population (Table 2.6). The regression equation predicts an extraordinary average height of 2m at 200 years old. Even given the bias, these data indicate that the population of saplings in the understory represents the net effect of up to several hundred years of demographic processes. Patterns of release dates among saplings less than 5 m in height were not useful as indicators of the dates of gap creation. They were highly variable within a gap and broadly consistent among gaps. Averaged over all gaps, there was a mean number of years of released height growth of 10.4 years (SD=3.4, N=250). Most gapfillers that showed released sapling height growth were released within a few years of 10 years ago, even in very old gaps. This may be partially due to the multiple gapmaker origin of most gaps (if the releases date the most recent gapmaker to contribute to the gap), but it was often difficult to establish a clear correspondence between particular released gapfillers and particular recent gapmakers. In summary, seedlings of most species may establish, and saplings persist, under any canopy class, though western hemlock may be limited by substrate in all of them. Pacific silver fir establishes and persists in greater numbers than the other species under all canopy classes and on all substrates. Though fir saplings can persist under a closed canopy, growth rates are substantially reduced relative to both types of gap. The majority of saplings in the forest in general, and even in gaps, experience suppression at some point during their life.  Diversity of the understory layer. Though the differences were sometimes slight, gaps (expanded and canopy) had consistently higher vascular plant diversity than closed canopy areas (Figure 2.6). Neither canopy nor expanded gap was consistently more diverse than the other. Stands varied in diversity of the understory layer, with HLY1 the most diverse, and NST1 the least. HSC2 was the only stand where mosses were a substantial  29 component of the understory. Adding them to the species diversity calculation increases the values for HSC2 to a level comparable to that of HLY1. Western hemlock is known to produce especially dense shade beneath its canopy (Stewart 1986a,b; Spies and Franklin 1989b), so I regressed understory species diversity on the percentage western hemlock in the canopy (arcsin square root transformed) for each stand to examine the extent to which this might explain the variation among stands. For stand diversity as a whole (data from canopy gaps, expanded gaps, and closed canopy combined), the r2 was 0.99 and the regression highly significant (Figure 2.7). For diversity in each canopy class taken individually, r2's were all quite high (>0.80), but only the regression for canopy gap diversity on percent western hemlock was significant (r2=0.965; p=0.018). Though there are only four points being fit, there is wide variation in the independent variable, and the r2's are quite striking.  Number of Gapmakers Per Gap and Mode of Mortality. I separated gapmakers into those of primary and secondary importance in each gap. A primary gapmaker was one whose crown was judged to have represented a substantial portion of the current canopy gap. Secondary gapmakers are either on the edge of the expanded gap (thus the removal of their crown did not greatly influence the current gap), are smaller trees knocked down by primary gapmakers, or represent earlier mortality events whose influence on gap geometry has been overwhelmed by the more recent mortality of a primary gapmaker in the same part of the gap. Most gaps are a result of the mortality of more than one tree: 90% of the gaps had more than one gapmaker, and 63% had more than one primary gapmaker. Many gaps had several gapmakers, to a maximum of 16 (8 primary and 8 secondary; Figure 2.8). The mean numbers (± one standard deviation) of primary, secondary, and total gapmakers per gap were 2.8±1.9, 3.0±2.5, and 5.7+3.8, respectively. The gapmakers in a given gap were often of different decay classes (primary gapmakers usually more  30 recent, secondary gapmakers usually older), suggesting a long residence time for many gaps. Primary gapmakers had a larger mean diameter than secondary gapmakers (0.92 vs. 0.68 m; t-test, p<0.001). I divided the mode of mortality of gapmakers into standing mortality, windthrow (any gapmaker with roots pulled from the soil and its bole intact), snap-off (roots remaining in the soil and bole broken), and unknown (Table 2.10). Of the 143 gapmakers for which the mode of mortality could be identified (95 primary gapmakers total), 55% died standing (64% of primary gapmakers). Only 13% of the total gapmakers were windthrown (13% of primary gapmakers). The remaining 31% (23% of primary gapmakers) had snapped between two and ten metres above the ground. Some of the gapmakers classified as snap-off mortality probably died standing intact and snapped later. Many of the snapped and windthrown trees were small individuals knocked down by larger trees when they fell (note the lower proportion of primary gapmakers in these mortality types). "Windthrow" in this classification does not mean that wind was necessarily the direct causative agent. Mean diameters did not differ significantly between gapmakers experiencing different modes of mortality (ANOVA; p=0.09), but the mean diameters of windthrown or snapped trees were smaller than those that had died standing (means = 0.71, 0.76, and 0.92m, respectively). Though a large proportion of the total gapmakers could not have a mode of mortality assigned to them (57% of all gapmakers; 41% of primary gapmakers; Table 2.10), 98% of the young, young-medium and medium age-class gapmakers could be assigned to mortality classes (gapmaker age classification scheme, Table 2.1; data, Table 2.11). The proportion of standing dead among these trees increases to 62% (71% of primary gapmakers) due to a reduction in the proportion of snapped gapmakers. This suggests that an even higher proportion of all gapmakers died standing than Table 2.10 indicates, and that many of the primary gapmakers that now appear as snap-off mortality died standing.  31 Windthrown gapmakers were all complete, plate and ball "hinge type" tree falls (sensu Beatty and Stone 1986). Almost all the windthrown gapmakers were Pacific silver firs (17 out of 19; Table 2.8), though these represented only 20% of all Pacific silver fir gapmakers, and 18% of Pacific silver fir primary gapmakers. Of the 160 gapmakers identified at least to genus (101 primary gapmakers), 64% were Pacific silver firs and 23% western hemlocks (56 and 27% of primary gapmakers; Table 2.10). Pacific silver fir thus occurs as a gapmaker in a much higher proportion than its representation in the canopy (Table 2.3a). The mean diameter of western hemlock gapmakers was greater than for Pacific silver firs (1.1 vs. .70 m, t-test; p<0.001), consistent with the higher proportion of western hemlock gapmakers that were primary (73% of the western hemlocks vs. 56% of the Pacific silver firs). Does the species composition of gapmakers change with gapmaker age class? Figure 2.9 shows the ratio of Pacific silver firs to western hemlocks and to all hemlocks for young through old age classes. There are roughly double the number of firs per hemlock for medium aged gapmakers than there are for both younger and old gapmakers. Two alternative hypotheses could explain this pattern: 1) a pulse of mortality among Pacific silver firs in the past, or 2) a general increase in the representation of hemlocks among gapmakers (leading to a lower fir:hemlock ratio among younger gapmaker age classes), combined with a more rapid rate of decay for Pacific silver firs (leading to a lower ratio among older age classes). There are no data with which to evaluate the first hypothesis, but the second is consistent with a gradual increase in firs in the canopy (see Chapters 4 and 5). Since the same pattern holds for the comparison of firs to both mountain and western hemlock, the higher proportion of firs among medium gapmakers is less likely to be a result of a directional change in climate. With increasing time since mortality of the gapmaker, an increasing proportion of the individuals were stumps that could not be assigned to a mortality class (Table 2.11) These older age classes of gapmakers were only common as primary gapmakers in  32 HSC2. Gap filling in HSC2 may be delayed by persistent spring snowpack because snow persists longer in the spring in this stand than in the others (see discussion).  Gap Size and Geometry. Canopy gap sizes varied from 5 to 525 m2, with a median of 41 m2, and exhibited a distribution that was negative exponential in form (Figure 2.10, Table 2.12). Expanded gaps varied from 25 to 1127 m2 in area, with a median of 203 m2, and more closely approached a lognormal distribution, with an intermediate modal size. Gap apertures were even more symmetrically distributed, with a median and mean of 130. Gap aperture was the least variable of the three measures (coefficient of variation=0.48), and canopy gap area the most variable (CV=1.4 for canopy gaps, 0.85 for expanded gaps). Though most gaps are small, large gaps contribute disproportionately to the total gap area in the forest. The five largest gaps (14% of the total number of gaps) contained 52% of the canopy gap area, and the 10 largest gaps (16% of the gaps) contained 44% of the expanded gap area. In some forests, gap shape is well approximated by an ellipse (Runkle 1982; Veblen 1985), but at Cypress, both canopy gaps and expanded gaps are often quite irregular in shape and, especially for larger gaps, the relationship between perimeter and area deviates substantially from that predicted for a circle (Figure 2.11). With increasing size, gaps become increasingly irregular. This is consistent with the multiple gapmaker origin of most gaps: while the crown of a single tree may be roughly represented by an ellipse, the intersecting crowns of several are unlikely to be. Lawton and Putz (1988) suggested that gap aperture is much better correlated with photosynthetically active radiation (PAR) in the center of a gap than is gap area per se. Gap aperture integrates projected gap area with stand stature and slope. Incident light in the gap will be a complex function of how these 3 interact with canopy geometry and latitude. There was a strong relationship between gap aperture and gap area and perimeter (0.77<r2<0.82 for all comparisons between canopy and expanded gap area  33 and perimeter with gap aperture). Of the other measures of gap size, gap aperture was best correlated with canopy gap perimeter (r2=0.82).  2.4. Discussion.  Percent of the forest in each canopy category. The 18% canopy gap observed at Cypress is higher than the mean for either temperate or tropical forests (temperate mean=17.0±10.4; tropical mean=13.3+15.0; Table 3.1), but less than the extreme values for either. There are fewer studies that have reported values for expanded gap, but the 52% reported here is substantially higher than has been reported elsewhere (Runkle 1982; Veblen 1985). The amount of open space in a forest at steady state is a balance between the rate of creation of gaps and filling in of gaps (see chapter 3; Paine and Levin 1981; Runkle 1982). The high overall percent gap in the forest at Cypress Park (canopy gap + expanded gap = 70%) probably results from a lower rate of filling in of gaps, leading to a longer mean residence time for open space and a higher steady state percent gap. Why would gaps persist longer before being filled at Cypress Park than elsewhere? Growth rates in gaps, especially in the smaller size classes of saplings, are very slow compared to other gap-regenerating systems that have been studied (Runkle 1982; Brokaw 1985b; Runkle and Yetter 1987; Canham 1988a; Uhl et al. 1988). Snow persists in gaps up to several weeks longer in the spring than in the adjacent closed canopy areas (in each of the four stands, in every year of the study, spring snowmelt under a closed canopy preceded adjacent gap areas by one to three weeks), and, until trees are tall enough to extend above the snowpack, this reduces their growing season substantially. Larger trees (>~3m), which become exposed while there is still a snowpack, are able to grow at substantially higher rates than those reported here. A prolonged delay between the creation of a gap and it being filled can also be inferred from the distribution of "ages" of primary gapmakers (many gaps may not have been  34  filled even after 100-200+ years; see below). The large proportion of gap in the forest is thus a consequence of the long tenure of gaps before they are filled, which is in turn a consequence of short overall growing season, slow growth rates of young trees, and initial suppression of growth in gaps by persistent snowpack.  Size and age stmcture of the canopy trees. Most canopy trees examined exhibited three general patterns of radial growth: a period of early suppression, short bursts of growth over much of their life, and peak growth during the 20th century. These growth histories reflect a long history of interactions between competitive status and climate. Similar individual growth histories have been found in the Cascade Range of western Washington state where individual dominant trees of both western hemlock and Pacific silver fir can exhibit substantial volume and height growth late in their lives (Long 1976; Grant 1980). A maximum in growth during this century has been observed regionally in montane to high elevation forests, and appears to be the result of long-term climatic warming (Graumlich and Brubaker 1986; Graumlich et al. 1989). Graumlich et al. (1989) found, for similar forests in Washington state, that net primary productivity was under fairly close climatic control, and correlated well with changes in summer temperature. However, most trees in the forest do not express this climatic release: most trees are below canopy status and are suppressed by their larger neighbors. Only under periods of competitive release will they reflect such climatic trends. The short bursts of growth are much more likely to reflect fluctuations in the local competitive environment, or the coincidence of reduced competition and climatic amelioration, than climate alone. They were not synchronous among trees, and were of a frequency not apparent in the palaeoclimatic record (Graumlich and Brubaker 1986). Patterns of multiple periods of suppression and release in a sequence of gaps are common in a variety of gap regenerating forests, with trees often requiring 2-5 periods  35 of release to reach the canopy (Henry and Swan 1974; Oliver and Stephens 1977; Lorimer 1980; Canham 1988a, 1989; Runkle and Yetter 1987). The ability of large trees to respond strongly to changes in climate and local competitive regime suggests that losses of live biomass through mortality will be compensated for by the subsequent growth of neighbors or new recruits. Long (1976) found that aboveground tree biomass continued to increase through the oldest stand examined (550 years) in an Abies amabilis Zone watershed in the Cascades. He concluded that through the period of canopy thinning and gap formation characterizing the transition from a mature stand to old growth, mortality was compensated for by the growth of both large canopy trees and younger recruits to the overstory.  Few data  exist on the carbon, budgets of old growth forests in this region, but those that do, suggest that, at worst, stands approach a carbon steady state (Grier and Logan 1977; Spies et al. 1988). Given the apparent sensitivity of the system to a warming trend in climate (Graumlich et al. 1989), and the growth potential of older trees, we might anticipate substantial increases of aboveground living biomass in the future.  The sapling layer. Saplings were present in all parts of the forest, but occurred most densely in expanded gaps, and sustained the best growth in gaps. The great majority of saplings were Pacific silver firs, and most of them (81%) showed evidence of suppression at some time in their life. The distribution, size, and age structure of the sapling population indicate that it represents the net result of demographic processes over more than 200 years. Expanded gaps had the highest density of saplings because they included both new recruits and the suppressed population that predated the gap. The low density of saplings in canopy gaps probably results from the increased disturbance in the gap centre by falling bark, branches, and boles of gapmakers, and disturbance to the soil by windthrown root mounds. The high incidence of open growth patterns among silver firs in canopy gaps may be as much a consequence of gapmaker-induced  36 mortality among the suppressed population as increased resource levels at the gap centre. The restriction of western hemlock to woody substrates is consistent with what has been seen at lower elevations (McKee et al. 1980; Christy and Mack 1984), and in montane and sub-alpine forests in the Cascades of Washington (Thornburgh 1969; Long 1976). However, the increased importance of stumps relative to logs has not been reported previously. The main hypotheses advanced to explain the nurse log phenomenon in western hemlock are 1) logs remain wetter during late summer droughts (Place 1950 - from Thornburgh 1969; Long 1976), 2) logs shed leaf-litter that can bury small hemlock seedlings (especially large accumulations of leaf litter melting out of the winter snowpack) (Christy and Mack 1984; Thornburgh 1969), 3) logs emerge from the snow in the spring prior to the forest floor (Thornburgh 1969), and 4) competition from herbs and mosses is more intense on the forest floor than on logs (Harmon and Franklin 1989). These hypotheses are not mutually exclusive, and all appear to be supported by the data in the above references. These arguments should apply equally well to stumps. At Cypress Park, stumps often stand higher than most logs, and are the first substrates in gaps to emerge from the snow in the spring - well before the logs in most cases. Thornburgh (1969) presents data indicating that logs that originated as snags provide a poor substrate for seedlings because their smooth surface doesn't collect litter and build a humus layer to nearly the extent that windthrown logs do. The rough surface of logs which have retained their bark collects leaf litter and builds a humus profile similar to that of a soil horizon. The high proportion of snags among gapmakers at Cypress may make logs there less suitable in general as nurse-substrates. .Also, logs from Pacific silver fir, the most common gapmaker species, do not provide as good a substrate as Douglas fir (Thornburgh 1969). Stumps at Cypress Park commonly have a jagged surface that collects leaf litter, and "old" through "very old" stumps often display an accumulating humus profile of the  37  sort described by Thornburgh (1969). Unlike most logs, stumps also raise seedlings above the dense shrub layer in gaps. Though canopy trees commonly exhibited evidence of having germinated on stumps, they never showed the collonade pattern typical of trees which have established in rows on logs.  Species diversity of the understory layer. Species diversity was higher in gap environments than under closed canopy because, while there were no species that occurred exclusively or primarily under a closed canopy, there were many which occurred at least predominantly in gaps. The strong influence of western hemlock canopies on the composition and dynamics of the surrounding forest has been observed previously (Stewart 1986a,b; Spies and Franklin 1989b). However, the strong inverse correlation between herb and shrub species diversity and the representation of western hemlock among canopy trees is remarkable.  Similarities and differences among stands. There was substantial variation among the four stands in the species composition of the canopy layer, in density of the sapling layer, and in the diversity of the herb and shrub layer. Some of this variation is likely topographically induced (e.g. for HSC2), and some less easily explained. However, despite this variation, overall gap-phase structure (% closed canopy, % gap) was similar between stands, as was the species composition of the sapling layer. The gap disturbance regime, and its consequences for regeneration patterns, appear consistent among stands. Based on the species composition of the sapling layer, we would predict similar successional processes in all stands: the steady replacement of all other species by Pacific silver fir. Patterns of replacement and their community-level consequences will be dealt with further in Chapters 4 and 5.  38 Gapmakers: mode of mortality and number per gap. The stands at Cypress Park are unusual in the degree to which standing mortality dominates the gap forming processes. Most studies which have examined the mode of mortality and its consequences among gap-forming trees have focussed on windthrow and snapping of boles (Brewer and Merrit 1978; Putz and Milton 1982; Naka 1982; Romme and Martin 1982; Foster and Reiners 1983; Orians 1982; Putz et al. 1983; Brokaw 1985a; Veblen 1985; Lawton and Putz 1988; Uhl et al. 1988). However, standing death has been reported as an important mode of mortality in a variety of tropical and temperate gap-regenerating forests (Lieberman et al. 1985; Veblen 1986; Armesto and Fuentes 1988; Martinez-Ramos et al. 1988; Taylor and Zisheng 1988b), and it is increasing in importance among high elevation spruce in north eastern North America, possibly as a consequence of air pollution (Foster and Reiners 1986). Windthrow or wind-related breakage appear to be dominant forms of non-catastrophic mortality in many other coniferous forests of northwestern North America (Franklin et al. 1987; Franklin and Debell 1988) as well. Franklin et al. (1987) described a gradient of decreasing importance of windrelated mortality from coastal areas of Washington and Oregon to the drier forests east of the Cascades. However, Cypress Park is within 3 km of the Straits of Georgia to the west and 7 km to the south, and experiences a climate with a strong maritime influence (though on a relatively sheltered coast). Strong winds do occur, and the south facing edge of stand HSC1 bordering on a clear cut has experienced a high rate of windthrow since the harvesting took place. Because windthrow is more common in poorly drained areas where rooting is restricted (Gratkowski 1956 [in Franklin et al. 1987]; Hartshorn 1978), it may be that the generally deep, well drained soils at Cypress reduce susceptibility to windthrow. The forests at Cypress Park are also unusual in the predominance of multiple gapmakers (though see Armesto and Fuentes 1988; Taylor and Zisheng 1988b). This is particularly notable because, while some gapmakers died at the same time, most  39 gaps represent the combined results of different gapmaking events separated substantially in time. Repeat disturbance and expansion of canopy gaps by mortality among peripheral trees has been observed in other systems (Foster and Reiners 1986; Runkle and Yetter 1987), and in one tropical forest (Lawton and Putz 1988) new gaps were shown to occur closer to old ones than expected by chance. However, the slow rate of filling in at Cypress Park enhances the consequences of such processes. Though treefalls are more frequent at the edges of gaps (Hubbell and Foster 1986b; Lawton and Putz 1988), sequential gap-making episodes appear to be less often of primary importance in the tropics, where turnover times and rates of filling in of gaps are faster (compare Runkle 1985 with Brokaw 1985a; and Runkle 1982 and Runkle and Yetter 1987 with Brokaw 1985b and 1987; see Chapter 3). A gap disturbance regime dominated by standing mortality produces regeneration opportunities very different from a regime dominated by windthrow. Windthrow occurs suddenly, and just as suddenly changes the moisture and temperature regimes at the soil surface, and the competitive regimes above and below ground. A tree succumbing to pathogens may die over a number of years, slowly losing foliage and root volume. Standing mortality creates relatively small gaps slowly, and favors tolerant species, such as Pacific silver fir, that are able to take advantage of intermediate light environments. It also favours those individuals already present relative to those which germinate after gap formation (Putz et al. 1983). In such a system, species dependent on woody substrates that do not become available until some time after gap formation are at a disadvantage.  Size of gaps. Canopy gap sizes are small relative to many other forests (Naka 1982; Romme and Martin 1982; Runkle 1982; Brokaw 1985a), though expanded gap areas are similar to other forests where expanded gaps have been measured (Runkle 1982; Veblen 1985). A predominance of small canopy gap sizes is consistent with the high frequency  40 of standing mortality, which should form smaller gaps than windthrow (Putz et al. 1983). The size of expanded gaps should be less dependent on the mode of gapmaker mortality because it is a function of the spacing of the canopy trees that remain after the gap has formed. Also, with time after the formation of a gap, bordering trees grow laterally to fill a canopy gap (Runkle 1982; Runkle and Yetter 1987), so that the canopy gaps of the very old gaps I measured will have shrunk substantially relative to their original size. Expanded gap size will only be affected by mortality among bordering trees, or recruitment of new bordering individuals to canopy size classes, and will thus be less affected by time since gap formation. The disproportionate contribution of large gaps to the total gap area in the forest has been observed in other systems (Lawton and Putz 1988). This should be particularly important where gap colonizing species have a threshold size for gaps in which they can successfully recruit (Brokaw 1985a,b; Denslow 1980). I will examine this idea for the Cypress Park system in Chapter 4. Both negative exponential (Brokaw 1982a; Foster and Reiners 1986; Lawton and Putz 1988) and lognormal (Naka 1982; Runkle 1982; Hubbell and Foster 1986) patterns of gap size distributions have been observed. To some extent, different processes of gap formation may account for differences in the shape of size distributions. For instance, in many tropical forests, branchfalls create many small gaps (Brokaw 1982a; Lawton and Putz 1988), but branchfall is not a significant gap forming process in temperate forests.  Forest Turnover Time It would be useful to be able to summarize data on the creation and filling of gaps in a single statistic that represented the overall rate of turnover of space in the forest. Forest turnover time can be calculated in several ways (Hartshorn 1978; Runkle 1982; Romme and Martin 1982; Bongers et al. 1988). I used a method based on the proportionality between the amount of forest in gap and canopy phases and the amount  41 of time spent in each phase (see equation 3.2 for calculating TT2; see Chapter 3 for discussion of turnover times and analyses of the methods for calculating them). Table 2.13 shows turnover times and the time that trees are resident in the canopy after filling a gap (Tres). given the observed 18% of the forest in gap, and a range of estimates for the time taken to fill gaps (Tfjn). For a known percent of the forest in gap and a given Tfill, Tres is fixed, and turnover time = Tfj|| + Tres- If, on average, gaps take 125 years to fill, mean turnover time for this forest should be in the neighborhood of 694 years, and residence times for canopy trees should be near 569 years. This is consistent with the data on tree ages and lengths of suppression. If this turnover time is accurate, it would be among the longest that has been reported for a gap regenerating forest. The estimates I used for Tfj|| in Table 2.13 are substantially longer than those reported in other systems (Romme and Martin 1982; Runkle 1982; Veblen 1985; Bongers et al. 1988), though Foster and Reiners (1986) estimated a maximum gap age of 100 years for a sub-alpine forest in New Hampshire. These values are consistent with the expected Tfj|| from the decay class distribution of gapmakers. As well, such long Tfjn's are a necessary consequence of the observed combination of percent gap, ages of canopy trees and the length of time commonly spent in a suppressed state before reaching the canopy. The period of suppression observed in canopy trees, however, is not equivalent to the Tfj|| for the gap(s) in which they recruited - it is probably substantially longer and represents portions of several Tfjn's for different gaps. Thus, the total longevity of canopy trees will often exceed the turnover time (Runkle 1982). Without data on decay rates and/or detailed stem analysis of gapfillers, it is not possible to estimate accurate Tfjn's. In addition, gaps will take a variable length of time to fill, and trees will occupy the space filled for a variable length of time, irrespective of how long the gap took to fill. Many combinations of Tfj|| and Tres probably occur, and an accurate estimate of the true mean (or median) turnover time could only be  42 calculated if the distribution of Tfjn's were known. The range 50-200 years used in Table 2.13 should bracket most of the variation in actual Tfj||. There were several gaps in HSC2 that only contained gapmakers in an advanced state of decay, yet had little tree regeneration growing within them. They resembled the shrub meadows of the sub-alpine parkland 100-200m higher in elevation. The trees that were present were mostly less than 1 m in height and resembled the young trees invading sub-alpine meadows in response to the warmer 20th century climate (Brink 1959; Franklin et al. 1971). It may be that the little ice age effectively suppressed tree growth in some gaps in HSC2, maintaining them as persistent gaps with effectively infinite Tfjn's. The accuracy of the turnover time estimate also depends critically on the amount of gap in the forest being at steady state. There are no data with which to assess this directly. Two lines of evidence suggest the possibility that stand structure could be at steady state. The first is the extremely long period of stand development (1,500+ years; Lertzman and Brubaker, in prep); if such forests can ever reach steady state, then this one should be a candidate. Second, the models of Chapter 5 demonstrate that despite the dominance of Pacific silver fir among gapfillers, species composition may be close to a stationary distribution. However, even if the stand structure does not reflect the initial disequilibrium of early stand development (Figure 3.3), given the known changes in climate during the stand's history (Lamb 1982; Graunmlich and Brubaker 1986; Ritchie 1986), it seems likely that stand structure contains climatically induced deviations from steady state (as in Figure 3.6). If this is true, then the turnover time estimate developed here may be in error in either direction. It is less likely to be very wrong than an estimate based on yearly gap creation rate (Chapter 3). Even if the estimate of 694 years is high by 30%, the stands at Cypress are well above the average for temperate forests (Tables 3.1, 3.2), and the general conclusion of slow turnover is valid.  43  In summary, the long turnover time for the Cypress Park forest reflects the slow time scale of several important component processes of turnover. In the absence of major disturbances to the canopy, potential recruits are first suppressed by canopy dominants, and then by persistent spring snowpack in gaps, effectively delaying the process of filling gaps. Because canopy gaps are narrow, due to the primacy of standing death as a mode of mortality, they may often be filled by the lateral growth of adjacent canopy trees before new recruits are able to fill them from below. This results in new recruits requiring several periods of suppression and release in successive gaps before they can recruit to the canopy. As a consequence, the faster growing, more light demanding species, western hemlock, occurs in much fewer numbers than the slower growing, but more shade tolerant Pacific silver fir. There are few data with which to assess the potential generality of such long turnover times in the old growth forests of northwestern North America. Old ages of individual trees or stands are well known (Fowells 1965; Waring and Franklin 1979; Hemstrom and Franklin 1982), but data on the dynamics of space or individuals are rare. Fire interrupts stand development on a time scale precluding such long turnover times in many forests (Franklin and Hemstrom 1982; Hemstrom and Franklin 1982; Franklin 1988), but there are many wet forest types on the British Columbia Coast for which no data exist. The Cypress Park forest may be unusual in the length of time without catastrophic disturbance, but it is unlikely that it is unique.  44  Table 2.1 Decay classes for gapmakers.  Characteristic Class  Small Large Branches Branches Bark  Heartwood  Sapwood  Log  Young  present  present  intact  solid  solid  present  Medium  absent  present  sloughing friable  shell  Old  absent  absent  absent  variable^  variable  decayed  V. Old  absent  absent  absent  variable  variable  absent  3  0  0  present e  a  The sub-categories young-medium and medium-old were distinguished largely on how much bark had sloughed off; y-m, 25%; m, 50%; m-o, 75%.  D  The heartwood of standing dead trees rotted first, leaving a hard exterior shell (Thornburgh 1968). The pattern was the opposite for logs on the ground, as in Triska and Cromack (1979) and Sollins (1982).  c  Bark is sometimes present on stumps of old and very old gapmakers, but never on the logs associated with the stumps. Bark persisting on the stumps of most old and all very old gapmakers was waterlogged and covered by decomposing litter.  d Portions of old and very old gapmakers belowground or covered with leaf litter and moss were often waterlogged and sound, with distinct ring structure, though heavily stained and soft. e  Logs associated with old gapmakers were being incorporated into the forest floor and are well characterized by the descriptions of Class 5 logs in Triska and Cromack (1979) and Sollins (1982).  45  Table 2.2 Frequency of canopy gap, expanded gap, and closed canopy among the four stands. Values are means of the among transects in each stand ± 1 standard deviation. N is the number of transects; each parallel transect wihtin each stand is a datum, and 50m and 100m transects are not weighted differently. The stands do not differ significantly, ANOVA (on arcsin square root transformed data); C G , p=.13; E G , p=.54; CC, p=.99.  Stand  % Canopy Gap  %Expanded Gap  % Closed Canopy  N  HSC1  21±6  50±15  29±20  4  HSC2  22±4  47+19  28+19  3  HLY1  20+6  51+4  29+9  3  NST1  12±6  60±6  28±11  4  Overall Mean  18+7  52+12  29+14  14  46  Table 2.3 Species composition of canopy trees. Densities within each table are + one standard deviation. Total stand density at the top of each table is given with its 95% confidence interval, a. Overall summary of stands HSC1, HSC2, and HLY1. The values in parentheses for percent composition also include data from stand NST1. b. Data for stand HSC1 alone, c. Data for stand HSC2 alone, d. Data for stand HLY1 alone. a. All Stands: Total density=200.3/ha (172.8-231.7). Species  Density (trees/ha)  Percent Composition  Relative Dominance  N  Pacific silver fir  90.2+6.8  45.0 (43.2)  29.8  81  western hemlock  49.0+3.7  24.4 (35.6)  38.6  44  mountain hemlock  42.3+3.2  21.1 (15.0)  18.7  38  yellowcedar  18.9+1.4  9.4 (6.3)  13.0  17  a  a Composition for stand NST1 alone: Pacific silver fir: 46.2%, western hemlock: 50.2%, mountain hemlock: 3.3%, yellow-cedar: 0.3%.  b. Stand HSC1: Total density=154.3/ha (119.4-198.0) Species  Density (trees/ha)  Percent Composition  Relative Dominance  N  Pacific silver fir  86.0+11.3  55.7  39.5  34  western hemlock  50.6±6.6  32.8  51.1  20  mountain hemlock  17.7±2.3  11.5  9.4  7  0  0  0  yellowcedar  0  47  Table 2.3 continued.  c. Stand HSC2: Total density=273.5/ha (210.7-352.4) Species  Density (trees/ha)  Percent Composition  Relative Dominance  N  Pacific silver fir  55.6±7.3  20.3  12.7  12  western hemlock  64.9±8.5  23.7  29.3  14  mountain hemlock  97.3±12.8  35.6  36.9  21  yellowcedar  55.6±7.3  20.3  21.1  12  d. Stand HLY1:: Total density=201.7/ha (155.7-259.3) Species  Density (trees/ha)  Percent Composition  Relative Dominance  N  Pacific silver fir  117.6+15.4  58.3  31.5  35  western hemlock  33.6±4.4  16.7  31.9  10  mountain hemlock  33.6±4.4  16.7  15.6  10  yellowcedar  16.8±2.2  8.3  21.1  5  48  Table 2.4 Comparison of the mean diameters (at breast height) of canopy trees of all species among the four stands.; HSC2 differs from all the other stands (ANOVA with Tukey-Kramer HSD test, p<0.005 for all comparisons), which do not differ from each other significantly (ANOVA with Tukey-Kramer HSD test, p>0.75 for all comparisons).  HSC1  HSC2  HLY1  NST1  Diameter  82.2  65.2  83.8  85.2  S.D.  30.7  22.4  39.7  34.9  Mean  49  Table 2.5 Species composition of the sapling layer. Data are from stands HSC1 and HSC2. Densities within the body of each table are + one standard deviation. Overall stand density is given with its 95% confidence interval. a. Overall Composition: HSC1 and HSC2 combined: Total density=1512/ha (1331 -1714). Density (trees/ha)  Percent Composition  N  Pacific silver fir  1230±80  81.3  196  western hemlock  188+12  12.5  30  mountain hemlock  63±4  4.2  10  yellowcedar  31+2  2.1  5  Species  b. Stand HSC1: Total density=1189/ha (1008-1399) Density (trees/ha)  Percent Composition  N  Pacific silver fir  933+78  78.5  113  western hemlock  206+17  17.4  25  mountain hemlock  25±2  2.1  3  yellowcedar  25+2  2.1  3  Species  50  Table 2.5 continued. c. Stand HSC2: Total density=2493/ha (2037-3038). Density (trees/ha)  Percent Composition  N  Pacific silver fir  2133+220  85.6  83  western hemlock  129+13  5.2  5  mountain hemlock  180+19  7.2  7  51 ±5  2.1  2  Species  yellowcedar  51  Table 2.6 Densities, heights, and growth rates of the sapling layer under each canopy category. Values are given + one standard deviation. Data are from stands HSC1 and HSC2.  Density (Trees/ha)  N  Height (m)  N  Growth Rate (cm/year)  Canopy Gap  339±22  53  1.0±1.0  53  3.4±4.8  Expanded Gap  786+51  123  1.4+3.3  123  3.1+3.5  202  Closed Canopy  409±27  64  1.1+1.1  64  1.4+1.2  98  D  C  3  N 80  a Measured as leader extension growth for each of the 2 previous years for all Pacific silver firs. b Mean heights not significantly different among canopy classes; ANOVA, p=0.492. c Canopy Gap growth rate = Expanded Gap growth rate > Closed Canopy growth rate; ANOVA and Tukey-Kramer HSD test. For Canopy Gap:Closed Canopy, p<0.001, for Expanded Gap:Closed Canopy p=0.001.  52  Table 2.7 Frequency of different growth patterns in the sapling population. Data are from stands HSC1 and HSC2. Percentages are in parentheses. Open grown saplings showed regular, substantial height increment in each branch internode, and an even, full distribution of foliage along their branches. Suppressed saplings showed irregular patterns of branches and internodes with no periods of substantial height increment, and a sparse and uneven distribution of foliage. Saplings in the "suppression-release" category showed one period of early suppression followed by one of released height growth. "Multiple suppressionrelease" saplings had several periods each of suppression and released height growth.  Open Grown  Suppressed  SuppressionRelease  Multiple Sup.Release  Total  36  78  33  43  190  (19)  (41)  (17)  (23)  (100)  53  Table 2.8 Distribution of saplings of each species among canopy classes. Numbers are frequencies, row percents in parentheses.  Canopy Class Species  Canopy Gap  Expanded Gap  Closed Canopy  Total  Pacific silver fir  45 (23.0)  97 (49.5)  54 (27.6)  196  western hemlock  2 (6.7)  19 (63.3)  9 (30.0)  30  mountain hemlock  6 (60.0)  2 (20.0)  2 (20.0)  10  yellowcedar  0 (0.0)  5 (100.0)  0 (0.0)  5  53 (22.0)  123 (51.0)  65 (27.0)  241  Total  54  Table 2.9 Distribution of saplings of each species among substrates. Numbers are frequencies, row percents in parentheses.  Substrate Species  Forest Floor  Decaying Wood  Stumps  Total  Pacific silver fir  141 (72.0)  23 (11.7)  32 (16.3)  196  western hemlock  4(13.3)  12 (40.0)  14 (46.7)  30  mountain hemlock  1 (20.0)  2 (40.0)  2 (40.0)  5  yellowcedar  5 (50.0)  2 (20.0)  3 (30.0)  10  151 (62.7)  39 (16.2)  51 (21.2)  241  Total  55  Table 2.10 Number of gapmakers of each species, by type of mortality. Values are totals for primary plus secondary gapmakers, in parentheses are numbers of primary gapmakers. HEM are hemlocks identifiable only to genus, UNKN are individuals in a sufficiently advanced state of decay that their species or the original mode of mortality could not be determined.  Type of Mortality Species  Windthrown  Snapped  Standing  UNKN  Total  Pacific silver fir  17(10)  21 (8)  49 (35)  15(4)  102 (57)  western hemlock  1(1)  9(5)  20 (19)  7 (2)  37 (27)  mountain hemlock  0(0)  1 (0)  3(2)  1(1)  5(3)  HEM  0(0)  7(5)  3(3)  3(3)  13(11)  yellowcedar  0(0)  0(0)  1(1)  2(2)  3(3)  UNKN  1 (1)  7(4)  3(1)  165(54)  176 (60)  19 (12)  45(22)  79 (61)  193 (66)  336 (161)  Total  56  Table 2.11 Number of gapmakers in each age/decay class, by type of mortality. Values are totals for primary plus secondary gapmakers, in parentheses are numbers of primary gapmakers. See Table 2.1 for explanation of decay classes. UNKN are individuals in a sufficiently advanced state of decay that their original mode of mortality could not be determined. Column totals as in Table 2.10.  Type of Mortality Class  Windthrown  Snapped  Standing  UNKN  Total  Young  4(2)  6(2)  20 (16)  0(0)  30 (20)  Yng-Med  4(2)  5(2)  28 (23)  0(0)  37 (27)  Medium  9(7)  15(7)  22 (16)  2(1)  48 (31)  Med-Old  1 (0)  13(8)  5(4)  12(6)  31 (18)  Old  0(0)  6(3)  4(2)  86 (25)  96 (30)  V. Old  1  0(0)  0(0)  93 (34)  94 (35)  0)  57  Table 2.12 Summary statistics for gap size measures. CG=canopy gap, EG=expanded gap-  C G Area  CG Perim  EG Area  EG Perim  Aperture  (m2)  (m)  (m2)  (m)  (degrees)  37  37  60  60  37  MIN  5  9  25  20  5  MAX  525  143  1127  176  33  MEDIAN  41  31  203  56  13  MEAN  77  40  286  67  13  SD  108  29  244  34  6  N  58  Table 2.13 Forest turnover times calculated according to equation 3.2, given 18% canopy gap in the forest. Tfj|| = time for a newly created gap to fill to the point where it is no longer gap. Tres = time that a tree is resident in the canopy after filling a gap. Turnover Time = (Tfj||)+(Tres) (see text and Chapter 3 for explanation).  Tfj||  T  50  228  278  100  456  556  150  683  833  200  911  1,111  r e s  Turnover Time  59 Figure 2.1 Definitions of canopy gap, expanded gap, and closed canopy. CG=canopy gap, EG=expanded gap, CC=closed canopy, GM=gapmaker, GF=gapfiller.  60  Figure 2.2 Size distributions of canopy trees. PSF=Pacific silver fir; W HEM=western hemlock; M HEM=mountain hemlock; AYC=yellow-cedar.  T  1  1  1  1  1  1  DIAMETER (cm)  i  r  61 Figure 2.3 Ages of canopy trees. Hemlocks refers to individuals of both hemlock species (identified only to genus).  »  I  C\]  I  i  i  ^  1  1  CM  ADN3n03cdd  L  62  re 2.4 Decadal increments of basal area for 6 representative canopy trees. All 6 samples are from clearcut adjacent to stand HSC2. Time series all terminate in the same calendar year (date cut, ca. 1970). Lines are DWLS (Distance Weighted Least Squares) smoothing of data (Wilkinson 1988a,b). a, b, d = Pacific silver fir. c, e, f = hemlocks.  1 J  2.0  b.  1.5  C\J  1.0  E  .o ID Q <  o  0.5  0  100 200 300 400 500 600 700  0.0  •  0  100 200 300 400 500 600 700  ID Q \  hZ LD LLI CC  O  < LD CC <  0  200  400  600  800  1000  0  100 200 300 400 500 600 700  < CO  <  CO  100  200  AGE (years)  300  400  500  63 Figure 2.5 Height vs. age for dead, suppressed Pacific silver firs. Regression line with 95% confidence intervals is shown.  O  100  200 AGE (y«aro)  300  64 Figure 2.6 Changes in vascular plant species diversity of the understory layer among canopy classes. The class "understory" refers to areas under a closed canopy. Diversity is measured as the inverse of Simpson's index (Krebs 1989). Data from each stand are as per the key on the figure. The hollow stars represent stand HSC2 when mosses are included in the species counts.  65 re 2.7 Changes in the diversity of understory layer among stands. Diversity values calculated for all canopy classes together. Diversity is measured as the inverse of Simpson's index (Krebs 1989). Linear regression and 95% confidence intervals are shown. Regression equation is: y = -0.059x + 6.7 (p=0.005). In order of decreasing y value, stands are: HYL1, HSC2, HSC1, NST1. Mosses are included in the calculation of diversity for HSC2.  % WESTERN HEMLOCK IN CANOPY  66  re 2.8 Frequency distribution of the number of primary and total gapmakers per gap. Triangles and the thinner line = total gapmakers; circles and the thicker line = primary gapmakers. Lines are distance weighted least squares smoothings of the data.  0  5  10  GAPMAKERS PER GAP  15  67  Figure 2.9 Ratio of Pacific silver fir to other species among gapmakers of different ageclasses.  T  1  1  GAPMAKER AGE CLASS  r  Figure 2.10 Size frequency distributions for canopy gaps, expanded gaps, and gap aperture. M=median, m=mean.  69  Figure 2.11 Area-perimeter relationships for expanded gaps and canopy gaps. Solid lines plot the area-perimeter relationship for circles.  EXPANDED G A P A R E A (m2) 175 ,§  I  1  1  1  "  140 -  CANOPY G A P A R E A (m2)  r-  70 3. ON ESTIMATING FOREST TURNOVER TIMES 3.1 Introduction. The turnover of living space from one individual or set of individuals to another is a central process in the population and community ecology of sessile organisms (Dayton 1971; Horn 1975,1981; Paine and Levin 1981; Runkle 1981, 1982; Pickett and White 1985). Patterns in the freeing and subsequent capture of space play a critical role in determining the equilibrium status or trajectory of a community (Chapter 5), and a variety of models of communities focus on these transitions (Horn 1972; Paine and Levin 1981; Runkle 1982). Often we wish to estimate the rate at which space turns over in such systems. In this chapter, I will discuss some of the methods for estimating turnover time in forests where gap-phase processes dominate transitions of space (e.g. Brokaw 1985a,b; Runkle 1982; Foster and Reiners 1986; Bongers et al. 1988). We wish to know not the frequency of catastrophic, stand destroying disturbance (e.g. Heinselman 1973), but the net effect of individual growth, mortality and replacement processes on the average rate at which the space in a forest is cycled from one "generation" of trees to the next. Turnover time has been defined as both the number of years necessary for an area equal to the entire forest to be disturbed by gaps, and the average interval between disturbances at a given point (e.g. Runkle 1982; Romme and Martin 1982; Barden 1989). These two approaches have led to different ways of calculating turnover time. Both measure the turnover of space, and are intuitively reasonable, but they do not necessarily yield the same results (see below; Romme and Martin 1982; Bongers et al. 1988). This is potentially a serious problem if one wishes to interpret differences in estimated turnover times as differences, say, between temperate and tropical forests. Method 1 estimates turnover time (TT1) as the inverse of the rate at which open space is made available (the "birth rate" of new gaps). Method 2 estimates the mean time between gap creation events at any point in the forest (TT2) using a proportionality between the amount of gap and closed canopy area in the forest, and the time spent in  71 each phase. The purpose of this chapter is to examine each method and compare their assumptions, faults and results. Both methods assume that the disturbance regime is in equilibrium; that is, TT1 assumes that the gap creation rate over the measurement period is neither anomalously low or high (relative to the long-term mean), and TT2 assumes that the proportion of the forest that is in gap-phase is constant. These assumptions are easily shown to be invalid for cases in which patterns of mortality are known to have changed recently (Foster and Brokaw 1982; Foster and Reiners 1986; Hubbell and Foster 1986b). Additionally, gap formation rates measured over a short period are often not representative of the longer term average (Romme and Martin 1982; Martinez-Ramos et al. 1988). For many forests, the assumption that forest structure is in equilibrium with a constant disturbance regime cannot be tested. I will use a simple simulation of stand turnover to examine the consequences of departures from steady state on turnover time estimates. We can define 4 parameters associated with forest turnover: 1) Tfjn, the time from the creation of a canopy gap until it has been filled to a specified height and is no longer considered a gap (Brokaw 1982b), 2) Tres. the length of time that a tree is resident in the canopy after filling a gap, 3) BRg, the rate at which open space in canopy gaps is created by the death of canopy trees or fall of major branches (proportion of area in new gaps each year), and 4) Pgap, the proportion of forest area represented by canopy gap at any given time. The Pgap at steady state is a function of the relative values of Tfj|| and BRg, or alternatively, of Tfjn and T es- Turnover time can r  be calculated as either TT1=1 /BRg or TT2=(Tfj||+T"res)Some authors have recognized an area classed as "expanded gap" under the foliage of the trees whose canopies bound the canopy gap (Runkle 1982; Veblen 1985; Chapter 2), and Veblen (1985) calculates turnover times based on gap area including canopy and expanded gap as well as the traditional canopy gap-based estimate. Here, I will restrict the discussion to estimates based on canopy gap area. Though expanded  72 gap may be important for regeneration, the meaning of a turnover time based on expanded gap area is less clear.  3.2 One Parameter Method: The Inverse of Gap Birth Rate. The calculation of turnover time as 1/BRg (i.e. TT1), has been common in the literature, and somewhat more common in tropical than temperate studies (Table 3.1). In most cases, this has required repeated measurements of a mapped stand to identify the area in new gaps over a known period, though Runkle (1982) and Martinez-Ramos et al. (1988) estimated BRg from gap age structure. If mortality is episodic, even studies over modestly long time scales (10 years) may produce estimates of gap creation rate that are not representative when averaged over a longer period (Romme and Martin 1982). This method has the advantage that only one parameter needs to be estimated. In circumstances where we are uncertain about the quality of our data, this in itself can be important (Ludwig 1989; Ludwig and Walters 1989). It seems to be the preferred method since in cases where data are available to calculate both TT1 and TT2, often only TT1 is calculated (see cases in Table 3.1 where I have calculated TT2 from data presented in a reference where the authors did not). Estimating TT1 has the disadvantage that it requires permanently marked plots with censuses over a sufficiently long period to obtain "representative" gap creation rates. In practice, little attention has been paid to adjusting the duration of the census interval to match the variation in mortality rates because in most cases the periodicity of mortality is both unknown and likely to be longer than research programmes (Romme and Martin 1982; Bongers et al. 1988; Martinez-Ramos et al. 1988). Estimates of TT1 vary from 47 to 278 years (Table 3.1).  73 3.3 Two Parameter Method: Time to Fill Plus Residence Time. This method is based upon an assumption of proportionality between the amount of forest area occupied by different stages of maturity, and the time spent in each. We can express the time-area proportionality as: Tfill Pgap = - -  (TfiH+Tres)  (3.1)  The proportion of the forest that is gap equals the ratio of the time a particular point spends in gap to the total time between gap creation events. We can then calculate the turnover time as (Tfjn+Tres). so: Tfill TT2 =  Pgap  (3.2)  Because Pgap is fixed and relatively easily measured at any point in time, and both Tfjn and Tres are highly variable in space and time, equation 3.2 is preferred to calculating TT2 as (Tfjn+Tres) directly. This calculation has been applied primarily in studies of temperate forests (Table 3.1). Bongers et al. (1988) divide Tfj|| into two sub-components, the gap phase, lasting only a year, and the building phase, lasting until the gap is filled, but they combine the gap and building phases in calculating the turnover time. The estimation of TT2 has the advantages that it can be based on data gathered from a one-time census, and that the overall proportion of forest in gap may be less sensitive to yearly fluctuation in mortality rates than is 1/BRg. Because Pgap is determined by gaps of all age classes, it will exhibit less yearly variation than will the number of current year gaps. TT2 has the disadvantage that two parameters need to be estimated, and Tfjn will often be variable and difficult to measure. Estimates of TT2 vary from 60 to 794 years (Table 3.1). Figure 3.1 shows equation 3.2 plotted for the range of parameter values commonly found in the field (Table 3.1). The behaviour of equation 3.2 is as we might  74 expect: long tenure of canopy trees and long times for gaps to fill in result in long turnover times, and short lived canopy trees with fast growing saplings result in fast turnover times. What is perhaps more interesting, is that in the range of values for Tfj|| and Pgap common among tropical forests (Tables 3.1 and 3.2), the slope of the surface is quite steep. Small errors in estimates of Tfjn and Pgap will result in large errors in the estimated turnover time.  3.4 Comparing the Two Methods: Data.  Comparing TT1 and TT2 There are few studies for which TT1 and TT2 can be compared using data from a single forest. I found only 6 studies in tropical and 3 in temperate forests where estimates of TT1 and TT2 were presented, or where I could calculate the second estimate from data provided in the paper (Table 3.1). However, in all but one of these cases, the estimate for TT2 exceeded that for TT1, sometimes substantially. In the case where TT1 exceeded TT2 (Romme and Martin 1982), the authors suggest that the disparity in estimates may have been due to errors in estimating gap age (for calculating Tfj||) and variation over the study area in gap creation rate. Gap creation rates were also heterogeneous in time, causing the disparity in TT2 estimated from gaps created over a 10 year period vs. TT2 estimated from gaps created over a 20 year period (Table 3.1). Why might TT2 be consistently longer than TT1 ? One explanation is based on TT1 's greater sensitivity to fluctuations in disturbance regime. This was Bongers et al.'s (1988) rationale for the difference between their estimates of TT1 and TT2 - that the shorter TT1 expressed a recent increase in disturbance in the plot. However, it seems unlikely that 8 out of 9 forests in Panama, Costa Rica, Mexico, the Eastern United States and Japan should all have experienced a recent increase in disturbance frequency (compare TT1 and TT2 for each forest in Table 3.1). Another explanation is  75 suggested by the simulation results presented below: that the forests have not reached a steady state structure, and the two estimators are responding differently to changing stand structure. Overall, the TT2's are substantially longer than the TT1 's (Table 3.2). However, we cannot separate this from the cross-system comparison because the TT1 's are largely from the tropics and the TT2's largely from the temperate forests (see below). We are left with the tentative conclusion that TT1 and TT2 are either measuring different aspects of the turnover process, one of which is often substantially longer than the other (which doesn't make sense intuitively), or that as estimates of the same thing, "the true turnover time", either or both measures are biased.  Comparing Estimates From Tropical and Temperate Forests Hartshorn (1978) predicted that tropical forests are more dynamic than temperate forests. One measure of dynamism is turnover time: tropical forests should have faster turnover times. Do the data summarized in Tables 3.1 and 3.2 support this prediction? Mean turnover time (both TT1 and TT2) is longer for temperate than tropical forests (Table 3.2). However, the sample sizes are small and the variances large. The comparisons are made more difficult because there are more examples of TT1 for tropical forests and TT2 for temperate forests. The difference between systems cannot be separated from the effects of the method of calculating turnover time. At this point then, though the data give the appearance of supporting Hartshorn's (1978) prediction, nothing can be concluded with certainty. This analysis awaits more data that have been collected in a more directly comparable manner. If temperate forests really do have longer turnover times, then obtaining an accurate estimate of TT1 for some systems may require truly long term studies. For instance, at my own sites (this thesis), between 1984 and 1988 I observed no new gap creation events. Until such data are available, the development of TT2 estimates for tropical forests may be  76 the best approach to the temperate-tropical comparison. Because of the sensitivity of the estimate to small errors in Tfjjj for the range common in tropical forests, particular care will be required in estimating Tfj||. One factor leading to faster turnover in tropical forests is much shorter Tfjn's (Table 3.1; and see simulations below, Figure 3.3). An explanation of the shorter Tfjifs may relate to the presence of very rapidly growing pioneer species which can rapidly create a nearly complete canopy at 2-10 m even in large gaps (Brokaw 1985a,b).  3.5 Comparing The Two Methods: Simulations.  The Model To examine the discrepancy between the estimates of TT1 and TT2, I built a model which simulates the creation of gaps in an age-structured stand at a known rate, calculating both TT1 and TT2 at each time-step. The model simulates a 1,600 cell matrix (40x40) of trees, each a potential location for a gap (though gaps from adjacent trees can anastomose, all are treated as independent one-tree gaps). There are 3 input parameters: Tfill, Trmin, and MR (here Tfill and Trmin are not subscripted to make the distinction between model parameters and field estimates). Tfill is, as above, the length of time after a tree dies that the space formerly occupied by its canopy is considered to be gap. Trmin is the minimum time that a tree is resident at a location after filling a gap, before it is subjected to mortality. MR is the mortality rate applied to trees older than (Tfill+Trmin). In the baseline simulation, Tfill=50 years, Trmin=100 years, and MR=0.00606. I then varied Tfill from 25 to 150 years. I also ran simulations with a higher mortality rate (0.01212), shorter Trmin (15 years), and a lower range of Tfill's (5-20 years) to produce turnover time estimates more consistent with tropical forest data. Finally, I examined the effect of short pulses of lower and higher mortality rates on variation in TT1 and TT2. The base value for MR of 0.00606 is the mean mortality rate over 36 years, over  77  all species, for trees over 24 cm dbh in an old-growth Douglas-fir stand (Pseudotsuga menziesii) (Franklin and DeBell 1988). This is probably a low estimate of mortality for many forests outside western North America. This model does not represent the competition and sometimes substantial mortality among young trees filling gaps (Brokaw 1985b; Lawton and Putz 1988). In reality, there is no minimum canopy residence time. However, by definition, those individuals that do survive to successfully occupy gaps do not experience mortality until sometime after reaching canopy status. Additionally, mortality among smaller trees will not create gaps in the canopy layer. Trmin can be thought of as the point when a gap would no longer be classified as "gap" in a field census (for instance when vegetation has formed a continuous cover at greater than 2 m height; Brokaw 1982b), and when the tree(s) filling the gap have reached codominant status. There is some empirical support for this approach. For instance, young forest on Barro Colorado Island has a longer TT1 than old forest (Lang and Knight 1983; Brokaw 1982a). In each simulation, all cells begin at 0 years old, and no mortality occurs until (Tfill+Trmin) years. Every year, all cell ages are incremented by 1, and when a cell's age passes Tfill, the total number of gaps is decremented. When cell age passes (Tfill+Trmin) then the tree occupying that cell is available for mortality. Each year, a binomial trial is run for every cell over the threshold age: it is killed or not depending on the result of a random number call. The probability of mortality for each cell is equal to MR.  For each tree killed by this process, cell age is set to 0, and the total number of  gaps incremented. Thus gap-creating mortality is a roughly constant proportion of the population of older canopy trees. There is variation around this proportion because the total mortality each year results from the sum of a large number of stochastic processes. I calculated TT1 in two ways. The theoretical TT1 is equal to MR*(the number of trees older than Tfill+Trmin). The actual TT1 was calculated as the inverse of the proportion of cells reset to age=0 in a given year (the proportion of area in new gaps  78 each year: 1/BRg). As discussed above, this varied from year to year in the proportion it represented of the population of older trees, making it difficult to assess the steady state TT1 value. Therefore, in most cases (i.e. Figures 3.2, 3.3, and 3.5), I compared the theoretical TT1 with TT2. Note that TT1 is not the same as the inverse of MR because the number of trees with ages > (Tfill+Trmin) is equal to the total number of cells only at year=(Tfill+Trmin) - the first year that mortality is applied to the young, even aged canopy. I calculated TT2 as specified in Equation 3.2: TT2 = Tfill/(proportion of total area in gaps of all ages < Tfill).  Model Results Both estimates of turnover time are infinite until after (Tfill+Trmin) because they are both dependent on functions of mortality, which doesn't occur until then (Figure 3.2). They decrease rapidly and then come to steady state in damped oscillations as the spatial structure and age structure of the forest equilibrate. Though TT2 and TT1 are approximately equal at steady state, during middle stand development the estimates can be quite different. The shape of the response of each measure of turnover time is easily explained as a consequence of the pattern of change in the number of older trees (TT1) and the proportion of the forest in gap (TT2) (Figure 3.2). Because age structure and annual mortality rates stabilize faster than the proportion of open space, TT1 stabilizes earlier than TT2. The time taken to fill gaps exerts a strong influence on both the steady state turnover time for each estimator, and on the pattern of change in the estimates (Table 3.3, Figure 3.3). As expected, increasing Tfill results in longer steady state turnover estimates. However, it also prolongs the period when the two estimates can produce very different results, and increases the magnitude of the oscillations about the steady state value. For the shortest Tfill in Figure 3.3 (25 years), the estimates differed little over the course of stand development.  79 The simulations with higher mortality rates and shorter Tfill's and Trmin's produced turnover time estimates similar to those found for tropical forests (Tables 3.1, 3.2; Figure 3.4). Both TT1 and TT2 quickly damped to steady state, and only differed for a short period initially. These simulations have assumed a young stand proceeding to steady state under a constant mortality regime: the TT1's plotted are what would be achieved with no year to year variation in mortality. In the real world, mortality can be quite patchy in space and time (Oliver and Stephens 1977; Romme and Martin 1982; Hubbell and Foster 1986a,b,c; Bongers et al. 1988; Martinez-Ramos et al. 1988; Lorimer 1980). Figure 3.5 shows the estimates for TT1 and TT2 when the actual variation in TT1 as a result of the stochastic nature of the binomial trials is expressed (the variation in TT2 is the same as Figures 3.3). As expected, TT1 is tremendously more sensitive to annual fluctuations in mortality rate than is TT2. It would be difficult, given a short section of the time series shown in Figure 3.5, to estimate the true steady state TT1. Another way to examine the effects of variation in mortality rates on turnover time estimates is to incorporate short pulses of higher or lower mortality rates (in Figure 3.5 above mortality rates were constant over time). For the data shown in Figure 3.6, MR was its baseline value of 0.00606 from 0 to 400 years, increased to 0.01212 for 20 years, dropped back to 0.00606 from 420 to 800 years, dropped to 0.00303 for 20 years, and increased back to 0.00606 from 820 years until the end of the simulation. By 400 years, both TT1 and TT2 were approaching their steady state values. As expected, TT1 responded immediately to the change in gap creation rate, and TT2 more slowly as total area in gaps changed incrementally. TT1 recovered to near its steady state value much faster than did TT2, but both show the effects of the 20 year pulses for > 150 years afterwards, and give fairly different results over much of that time.  80 3.6 What is the "Real" Turnover Time? In general, the model indicates that both TT1 and TT2 are sensitive to the nonequilibrium structure of younger stands. At steady state, TT1 is the same as TT2, but TT2 is farther from its steady state value for a longer period of stand development than is TT1. However, TT1 is much more sensitive to the short term variations in gap creation rate that appear to characterize most systems. TT1 approaches steady state earlier in stand development, and returns faster after perturbations to the gap creation rate, but will more easily give estimates that do not reflect longer term trends. What rules for using TT1 and TT2 can be abstracted from the simulations? 1. If it is likely that the stand has not reached a steady state distribution of age structure or % gap, then neither estimate of the long-term turnover time is likely to be accurate. However, in most cases, TT2 is likely to be less accurate than TT1. 2. If it is certain that the stand has reached a steady state structure, then either estimate will do, and they will produce nearly the same result. 3. For systems with short Tres and Tfill (such as some tropical forests), both estimates should produce similar results much of the time, but if there is any doubt about the steady state status of canopy structure, TT1 should be used. This is reinforced by the steepness of the curves in Figure 3.1 in the region occupied by such systems: small errors in estimating Tfill will cause larger errors in estimating TT2. 4. For systems with very long Tfill's, caution should be exercised in assuming that stand structure has reached a steady state with respect to its disturbance history. The effects of short past fluctuations can exert an influence different from the overall trend for a substantial time after they are over. If the stand really has come to steady state, then either estimate will do, but if not, then both are likely to be in error.  81 5. Any time that gap creation rate is episodic, TT1 estimates should be suspect unless they have been measured over a long period relative to the pattern of mortality fluctuation, or the history of mortality is known in some detail. The problem with these prescriptions, is that it is difficult to assess whether a stand is at steady state or how episodic mortality has been. Certainly we would like to have such data, but in many cases we are many person-years of research away from it. In the mean time, we would still like to estimate turnover time. In this case, it seems most reasonable to obtain both estimates whenever possible, recognizing that either or both may be in error. There is no a priori reason to say that one of the estimates of turnover time is the "true" turnover time. They both measure the turnover of space, and each has its problems in estimation. What do the simulations indicate about the discrepancy between TT1 and TT2 seen in the empirical examples of Table 3.1 (TT2 almost always greater than TT1)? In the model output, the only time when TT2 was consistently higher than TT1 was early in stand development when the proportion of the forest in gap had not reached steady state. This suggests that many of the stands in which both turnover time estimates have been calculated have not reached steady state. If so, the TT1 's may be closer to the steady state value than the TT2's. As well, it suggests that the "appearance" of steady state (based on a researcher's intuition) is a poor indicator of a stand's true status. Many old forests are still undergoing substantial compositional change (Stephens and Waggoner 1980; Harcombe and Marks 1983; Whitney 1984; Glitzenstein et al. 1986; Steijlen and Zackrisson 1987; Foster 1988; Franklin and DeBell 1988). Given differences in canopy geometry, mode of mortality, and longevity among species (Putz et al. 1983; Spies and Franklin 1989b), it is not surprising that they would be far from structural equilibrium as well. There are several ways in which the model oversimplifies stand dynamics. The amount of gap area in a stand is not greatest early in its mature stage (as shown in the middle graph, Figure 3.2). Gap area probably increases more gradually over a longer  82 period of time. This should improve the quality of TT2 estimates. However, the model also neglects changes in canopy structure as a consequence of changes in species composition with stand development (Stewart 1986b; Aplet et al. 1988; Spies and Franklin 1989b). For instance, the late serai replacement of Douglas-fir by western hemlock in the Pacific northwest, or the loss and reinitiation of Englemann spruce (Picea englemanniO in the spruce-fir forests of the Rocky Mountains should greatly prolong the period during which turnover time estimates do not reflect their steady state values. It is likely that the model results are conservative in their assessment of the period during which turnover time estimate are far from their steady state values. Finally, it is not clear that it is the steady state turnover time that is of interest after all. If most forests are not at steady state, or if they spend most of their history in a pre-steady state status, the steady state turnover time may not be that meaningful. A more useful approach might be to assume that turnover time estimates are equivalent to "instantaneous velocities" until proven otherwise. The instantaneous rate of change, together with a sense of the system's history and its likely future, might be combined to provide a longer term estimate of turnover. Short-term observations alone appear inadequate as a basis for conclusions about long-term changes.  83  Table 3.1 Parameters and turnover times from several tropical and temperate forests.  Location  Tfill (years)  %GAP  TT2 (years)  TT1 (years)  Source  Tropical Forests Panama Panama Costa Rica Costa Rica Mexico Mexico Malaysia  10a 10a  -  10b 25-35 25-35C 20-30  2.8 3.8  -  4.8 16.5 42 9.9  360 263  -  211 152-212 60-83 250-375  114 137 94 118 138 47 -  Brokaw 1982a Lang & Knight 1983 Lawton & Putz 1988 Hartshorn 1978 Bongers et al. 1988 M.-Ramos etal. 1988 Poore 1968  123 278 -  Runkle 1982 Romme & Martin 1982 Fosters Reiners 1986 Lertzman 1989 Veblen 1985 Veblen 1985 Armesto & Fuentes 1988 Naka 1982  Temperate Forests E. USA SE. USA NE. USA W. Canada Chile Chile Chile Japan  22.9d 10,20e 100 125 32 52 1009 50h  12.2 6.5,7.7 33 19 8.6 3.3-6.6 29 20  204 156,260 303 658 392 633-794f 345 250  -  -  179  a The authors calculate only TT1,1 have assumed a Tfj|| of 10 years and calculated TT2's (see footnote 2). Data from Lang and Knight (1983) are for old forest, those from Brokaw (1982a) are for younger forest. b Hartshorn (1978) only calculates TT1, but gives mean % gap. 10 years seems a reasonable minimum estimate for Tfj|| (and much less than for the other tropical studies which do estimate Tfj||. To get a TT2 close to TT1, a Tfj|| of 5-6 years is required. Thus this is a conservative estimate for Tfj||, in that it minimizes the difference between TT1 and TT2. c Martinez-Ramos et al. (1988) do not calculate TT2. Values for Tfjn are from Bongers et al. (1988); the studies were located within 1 km of each other. Mean values for MartinezRamos et al. (1988) and for Bongers et al. (1988) data in Table 3.2. d These values are the average obtained by recalculation from Runkle (1982). e TT2 calculated using 2 estimates of Tfill, 10 and 20 years. The TT1 given is estimated over the entire 8 year observation period, both lower and higher estimates were obtained with various subsets of this period. Estimates with Tfj||=20 years used in Table 3.2. f Though Veblen (1985) gives TT2 for both expanded and canopy gaps, I have presented only those from canopy gaps here. g The authors do not calculate a turnover time, I have calculated this from their data. h No Tfjn given, but gaps dating from a typhoon nearly 50 years earlier suggest Tfj|| > 50. Thus the TT2 calculated here is less than it would be if the true Tfj|| were known.  84  Table 3.2 Summary statistics for Turnover parameters.  Mean  Std. Dev.  N  136.4  63.9  9  TTItrop  108.0  34.1  6  TTItemp  193.3  78.5  3  TT2total  323.3  174.2  14  TT2 rop  233.4  102.6  6  TT2temp  390.7  191.7  8  %GAP tal  15.4  12.0  14  a  13.3  15.0  6  %GAP emp  17.0  10.0  8  7  1  1  total  t  to  %GAP rop t  t  a Without Martinez-Ramos et al. (1988): mean=7.5, Standard deviation=5.7, N=5.  85  Table 3.3 Steady state turnover times for simulations with varying input parameters.  Model  TT1  TT2  %Gap  Corresponding Figure  "Temperate": MR=0.00606 Trmin=100 Tfill=25 50 100 150  289 315 370 411  287 309 365 415  8.7 16.2 27.5 24.0  Figure 3.3  "Tropical": MR=0.01212 Trmin=15 Tfill=5 Tfill=10 Tfill=20  102 107 117  100 106 117  5.0 9.4 17.0  Figure 3.6  86  Figure 3.1 Plot of equation 3.2: TT2.  87 Figure 3.2 Output from baseline simulation: Tfill=50, Trmin=100, MR=0.00606. a. The number of trees older than (Tfill+Trmin), from which mortality is calculated, resulting in the TT1 estimate, b. The percent of forest area in gap. TT2 is based on this value, c. TT1 (thin line) and TT2 (thick line).  co  8000  UJ CC  6000  £  4000  Q  2000  °  0  1000 LU r—  QC LD > o z CC ID 100  200  300  400 YEARS  500  600  700  800  88  ure 3.3 Change in TT1 and TT2 overtime with varying durations of Tfill. Tfill=25 years, 50 years, 100 years, and 150 years. In ail cases Trmin=100 years and MR=0.00606. TT1: thin line, and TT2: thick line. In this and the following Figures, any time when a TT2 value reaches the maximum value on the vertical scale, it has actually exceeded that value and has been reseated for graphical clarity.  00 LU  cr  LU > O  z cr z>  200  400  600  YEARS  800  1000  1200  89 Figure 3.4 Change in TT1 and TT2 overtime with varying durations of Tfill, fast turnover time scenarios. Tfill=5 years, 10 years, and 20 years. Trmin=20 years and Mr=0.01212 in all cases. See comments for Figure 3.3.  CO LU  CC LU >  O  z cr  200  90 Figure,3.5 Change in TT1 and TT2 overtime, showing true variation in TT1. Tfill=50 years, Trmin=100 years, MR=0.00606. See_ comments for Figure 3.3.  1000  I  0  ,  I  200  ,  I  400 YEARS  ,  I  600  I  I  800  91 Figure 3.6 Change in TT1 and TT2 overtime with two 20 year fluctuations in mortality rate. From 400 to 420 years MR was doubled, ana from 800 to 820 years MR was halved. Dotted line shows steady state turnover time for both estimates.  YEARS  92 4. PATTERNS OF GAP-PHASE REPLACEMENT IN A SUB-ALPINE, OLD GROWTH FOREST.  4.1 Introduction. Two classes of hypotheses about the species of new recruits filling gaps have been suggested: those relating gapfillers to the species of the gapmaker they replace, and those relating gapfillers to variable characteristics of the gap environment. In a gap-regenerating forest, the sum of the individual transitions of space from gapmakers to gapfillers determines whether or not there is a net change in species composition overtime. The patterns of replacement among gapmakers can thus be used to assess whether, in sum, the current regime of disturbance to the canopy is sufficient to maintain the forest community or whether periodic larger scale disturbances are neccessary (Harcombe and Marks 1978; Runkle 1981; Veblen 1986). Variation among and within gaps, irrespective of the species creating them, may provide opportunities for species which could not establish either under a closed canopy or a subset of the gap environments. Such variation has been suggested as a prime mechanism promoting coexistence in a variety of forests (Denslow 1980, 1987; Orians 1982; Canham 1989; Poulson and Piatt 1989; Whitmore 1989). Alternatively, if most of the successful individuals in gaps were present before the gapmakers died, the role of gaps may be more distinct at a population than a community level, and their effect more one of reorganization than colonization (Glitzenstein et al. 1986). In this chapter, I will examine the relationships of gapfiller species to gapmaker species and gap environment. The overall question is: will the processes occurring in gaps maintain the current canopy composition or change it? Community composition could be maintained by either specific non-random gapmaker-gapfiller relationships (i.e. reciprocal replacement; Fox 1977), or reversals of competitive asymmetries in particular parts of gaps, or gaps of different sizes (Denslow 1987).  93 I will focus this analysis on the following (mostly non-mutually exclusive) hypotheses:  Hypotheses about the identity of gapfillers: gapmaker-gapfiller comparisons. 1. Self-replacement by each species will maintain the current canopy composition. 2. Reciprocal-replacement of canopy dominants will maintain current canopy composition. Reciprocal replacement has been proposed as a mechanism of long-term co-existence in several temperate forests dominated by 2 species (Fox 1977; Woods 1979; 1984; Runkle 1981). 3. Replacement of most gapmakers, irrespective of species, by one species of gapfiller will result in succession: a directional change in species composition overtime (Druryand Nisbet 1973; Connell and Slatyer1977; Horn 1981). 4. Random transitions between gapmakers and gapfillers will result in a slowly drifting community composition (Chesson and Warner 1981; Chesson 1986; Hubbell 1979; Hubbell and Foster 1986a).  Hypotheses about the identity of gapfillers: gap-envrironment - gapfiller interactions. 5. Certain species will be most successful in gaps of different sizes (Denslow 1980, 1987; Bazzaz 1983; Brokaw 1985a,b; Hubbell and Foster 1986b; Whitmore 1989). 6. Due to an abundance of local seed sources, each species will be most successful in gaps where it is most abundant in the surrounding canopy (Sousa 1984). 7. Certain species will be most successful as gapfillers in particular locations within gaps (i.e. center vs. periphery; Orians 1982; Brokaw 1985a; Putz 1983; Beatty and Stone 1986). 8. Certain species will be most successful on particular substrates within gaps (Christy and Mack 1984; Putz 1983; Lawton and Putz 1988).  94 4.2 Methods.  General gapfiller data and definitions. The data discussed in this chapter were collected using the methods described in Chapter 2. They are the detailed descriptions of the gaps which were located along belt transects in stands HSC1, HSC2, NST1, and HLY1. As described in chapter 2,1 collected data on a total of 60 gaps; 23 in 1985 and 1986 (for which I do not have canopy gap areas or gap aperture), and 37 in 1987 (for which I have the full data set). Table 4.1 shows the number of gaps and definitive gapfillers (see below for definition of definitive) from each stand. The four stands can be divided into two without substantial components of mountain hemlock (HSC1 and NST1), and two with mountain hemlock well represented among canopy trees (HSC2 and HLY1). Though there are more gaps in the sample from the first category, there are equal numbers of definitive gapfillers from each. In this chapter, data from the four stands were combined and dealt with as a single data set for the forest as a whole. I divided gapfillers into two categories in the field: the general population of gapfillers (all trees greater than 1 m and less than 10 m in height), and definitive gapfillers. Definitive gapfillers are those which, by virtue of their height, growth rate and location, are those individuals deemed most likely to successfully occupy the open space in a gap. There were usually one to a few individuals which were substantially taller than the rest of the gapfillers, were growing faster (for silver firs), and could be classed as definitive with little fear of error. Where two or more gapfillers could not be distinguished in terms of dominance, I considered all definitive. Definitive gapfillers could be in the canopy gap (the central portion of the gap, defined by the vertical projection of the opening in the canopy onto the forest floor) or in the expanded gap (the peripheral portion of the gap, underneath the foliage of the trees whose canopies defined the edge of the canopy gap). However, I did not consider gapfillers as definitive which were on the outer edge of the expanded gap and were  95 therefore truly marginal to the gap, having little chance of growing foliage which would occupy part of the canopy gap. Some gaps had no definitive gapfillers. This could occur in gaps that were recently created or in very small gaps that are filled by lateral growth of adjacent canopy trees (Runkle 1981; Runkle and Yetter 1987). All gaps in this study which lacked definitive gapfillers were being filled laterally by canopy trees.  Constructing the species-specific transition matrix. To establish gapmaker-gapfiller pairs, I noted the gapmaker with which each gapfiller was associated. In most cases, this was the gapmaker whose stump was closest to the gapfiller. In multiple-gapmaker gaps, however, there were gapfillers which could not be associated exclusively with one gapmaker. For all cases where gapfillers appeared associated with more than one gapmaker, I recorded all gapmakers with which the gapfiller appeared associated. I recorded the species of each gapmaker which could be identified in the field. The longer the time since the gapmaker died, the less likely I was to be able to identify it (Chapter 2). Many of the hemlock stumps could not be identified to species in the field, and it is not possible to separate mountain from western hemlock based on microscopic wood anatomy (Panshin and de Zeeuw 1980). To construct the matrix of species-specific transition probabilities, I tallied the number of gapmakers of each species that were replaced by definitive gapfillers of each species. I represented each definitive gapfiller by a transition of the form: (gapmaker species) --> (gapfiller species). Definitive gapfillers which were associated with more than 1 gapmaker of different species were given a fractional transition for each gapmaker species. For instance, if a Pacific silver fir was associated as a definitive gapfiller with two gapmakers, a western hemlock and a Pacific silver fir, then I tallied two transitions: 0.5(WHEM->PSF) and 0.5(PSF->PSF). After tallying all definitive gapfillers in this way, I summed the transitions for each species pair.  96 I treated the transitions that involved hemlock gapmakers only identified to genus (HEM; Chapter 2) separately. For gaps where 100% of the hemlocks in the surrounding canopy were of one species, I assumed that gapmakers recorded as generic hemlocks were of that species. Similarly, where the proportions of the two hemlock species were equal, and proportional transitions could be allocated equally within a gap, generic hemlock gapmakers were allocated species in proportion to their abundance among the canopy trees surrounding a gap (biasing the results toward a conclusion of no change). This left 14 "unit" transitions: 2 HEM~>WHEM transitions, 1 HEM-->AYC transition, 1 HEM->MHEM transition, and 10 HEM->PSF transitions. Since I recorded no other MHEM->WHEM transitions, but many WHEM->WHEM's, I assumed the 2 HEM~>WHEM's were WHEM->WHEM's. There were no precedents for transitions from either hemlock species to Alaska yellow-cedar, so I assumed that the single HEM->AYC was a MHEM->AYC on the basis of edaphic factors. Because there were no examples of MHEM->WHEM, I made the HEM->MHEM a MHEM->MHEM. Finally, the remaining 10 HEM->PSF were allocated among the 2 hemlock species in the overall proportions in which they occurred in the canopy surrounding the gaps in which the transitions occurred (11 WHEM:16 MHEM). A total of 56 of 118 transitions (56 of 75 transitions involving hemlock gapmakers) involved extrapolation of hemlocks from the genus to the species level.  Gapfiller species and gap characteristics. I presented size frequency distributions of canopy and expanded gaps in Chapter 2. These represent the same gaps that are dealt with in this chapter. For each gap in the sample, I counted the number of canopy trees of each species among the individuals whose canopies defined the canopy gap. These data are the estimate of local canopy composition for each gap.  97 For each gapfiller, I recorded whether it was in the canopy or expanded gap, and its rooting substrate. To estimate the proportion of each gap represented by stumps, I censused line transects along the 2 major axes of gaps, recording substrate class at 1 m intervals.  4.3 Results.  Composition of the overall gapfiller population. Pacific silver fir dominated the overall population of gapfillers with 75% of the individuals (Figure 4.1). This represents an increase of 31.9% in comparison with the canopy (see Chapter 2). Western hemlock, mountain hemlock and yellow-cedar decreased among general gapfillers 15.5,11.4, and 5.2% respectively, relative to their presence in the canopy. Pacific silver fir loses slightly and mountain hemlock and yellow-cedar gain slightly in the transition from the general gapfiller population to the definitive gapfillers (Figure 4.1). The proportion of western hemlock remains almost the same when the general gapfiller population is compared with the definitive gapfillers . The difference between the two distributions in Figure 4.1 is statistically significant (X2; p<0.001), with the major contributions to the estimated X2 value coming from mountain hemlock and yellow-cedar.  Gapmaker-gapfiller transition matrix: Hypotheses 1-4. Table 4.2 presents the species-specific transition frequencies based on gapmaker-gapfiller comparisons. The gapfiller species were not distributed randomly with respect to gapmaker species (X2; p<0.001). All species were replaced by Pacific silver fir with a greater frequency than they replace themselves. Thus hypothesis 1, self replacement, can be rejected for all species but Pacific silver fir, and hypothesis 4, random replacement, can be rejected for the community as a whole. There were few transitions between the low elevation species (western hemlock) and the 2 high elevation species (mountain hemlock and yellow-cedar). Western  98 hemlock never replaced mountain hemlock or yellow-cedar, and was rarely replaced by them, suggesting that there is edaphic or topographic differentiation between these species on a local scale. In fact, definitive gapfillers of mountain hemlock and yellowcedar were distributed non-randomly among the stands sampled: stand HSC2 was represented by only 13 out of the 60 gaps sampled, but contained 12 of 20 mountain hemlock definitive gapfillers and 7 of 10 yellow-cedar definitive gapfillers. The areas in the other stands where yellow-cedar or mountain hemlock definitive gapfillers were found were small patches similar to HSC2 in forest openness or stature: either flat benches or slopes adjacent to stream courses (both of which consistently retained snowpack later in the spring than the surrounding forest). Such differentiation in gapfiller species composition among stands is consistent with variation in their canopy composition (Chapter 2). Sample sizes for estimating transitions involving the two high elevation species are small, especially for yellow-cedar, and observations based on them must be tentative. However, just based on their relative abundances among gapfillers, there should have been more transitions between western hemlock and the two higher elevation species. For instance, based on its abundance as a definitive gapfiller alone (25 out of 112 = 0.22), western hemlock should have replaced mountain hemlock (0.22x13)=2.9 times (Table 4.2), rather than 0 times. Similarly, mountain hemlocks should have replaced western hemlocks 5.57 times rather than once. The four cells involving transitions between Pacific silver fir and western hemlock represent 80% of the observed transitions. When transitions between these two species are tested in isolation, the null hypothesis of no interaction between gapmaker and gapfiller species cannot be rejected (*2; p<0.190). Most gapmakers are replaced by silver fir, irrespective of species. Pacific silver firs represent 72% of the definitive gapfillers (65/90) in this sub-matrix, and they replace gapmakers in a proportion that doesn't deviate from this significantly. Because gapmaker and gapfiller species are independent for mutual transitions between Pacific silver firs and western hemlocks,  99 hypothesis 2, reciprocal replacement, is not supported. Within this sub-matrix, hypothesis 4, random transitions is supported. Of the four gapfiller-gapmaker hypotheses presented in the introduction, the only one supported strongly by these data is hypothesis 3, succession: no species except Pacific silver fir shows strong trends towards self-replacement, and all the others are replaced predeominantly by firs. However, this is not related to interactions between gapmaker and gapfiller species, but solely to the numerical dominance of firs among the gapfillers.  Gapfiller species and gap size. The number of pacific silver fir and western hemlock gapfillers both increased log-linearly with gap size, but over the range of measured gap sizes there was little variation in the proportion of Pacific silver fir among the gapfillers (Figure 4.2). The densities of Pacific silver fir and western hemlock were also approximately constant over the range of gap sizes (Figure 4.3a). Apparently, over most of the range of gap sizes, large gaps do not differ qualitatively from small ones for firs or western hemlocks, they are just larger. There were two differences between firs and western hemlocks. First, in gaps of all sizes, there are usually more firs present, and they are in greater densities than western hemlocks. Second, in very small gaps (expanded gap area < 100 m2), western hemlocks were rarely present (lower x intercept for firs in Figure 4.2, and figure 4.3). For mountain hemlock and yellow-cedar, neither numbers of gapfillers (Figure 4.4) or gapfiller density (Figure 4.3b) show a consistent relationship with gap size. In both cases zero values dominate the data. When gaps with no yellow-cedar or mountain hemlock gapfillers present are excluded, the coefficients of determination (r2) in Figures 4.3b and 4.4 increase, but the regression remains not significant (see below). The data in Figures 4.2, 4.3, and 4.4 show gapfiller relationships with expanded gap area. I estimated linear regressions for total gapfiller numbers and density, and for  100 each species individually on expanded gap area, canopy gap area and gap aperture (see Chapter 2 for explanation of gap aperture). No regressions of any of the gapfiller densities on any of the three independent variates were significant, and all r2's were less than 0.1. In the three cases where regressions of the number of gapfillers on gap area were significant (total number of gapfillers of all species, number of Pacific silver fir, number of western hemlock; all p<0.001), the regressions on expanded gap area had higher r2's than those on either canopy gap area or gap aperture (e.g. for total gapfillers: expanded gap r2=o.8; canopy gap r2=0.69; gap aperture r2=.50). Therefore, the figures shown plot gapfillers versus expanded rather canopy gap area. Can we conclude from the lack of significant correlations between gapfiller numbers (mountain hemlock and yellow-cedar), and density (all species), with gap area, that no relationship exists? This is equivalent to accepting the null hypothesis, and requires knowledge of the statistical power of the tests (Toft and Shea 1983; Cohen 1987). What is the likelihood, if there had been a relationship, that I could have detected it? Statistical power is a function of sample size, the significance level, and the effect size. In each of the cases, the sample size was 36-38. The effect size is r, the correlation coefficient. Assuming a 2-tailed test with alpha = 0.05, we can find the effect size neccessary to achieve acceptable power from Table 3.3.5 in Cohen (1987). Using 0.8 as a lower limit for acceptable power, I could not confidently expect to detect a relationship unless its correlation coefficient was at least about 0.45. In the above cases, the correlation coefficients were substantially less than this. We can conclude, therefore, that it is very unlikely that strong relationships (r > 0.45) exist between gapfiller densities and gap area, but no inference can be made about the presence of weak relationships. The relationships between numbers of definitive gapfillers and gap area are similar to those for gapfillers in general, but more variable (Figures 4.5 and 4.6). The numbers of firs and western hemlocks increase with gap size, but much less of the variation in definitive gapfiller numbers is explained (r2=0.166 and 0.055, and p=0.001  101 and 0.07, for firs and western hemlock respectively). The only difference in trend between gapfillers in general and definitive gapfillers, is that while firs represented a high proportion of the general gapfillers in very small gaps (Figure 4.2), they represented a smaller proportion of the definitive gapfillers in those gaps. Very small gaps tended to be filled by the lateral growth of adjacent canopy trees (4 out of 15 gaps with expanded gaps less than 100 m2 in area as opposed to 7 of 60 gaps overall). The numbers of yellow-cedar and mountain hemlock definitive gapfillers show no relationship with gap size. Another way to pose the question of whether species composition changes with gap size is to examine the distribution of the areas of gaps containing at least one individual of a given species. Figure 4.7 presents boxplots (Chambers et al. 1983; Wilkinson 1988a,b) of gap area for gaps containing definitive gapfillers of each species, and those being filled by lateral growth of adjacent canopy trees (lat). The distributions of canopy and expanded gap area for gaps containing Pacific silver fir and western hemlock are almost identical. The other three distributions (mountain hemlock, yellowcedar, gaps filled by lateral growth of canopy trees) for canopy gap area do not have the long tails at large gap areas, and for yellow-cedar and those filled laterally, the median gap size is smaller. However, for the distributions of canopy gap areas, the 95% confidence intervals for all medians overlap. Considering expanded gap areas does not change the conclusion, though there is more variability in the shape of the distributions. Yellow-cedar was present almost exclusively in small gaps, and only small gaps are filled laterally. However, other species were present in the small gaps as well. Though there appear to be some differences among species in the size of gaps in which they were successful, and larger gaps have more gapfillers in them, hypothesis 5, coexistence promoted by variability in gap size, requires a reversal of the dominance of Pacific silver fir at some gap sizes. Because, there is no suggestion of this in any of these data, hypothesis 5 is not supported.  102  Gapfiller species and local canopy composition. There was little relationship between the abundance of either Pacific silver fir or western hemlock in the canopy surrounding gaps, and their representation among definitive gapfillers (linear regressions on arcsin square root transformed proportions: firs: r2=0.1, p=0.06; western hemlocks: r2=0.07, p=0.12; Figure 4.8). Yellow-cedar was represented by too few definitive gapfillers for this analysis. However, mountain hemlocks increased in frequency among definitive gapfillers with increasing representation in the canopy (r2 of 0.50; p<0.001). We should be cautious in interpreting this correlation for mountain hemlock as causally related to the canopy composition surrounding the gap. Almost all gaps with a substantial proportion of mountain hemlock among definitive gapfillers were in stand HSC2, where mountain hemlock and yellow-cedar reach their greatest abundance. With these data, the effect of abundant local seed sources cannot be separated from potential microclimate or edaphic differences among stands. This does suggest that current differences between stands are self-maintaining. Following the reasoning above, we can assess the likelihood that the lack of relationship for firs and western hemlocks is meaningful by examining statistical power. Again, the sample sizes are 36-38, alpha = 0.05, and 0.8 is the lower limit for acceptable power, but this time the test is one tailed: the proportion of a species among gapfillers is expected to increase with its representation in the surrounding canopy a priori. There is a very high probability (beta < 0.01), that if as great a relationship existed for these two species as was seen for mountain hemlock (effect size = r = 0.7), that it would have been detected (Table 3.3.2; Cohen 1987). I could not confidently expect to detect relationships with effect size less than about 0.4. Thus for these two species, if there is an effect of local canopy composition, it is weak. In general, hypothesis 6, the effect of local seed sources, is supported weakly at best. Mountain hemlock is the only species which appears to show a strong effect of  103 local canopy composition on gapfiller species, and true frequency dependence cannot be separated from other factors.  Gapfiller species and location in gap. Though more than twice as many definitive gapfillers were located in canopy gap than in expanded gap, the proportions represented by each species were almost identical between the two populations {& n.s., p=0.84 - this test should have been able to detect a fairly small effect size of 0.25 with power = 0.8 (Table 3.15 Cohen 1987); Figure 4.9). The greater number of definitive gapfillers in canopy gaps is even more striking when the total area available of each class of gap is considered. The total area of all canopy gaps measured was 2,842 m2. The total area for their corresponding expanded gaps was 9,226 m2 - 3.2 times as much expanded gap area as canopy gap area (for n=37 gaps). In these gaps, there were 128 definitive gapfillers; 88 in canopy gap and 40 in expanded gap. This is many more in the canopy gap than would be expected on the basis of available area alone (X2; p<0.001). The abundance of definitive gapfillers in canopy gap is not explained by a greater amount of woody substrate in canopy gap relative to expanded gap. Based on the transects sampling the abundance of different substrates in the forest (Chapter 2), there was no association between woody substrates (stumps, logs, root throw mounds) and either gap class. Woody substrates were distributed among canopy and expanded gap in the same proportions in which the classes of gap occur in the forest (672 m of expanded gap, 342 m of canopy gap; 204 m of woody substrate in expanded gap, 99 m of woody substrate in canopy gap; X2\ 0.5<p<0.75). Further, most of the definitive gapfillers were Pacific silver fir, which were not substrate limited (see below). Though a disproportionate number of the definitive gapfillers are located in canopy gaps, there is no difference in species composition in the two locations. The center and the periphery of gaps may differ in the levels of critical resources they offer,  104 but they provide no basis for partitioning among the species, and hypothesis 7, partitioning by location in the gap, is not supported.  Gapfiller species and substrate. The different species of definitive gapfillers varied substantially in their distribution among substrates (&•, p<0.001). Only 23% of the Pacific silver firs, but 96% of the western hemlocks were on stumps. Western hemlocks represented 55% of the population of definitive gapfillers on stumps, but 0% on the forest floor (Figure 4.10). Mountain hemlock and yellow-cedar represented a small proportion of the individuals present on both stumps and the forest floor. In addition to the gapfillers shown in Figure 4.10, there were two firs and one hemlock which had germinated on logs. Though there was considerable variability among gaps in the proportion of gap area covered by woody substrates, there was no relationship between the proportion of western hemlock among either gapfillers or definitive gapfillers in a gap, and the proportion of gap area covered by woody substrate (for gapfillers: N=38, r2=0.011, p=0.52; for definitive gapfillers: N=38, r2=.005, p=0.69). Multiple regression of the proportion of western hemlock definitive gapfillers on the proportion of each gap covered in woody substrate and expanded gap area together had non-significant slopes and r2's approaching 0 (proportions arcsin square root transformed as above). These data support hypothesis 8, that differential success among species on  different substrates could play a role in maintaining some species in the community which would otherwise not recruit. Western hemlocks outnumber Pacific silver firs in the population of definitive gapfillers on stumps, the only circumstance found where firs were not overwhelmingly dominant among gapfillers. All but one of the western hemlocks in the transition matrix discussed above were on stumps.  105 4.4 Discussion.  Hypotheses relating gapfiller species to gapmaker species. Either reciprocal replacement (Fox 1977; Woods 1979,1984) or strong selfreplacement could act in a stabilizing manner on community composition. However, I found no evidence that the species of gapmaker exerts an influence on the species of gapfiller. Instead, over the community as a whole, transitions were skewed in favor of one species, Pacific silver fir. As well, for transitions involving just the two most common species (Pacific silver fir and western hemlock, representing 80% of all transitions), gapfiller species were allocated to gapmakers in proportion to their abundance among the gapfillers. Thus, in most cases, success in capturing space in gaps is proportional to a species' representation in the pool of potential gap colonists. Pacific silver fir is able to dominate this pool by virtue of its ability to persist as a suppressed sapling under a closed canopy and sustain release when a gap is created above it. The main deviations from this pattern do not involve the dominant gapfiller, Pacific silver fir, but rather the lack of mutual transitions between western hemlock and the two less common, high elevation species. The succession model seems well supported, and the processes causing these patterns of transitions should result in change in the forest community. Based on gapmaker-gapfiller transitions alone, Pacific silver fir should have a greater relative importance in the canopy in the future.  Predictions for gapfillers and gap environment at Cypress Provincial Park. Based on the dominance of Pacific silver fir among the sapling population (Chapter 2), we could predict a priori, that unless the gap environments strongly favour the regeneration of another species, the gapfiller population would be dominated by firs. Further, based on their life-histories, we would have expected that all species except firs should do better in larger gaps than small. Because it is present in the understory prior to gap creation, Pacific silver fir should occur in gaps of all sizes, but it should be  106 most successful in smaller gaps where relatively rapid canopy closure restricts colonization by new individuals (Hibbs 1982; Runkle 1982; Runkle and Yetter 1987). The other species, and especially western hemlock, should experience increasing opportunity with increasing gap size. A variety of studies comparing the biology of Pacific silver fir and western hemlock have emphasized the hemlocks greater drought tolerance, faster juvenile growth rate, greater seed production and dispersal, and in general, its greater ability to colonize open space (Thornburgh 1969; Kotar 1972; Long 1976; Grant 1980; Wagner 1980). Additionally, the production of the coarse woody substrates required by western hemlock is clearly associated with gaps. Based on these same arguements, it also seems reasonable to expect Pacific silver fir to be indifferent to location within gaps, but that the other species should be more successful in the center of gaps than in the periphery. If local seed sources limit gap colonization, it should show most distinctly in the less common species, mountain hemlock and yellow-cedar.  Hypotheses relating gapfiller species to gap environment. Pacific silver fir successfully replaces most gapmakers, based primarily on its overall abundance among the gapfillers. Even so, it's ability to dominate the community in the future would be limited if, at certain gap sizes, in certain locations in gaps, or on particular substrates, other species recruited to the canopy consistently (Denslow 1987). However, except for western hemlock recruiting on stumps, none of these appears to be the case. There was no evidence that larger gaps differed qualitatively from small ones for the two dominant species: large gaps contained more trees, but not at higher densities or in different proportions. For the two less common species, even the number of individuals present did not increase with gap size. The predictions regarding how species dominance should change with gap size were not supported except for the presence of Pacific silver fir in gaps of all sizes.  107 The "partitioning by gap size" model has received broad support in a variety of other forests (Denslow 1980, 1987; Whitmore 1978, 1989; Hibbs 1982; Runkle 1981, 1982; Brokaw 1985a,b; Poulson and Piatt 1989). Why not here? Though the four species at Cypress are very different in comparison to each other, they would all be classified as members of the same regeneration guild when compared to the diversity of regeneration niches among tropical trees or the mixed deciduous forests of the eastern United States. The range of variation among life history characters and the distinction among regeneration guilds is much greater in the tropical communities discussed by Denslow (1880; 1987), Bazzaz (1983), and Hubbell and Foster (1986b; and even they found inconclusive evidence for the gap size model). For instance, in Hubbell and Foster's plot in Panama, the 307 species of canopy trees belong to 58 families (Table 5; Hubbell and Foster 1986c), as opposed to two families at Cypress Park. Similarly, in her review of gap-dependent regeneration among a variety of tropical communities Denslow (1980) tabulated 62 species in 34 families. In particular, the Cypress Park system lacks canopy species belonging to either the seed bank or animal dispersed guilds, both of which are common among tropical gap colonists (Hubbell and Foster 1986b; Denslow 1987; Schupp et al. 1989). However, the low incidence of windthrow, and consequently, exposed mineral soil, might limit the effectiveness of buried seeds as a gap colonization strategy at Cypress Park. There are a variety of animal dispersed herbs and shrubs (e.g. Vaccinium spp.) that effectively colonize gaps and are uncommon under a closed canopy, so in principle such a strategy could work, but there are no canopy species that exploit it. In general, Cypress Park lacks tree species which can exploit the short-lived, fast-growing, ruderal niche (such as is filled by red alder (Alnus rubra) at lower elevations). In addition to the "limited variety of regeneration niches" argument, the range of variation in gap sizes and gap-formation processes may be too small for reversal of the competitive asymmetry between Pacific silver fir advanced regeneration and new western hemlock recruits. Wagner (1980) found that in an unburned clear cut in the  108 North Cascades of Washington State (970-1,100 m elevation), western hemlock increasingly overtopped Pacific silver fir saplings with greater distance from the stand edge, whereas silver fir growth rate and dominance peaked near the stand edge. Western hemlock reached its greatest growth rate > 50 m from the forest edge, where it would have experienced microclimates substantially more "open" than in the largest of the gaps I studied. The clear cuts near the stands at Cypress were free of snow up to a month earlier than any of the forest environments, suggesting that at larger gap sizes, microclimate would shift in favor of western hemlock. Though gap sizes at Cypress Park are comparable to those in many other gap regenerating forests (Chapter 2), large gaps there are the result of several gap creation events separated widely in time (Chapter 2). This would give the advantage to a species relying on a pool of suppressed saplings because light levels would increase only gradually to those associated with the current large gap. The most common type of gap formation process at Cypress is the slow mortality of standing dead trees (Chapter 2), which would further contribute to this. Thus the lack of an important effect of gap size on competitive dominance can be attributed to 1) little variation in regeneration niches compared to systems where gap size is an important driving variable, 2) large gaps not being large enough for western hemlock dominance, and 3) large gaps being the result of several small gaps occurring contiguously over a long period, rather than the simultaneous mortality of enough trees to cause a sudden, major alteration of microclimate and light environment. The cases where mountain hemlock and yellow-cedar were successful were mostly on flat benches that form cold air drainages and maintain late spring snowpack, such as stand HSC2. This provides circumstantial evidence for the role of within-stand topographic variation in maintaining these species within forests at the lower range of their distributions. However, such conclusions must be tentative because of the small number of transitions for these species, especially for yellow-cedar.  109 Though definitive gapfillers were much more frequently found in the center than the periphery of gaps, the location in expanded or canopy gap did not influence the species composition among the gapfillers. Similarly, local canopy composition did not influence gapfiller species for the two most common species. Mountain hemlock definitive gapfillers tend to be associated with gaps where they have a substantial representation in the canopy, but canopy influence cannot be separated from topographic or edaphic factors in these data. In general, the dominance of Pacific silver firs among gapfillers, and their origin as suppressed saplings whose distribution was established prior to gap creation, obscures the potential impact of local seed sources on gapfiller composition. Because Pacific silver fir can build a pool of suppressed saplings over many decades of seed production, its representation among gapfillers need have little relationship to its abundance among seed sources. The restriction of successful western hemlock regeneration to stumps may have several causes (see discussion in Chapter 2). However, its restriction to a low frequency substrate, combined with the current gap size distribution and mode of gap creation limit its ability to recruit in gaps. Given that the stands at Cypress Park are likely over 1,500 years old (Lertzman and Brubaker, in prep.), the amount of coarse woody debris has probably reached an equilibrium value (Spies et al. 1988). In fact, it seems likely that western hemlock would have had even more restricted opportunities for establishment when the stands were younger and the frequency of stumps and coarse woody debris in in the forest had not yet peaked. If so, the stands may be experiencing greater recruitment of western hemlock now than they did earlier in their history. In general, many of the relationships, or lack thereof, between gapfillers and gap environment are a consequence of the abundance of Pacific silver fir in the understory prior to gap creation. These individuals account for the majority of gapfillers, and their distribution has little to do with the characteristics of gaps that form around them. Similar lack of correlation between gapfillers and gap environment has been found in  110 other forests where the suppressed sapling population is the primary origin of gapfillers (Uhletal. 1988).  Gap-phase mediated coexistence? In these stands, gap-phase processes do not appear to foster coexistence of the tree community. Gaps primarily exert an influence at a population rather than a community level, resulting in a reorganization of the status of pre-existing individuals, rather than facilitating colonization by new species. Gap environments may be necessary for successful recruitment, but are rarely sufficient for any species except Pacific silver fir. Under a changed disturbance regime, (i.e. if the gap size-frequency distribution had a longer upper tail) this might change. Larger scale, or higher intensity disturbances than this stand has experienced in the last several hundred years may be necessary for substantial recruitment of species other than Pacific silver fir. However, under a changed climatic regime, we might see very different patterns than I found. This forest type appears quite sensitive to minor variation in both microand macro-climate. Strong growth responses to the 20th century warming trend have been documented at high elevations in Washington State (Graumlich and Brubaker 1986; Graumlich et al. 1989) and suggested for the Cypress Park community (Chapter 2). As well, invasion of sub-alpine meadows by trees during the early part of this century has been seen regionally (Brink 1959; Franklin et al. 1971). It seems likely that the current dominance of low to mid elevation species among gap recruits is a consequence of 20th century climate, and that during the colder period preceding this century, mountain hemlock and yellow-cedar experienced success in a broader range of sites within the stand than the cold air pockets to which they are now largely restricted. The large number of yellow-cedars among the smaller diameter classes of the canopy supports this idea (Chapter 2). On the one hand then, there is no evidence for strong mechanisms acting today that foster coexistence in this community; on the contrary, all the data indicate a rapid  111 increase in Pacific silver fir at the expense of the other species. On the other hand, we know from the pollen record that there has been no exclusion of species in either the 1,500-2,000 years since the last stand destroying disturbance, or the last 4,500 years of forest history (Lertzman and Brubaker in prep.). An alternative hypothesis to gap-phase mediated coexistence, is that coexistence is fostered by the "storage effect" and climatic variability, combined with long lifespans and slow rates of community change relative to environmental change (Chesson and Warner 1981; Warner and Chesson 1985; Chesson 1986). The conditions for Chesson and Warner's model appear to be met: 1) In the majority of cases, the probability of a species recruiting in gap is proportional to its representation in the pool of available saplings. 2) This proportion almost certainly will fluctuate with changing climate (as evidenced by irregular age structures among canopy trees (Stewart 1986a) and seedlings (Lertzman, unpub.). 3) Climate has fluctuated substantially over the lifetime of these stands (Lamb 1982; Graumlich and Brubaker 1986; Dunwiddie 1986). 4) The species involved are long-lived, with overlapping generations and a substantial capacity for "storage" at many stages of their lives. The next chapter will examine in more detail the potential for long-term coexistence to result from temporal fluctuations in recruitment.  112 Table 4.1 Number of gaps and gapmakers from each stand used to construct the matrix of transition frequencies.  Number of Stand  Number of Gaps  Definitive Gapfillers  HSC1  15  49  NST1  19  64  HSC2  16  77  HLY1  10  36  Total  60  226  113 Table 4.2 Matrix of transition frequencies between species of gapmakers and gapfillers. PSF = Pacific silver fir, W HEM = western hemlock, M HEM = mountain hemlock, AYC = Alaska yellow-cedar. Matrix entries are the proportion of gapmakers of each species replaced by a given species of definitive gapfiller (rows sum to 1).  GAPFILLERS GAPMAKERS  PSF  W HEM  M HEM  AYC  N  PSF  .61  .31  .08  0  48  WHEM  .77  .21  .02  0  47  M HEM  .69  0  .23  .08  13  AYC  .50  0  .25  .25  4  N  76  25  9  2  112  114 re 4.1. Proportion of each species among definitive gapfillers (DGFS) and the overall gapfiller population (GFS). Numbers of individuals making up each proportion are above each column.  cc LU  1.0  1202  148  U.  a <  O LL  O z o p ac O a. o cc a.  0.5  32243  ci  0.0 f W  *****  58  v^ * o C  20  18  10  • •  DGFS GFS  115 re 4.2. Log number of gapfillers of Pacific silver fir (PSF) and western hemlock (WH), arid proportion of Pacific silver fir among gapfillers vs. log expanded gap area, r PSF: 0.74, p< 0.001; r WH: 0.74, p<0.001. Linear regressions and 95% confidence intervals are shown for the number of gapfillers, the line through the proportion of Pacific silver fir is a Distance Weighted Least Squares Smoothing (DWLS; Wilkinson 1988a,b). 2  2  A WH • 10  100 EXPANDED GAP AREA (m2)  1000  PSF  116 Figure 4.3. Log density of gapfillers of Pacific silver fir (PSF) and western hemlock (WH), a., and of Alaska yellow-cedar (AYC) and mountain hemlock (MH), b., on log expanded gap area.  1.000  i  T I I — i "  i T T 11  :—r  """i  1—i—i—r  I T I T T -  • CM  S  a H i ••• •• • • • * S  0.100  A  A  0.010  WH  CO  cc 0.001 LU _J  Li_ Q_ <  •A'  • #A#A  1000  100  10  1.000 r-  -1  I I I I I I  CD LL  O >-  0.100  CO  z  LD Q  o  •  0.010  o  •  o 0.001  10  100  EXPANDED GAP AREA (m2)  1000  o  AY  A  MH  117 Figure 4.4. Log numbers of gapfillers vs. expanded gap area for Alaska yellow-cedar and mountain hemlock. Abbreviations as in Figure 4.3.  1 o  |  1 — i — i — i — i  i  111  1 — i — i — i — ' — i  > i  o CO  4D O  cn  LU  O  O  o  O  A O  CC  O  A  O  5  A AYC  O 10  1000  100 EXPANDED G A P AREA  (m2)  MH  118 Figure 4.5. Log numbers of definitive gapfillers vs. log expanded gap area for Pacific silver fir and western hemlock, and proportion of Pacific silver fir among definitive gapfillers. Abbreviations as in Figure 4.3. r PSF: 0.166, p= 0.001; r WH: 0.055, p=0.07. Linear regressions and 95% confidence intervals are shown for the number of gapfillers, the line through the proportion of Pacific silver fir is a Distance Weighted Least Squares Smoothing (DWLS; Wilkinson 1988a,b). 2  10  2  -  W  s  U. O.  <  a IL Ul  a  u. O c ui m 2 3 Z  A WH • 10  100  1000  EXPANDED GAP AREA (m2) 1.0 r  •  PSF  119 Figure 4.6. Log numbers of definitive gapfillers vs. expanded gap area for Alaska yellow-cedar and mountain hemlock. Abbreviations as in Figure 4.3.  10  i  i •1  r  r  i i i i | '  2 w _i =J  o  LL  A  z  O CO  LL  • 4 Qg> A O O  ®  fe  A.YC O MH A  10  100 E X P A N D E D G A P A R E A (m2)  1000  120 Figure 4.7. Boxplots of the size distribution of gaps containing at least one definitive gapfiller of each species, a: canopy gaps, b: expanded gaps. Abbreviations as in Figure 4.3. The horizontal lines in the middle of the boxes are medians, the horizontal lines marking the box ends are upper and lower quartiles, and the points where the angled sides of the box sides reach the maximum box width are equivalent to 95% confidence intervals around the medians (Chambers et al. 1983; Wilkinson 1988a,b). Sample sizes for expanded gaps are (from left to right): 8, 7, 12, 46, 23. For canopy gaps: 4, 5, 6, 27, 16.  121 Figure 4.8. Percent of each species among definitive gapfillers each gap vs its representation in the canopy surrounding the gap. The only species with a significant linear regression (for both variables arcsin(square root) transformed) is mountain hemlock (r =0.50; p < 0.001 ). 2  122 Figure 4.9. Species composition among definitive gapfillers in canopy and expanded gaps. Sample sizes refer to the number of definitive gapfillers in each location.  123 Figure 4 10. Species composition among definitive gapfillers on the forest floor and on stumps. Sample sizes refer to the number of definitive gapfillers in each location.  N=49  N=72  100  CO  O Q_  o o  hz LU  o  80  60  40  m AYC  20  cc  LU CL  ML  0 STUMP  FOREST FLOOR  B MHEM • WHEN\ • PSF  124 5. MODELLING LONG-TERM FOREST COMMUNITY CHANGE BASED ON GAPPHASE TRANSITIONS.  5.1 Introduction. Gap-phase processes are critical to the community dynamics of many forest communities (Runkle 1981, 1982, 1985; Brokaw 1985a,b; Denslow 1987). Because the gap environment often meets the requirements of species which could not otherwise establish, gaps may play a critical role in maintaining shade intolerant species in the forest community (Connell 1978; Denslow 1987; Whitmore 1989). Alternatively, they may be the site of tree-by-tree successional replacement (Horn 1975,1971; Glitzenstein et al. 1986). The sum of the individual transitions of space from the trees whose mortality creates gaps (gapmakers) to those who replace them (gapfillers) is an estimate of whether the species composition of the forest is changing. Thus, the net result of the gap creation and filling processes will determine the equilibrium status of the forest community. A common approach to projecting the consequences of different patterns of replacement is the use of transition matrix models (Horn 1971, 1975; Runkle 1981; Acevedo 1981; McAuliffe 1988). Each element in such a matrix represents the probability that a canopy tree of species (i) will be replaced by a recruit of species (j). Most applications of such models assume that transitions are a first order Markov process (Kemeny and Snell 1960; Cox and Miller 1965). That is, that the factors determining the transition probabilities are unchanging in time (the matrix is stationary), and that the state at time (t+1) depends only on the state at time (t) and not on any past states. The Markovian criterion is said to pertain in systems where "history" is not important (van Hulst 1979a,b). It is perhaps more correct to say that the sequence of events leading to the current state are important only in terms of how they are reflected in the current state. A common use of such models is to assess the equilibrium status of a community by comparing the predicted equilibrium composition with the current  125 canopy composition (Horn 1971, 1975; Runkle 1981; Veblen 1985; White et al. 1985; Taylor and Zisheng 1988b). Given the Markovian assumptions, the equilibrium composition is determined by the transition probabilities alone, irrespective of the initial state. The conditions which determine species-specific transition probabilities, such as climate, do change overtime scales relevant to the population and community dynamics of long lived organisms such as trees (Davis 1986a; Ritchie 1986; Kullman 1987; Steijlen and Zackrisson 1987). As well, such fluctuations may have profound implications for community composition (Chesson and Warner 1981; Chesson 1986; Chesson and Case 1986). For such systems, Markov projections are more appropriately treated as estimates of the current trajectory of community change than )  as predictors of future equilibrium states. For systems that are unlikely to reach equilibrium, the predicted equilibria are useful primarily as indices of the consequences of a particular set of transition probabilities or the environment that determined them. In its basic form, a Markov model also only specifies the fate of open space once it has been created. It does not incorporate patterns in the creation of open space, such as differential longevity among species. Differential longevity can be an important factor moderating the results of the transition probabilities in some communities, and the effects of longevity need to be added to Markov models of communities containing species with diverse lifespans (Horn 1975; Acevedo 1981; White et al. 1985; Veblen 1986; McAuliffe 1988). In the last chapter, I examined patterns among the transitions between gapmakers and gapfillers in the forest at Cypress Provincial Park and presented a matrix of species-specific transitions. I found that the patterns of transitions from gapmakers to gapfillers indicated a shift in composition in favour of Pacific silver fir, and that variability in the gap environment was not acting strongly to compensate for the firs' dominance. In this chapter, I will examine the longer term consequences of those transitions for forest composition. I will focus on the role of differential longevity among  126 species in moderating the numerical dominance of Pacific silver fir among the gapfillers, and on the possible effects of variations in climate on patterns of community change. I will do this through a series of Markov models that compare differing assumptions about differential longevity (expressed in species specific mortality rates). I will then incorporate a fluctuating climate by varying the empirically determined transition matrix of Chapter 4 in ways that imitate warmer and colder climates.  5.2 Methods and Description of the Models.  General gapfiller data and definitions. The data discussed in this chapter were collected using the methods described in Chapters 2 and 4. They are based on the detailed descriptions of the gaps located along belt transects in stands HSC1, HSC2, NST1, and HLY1. As described in Chapter 2, I collected data on a total of 60 gaps; 23 in 1985 and 1986, and 37 in 1987. Some applications of Markov models have been criticized on the basis of how the transition probabilities were estimated. For instance, basing transition probabilities on comparisons of overstory species composition with understory species composition assumes an equality in growth rates and survival of the understory trees that can result in substantial error (White et al. 1985). Better estimates of species-specific transitions are obtained by comparing the species of gapmakers with the species of those individuals destined to replace them in the gaps (White et al. 1985). The transitions discussed here are based on comparisons of the species of gapmaker with the species of its definitive gapfiller(s). Definitive gapfillers are those trees which, by virtue of their height, growth rate and location, are those individuals deemed most likely to successfully occupy the open space in a gap. The derivation of the individual transitions used in this chapter is presented in the methods for Chapter 4.  127  The Markov Model To build Markov models, I assumed that the probability with which each species replaces each other species is equal to the relative frequency of each gapmakergapfiller transition. I added a demographic component to the models to investigate the effects of differential longevity among species on the time-course of change to equilibrium, and on the equilibrium predicted by the models. I expressed differential longevity as variation in species-specific mortality rates. The model simulates a population of 1,000 individuals. At each time step, species-specific mortality rates are applied to the community: a fixed proportion of the individuals of each species is "killed". These gapmakers are then replaced according to the probabilities in the transition matrix.  Calculating Differential Mortality Rates I examined two cases with equal mortality rates among species (one high and one low), and two cases with differential mortality rates. The mortality rates for Differential Mortality 1 (Table 5.1) were based on the data from an old growth Douglasfir forest in Oregon (Franklin and DeBell 1988), combined with the distribution of species among gapmakers at Cypress Park (Table 5.2). Equal Mortality Rates 1 applied the mean mortality rate from this model to all species, and Equal Mortality Rates 2, the maximum mortality rate (Table 5.1). Differential Mortality 2 used mortality rates based on the ages of canopy trees at Cypress Park. There are no published mortality rates for old growth of the type found at Cypress, and few for any similar systems. I chose the Franklin and DeBell (1988) data set as the best available. To generate mortality rates for the Cypress Park system based on the Franklin and DeBell data, I made several assumptions. First, in their stand, mortality rate generally decreases with tree size, and Pacific silver fir and western hemlock are represented in smaller size classes than they are at Cypress. So, rather than use their species-specific mortality rates, I used the overall mean mortality  128 rate for individuals of all species greater than 37 cm dbh (0.57% per year) as an estimate of annual mortality for the stand. Assuming a steady state distribution of gapmakers among species at Cypress Park, I calculated species-specific mortality rates by assuming that total stand mortality would be allocated among species in the proportions in which they occur among gapmakers at Cypress (Table 5.2). For instance, of 1,000 total trees, 57 die each decade (0.57% annual mortality rate). Pacific silver fir represents 0.64 of the gapmakers at Cypress (Table 5.2), so (0.64x57) = 36 of the gapmakers during that decade would be allocated to Pacific silver fir. Stand HSC2, where mountain hemlock was most common in the canopy, had a higher frequency of unidentifiable gapmakers (Chapter 2), causing mountain hemlock to be underrepresented among gapmakers. To correct for this in the proportions of gapmakers applied to the Franklin and DeBell rate above, I lumped all hemlocks from Table 2.8 (western hemlock, mountain hemlock, and generic "HEM"s) and, given equal mortality rates between the hemlock species, assumed that they were distributed among gapmakers in proportions equal to their canopy representation. Because yellow-cedar decays more slowly than mountain hemlock, and is therefore recognizable for a longer period in the decay process, I did not apply any correction to its representation among gapmakers. Table 5.1 shows the mortality rates obtained through this process. The mortality rates for Differential Mortality 2 were based on the ages of trees found at Cypress Park (Figure 2.3, Chapter 2). For Pacific silver fir and the two hemlock species, I used the mean age of trees greater than the median age as an estimate of longevity, and its inverse as an estimate of the mortality rate of large, gapforming trees. For yellow-cedar,. I used 1/1100 years as the mortality rate. This generated mortality rates substantially lower than those in Differential Mortality Model 1, but with less overall differential between species. I do not assume that these estimated mortality rates accurately represent the true rates at Cypress Park; the data with which to assess actual mortality rates do not  129 exist. Rather, I suggest that these estimates are reasonable guesses based on partial information, and that they are adequate to assess the effect of changes in the overall rate of mortality and how mortality is distributed among species on patterns of relative abundance over time. To characterize each model, I ran it for 500 decades and defined the equilibrium as the proportion of each species after 5,000 years. In all cases, the changes in each species during the last 1,000 years of simulation was far less than 1 %. Equilibrium species abundances could be obtained analytically (Kemeny and Snell 1960; Cox and Miller 1965), but using the simulations allowed me to easily examine changes overtime when the forest is far from equilibrium, which would have been much more difficult to do analytically (Coale 1972). I obtained an estimate of the community disequilibrium by calculating at each time step, the sum over all species of the difference between their current abundance and their equilibrium abundance. Examination of change in this value over time allows an assessment of the rate of change in the community during its approach to equilibrium and of the time it takes to reach equilibrium conditions. Except where otherwise noted, when I refer to % change in a species' abundance, I am referring to its percentage in the community as a whole, not the change in its values expressed as a percentage of that value. Though the equilibrium species composition is independent of initial conditions for a given transition matrix, the time to reach equilibrium will vary greatly with the initial distance from equilibrium. By using the current composition of the Cypress Park forest as initial conditions for all simulations (Table 5.2), I was able to assess the implications of each model for change at Cypress Park and compare the equilibria and time-course of change between models.  Simulations with non-stationary matrix: Changing environmental conditions Climate has not been stable over time scales relevant to the population and community dynamics of long-lived trees, and we cannot assume that tree communities  130 are in equilibrium with climate today (Davis 1981, 1986a; Kullman 1983; Ritchie 1986; Colinvaux 1987; Steijlen and Zackrisson 1987). To examine the potential role of a changing environment in the dynamics of the Cypress Park community, I relaxed the assumption of stationary transition probabilities. I did this by constructing two alternate transition matrices, one for an arbitrary warmer climate, and one for a colder climate (Table 5.3). Both were based on the empirically determined matrix used in the above simulations. My goal was to make changes in the matrix that would reflect fluctuations in climate of approximately the magnitude of the change from the medieval optimum to the little ice age, or the little ice age to the early 20th century warming (Lamb 1982; Kullman 1983,1987; Ritchie 1986; Steijlen and Zackrisson 1987). The rules that I used for generating these matrices were that changes in a given cell should be 1) in the range of 10-20%, and 2) based on patterns observed in the intermediate empirical matrix (such as a low frequency of transitions between the high elevation species and the low elevation species, and a higher frequency of transitions involving Pacific silver fir and all other species). There were a few cases where changes were greater than 20%, such as the proportion of mountain hemlock and yellow-cedar gapmakers replaced by western hemlock in the intermediate matrix compared with the warm climate matrix. The warm climate matrix favors the low elevation species, western hemlock, at the expense of the others, especially mountain hemlock and yellow-cedar. The cold climate matrix favors the high elevation species, mountain hemlock and yellow-cedar, at the expense of the others, especially western hemlock. Pacific silver fir is most favoured under intermediate conditions, but never does badly. As with the estimates of mortality rates, these alternate climate matrices are not intended to reflect particular real climates, but rather are "best guesses" of what transitions might look like under changed climatic conditions in either direction. Because these species have coexisted in varying proportions in this stand under a variety of climates over the last five millennia (Lertzman and Brubaker, in prep), the  131 conditions reflected in these alternative matrices must have existed at some time in the history of the stands. In order to create smoother transitions from one climate state to another, I created two additional matrices with values halfway between the empirical matrix and those at either climatic extreme. In simulating changing environmental conditions, there were thus 5 possible climate states: colder, cold, empirical matrix, warm, warmer. I examined two components of a changing climate: the pattern in which climate states occur, and the duration of each state. I used two patterns of change in climate states; one in which states followed each other in a sinusoidal sequence (e.g. one period would be: neutral, warm, warmer, warm, neutral, cold, colder, cold, neutral), and one in which the climate states were randomly chosen. For each, I simulated cases where the duration of climate states was 50 years, 100 years, 200 years, and 300 years. These give periods for a neutral-warm-neutral-cold cycle from 200 to 1,200 years. Changing climate simulations were run with both Equal Mortality 2 mortality rates and Differential Mortality 1 mortality rates. To generate summary statistics for the random climate simulations, I ran 100 replicates of each scenario (with samples taken every 200 years from 5,000 year time series).  5.3 Results.  Equilibria and time to equilibrium with and without differential mortality With equal mortality rates, the forest experiences substantial change from its initial conditions, reaching an equilibrium where Pacific silver fir is substantially more common than it is now and the other species are less common (Equal MR 1 and 2; Table 5.4, Figure 5.1). Thus, based solely on the replacements currently occurring in gaps, we would conclude the forest is far from equilibrium and undergoing substantial change. Reducing the overall mortality rate, but holding it proportional among species does not change the outcome of succession, but can delay reaching equilibrium  132 (compare equilibrium compositions for Equal MR 1 and Equal MR 2 in Table 5.4, and rates of change in Figure 5.1). Even with equal mortality rates, species vary in the length of time they take to reach their equilibrium proportion, depending on the magnitude of change required. For instance, under Equal MR 2, Pacific silver fir increases 22% from initial conditions, and takes 310 years to approach within 5% of the equilibrium. Western hemlock decreases 9%, and takes 90 years to approach within 5% of its equilibrium value. Time to reach equilibrium defined for the community as a whole appears to be very different from what one might estimate based on examination of individual species. Each species' contribution to overall community disequilibrium will depend on its abundance within the commmunity, its rate of change, and how far it is from its own steady state value. However, mortality rates do differ among species, and the results of Equal MR 1 and 2 are mainly useful as baselines against which to judge the effects of differential mortality. Incorporating differential mortality into the models has two main effects: changing the equilibrium species composition and extending the time taken to reach equilibrium. In both the models with differential mortality among species, the equilibrium species composition is substantially moderated relative to that for the equal mortality models. Each species is intermediate in abundance between its current representation in the canopy and that predicted by the equal mortality rate models. The dominance of Pacific silver fir among definitive gapfillers is partially compensated for by its shorter lifespan (compare Differential Mortality 1 and 2 with Current Canopy, Equal MR 1 and Equal MR 2; Table 5.4). In fact, the initial conditions for all species are within 5% of their equilibrium values predicted by Differential Mortality Model 1. Differential Mortality models 1 and 2 predict different equilibrium compositions from each other because of the. difference in the ratio of mortality rates between species. For instance, the ratio of the mortality rates of Pacific silver fir and yellowcedar for Differential Mortality 1 is 4.95, whereas for Differential Mortality 2 it is 2.02. Because there is less overall differential in mortality rates in Differential Mortality 2, its  133 equilibrium composition is intermediate between that of Differential Mortality 1 and the Equal Mortality Rate models. The lower rate of total stand mortality in Differential Mortality 2, combined with the larger initial difference from its equilibrium composition, means that under this model the approach to equilibrium is quite slow and prolonged: after 1,000 years the community is still almost 10% from its equilibrium composition (Figure 5.1). Again, species take different lengths of time to reach equilibrium. Species with faster turnover of individuals, such as Pacific silver fir, more quickly approach to within a few percent of their equilibrium proportion, whereas the species with the slowest turnover (Yellow cedar) takes substantially longer. Though close to the equilibrium species composition initially, the forest may be hundreds of years to millennia from reaching a state of no change. Under Differential Mortality 1, with higher overall mortality, species vary from 330 to 1,050 years to reach within 1 % of the equilibrium value. Under Differential Mortality 2, with lower overall mortality, species vary from 1665 to 3095 years. The rate of change over the last few percent is so slow that it would likely be undetectable in field data. During this time, the turnover of individuals of the longest lived species may be rate-limiting for the equilibration of the community as a whole. Without knowledge of the actual mortality rates at Cypress Park it is not possible to choose one set of mortality rates over the other to assess definitively the equilibrium status of the forest at Cypress. It is clear though, that the initial assessment that the forest is far from equilibrium is unlikely to be correct. Using reasonable assumptions about differential mortality, it is possible for the current regime of gap-phase replacement to maintain the current canopy composition. For the rest of these simulations, I will use Differential Mortality 1 mortality rates because 1) they are in better accord with the little data that do exist, and 2) the sample of gapmaker species composition at Cypress Park is better than the sample of the age distributions.  134  Equilibria produced by warm climate and cold climate matrices The cold matrix and warm matrix results in Table 5.4 show the equilibrium species composition for Differential Mortality 1 using the two alternative climate matrices (Table 5.3). The equilibrium compositions reflect the intent of the matrices: the cold matrix produces a forest dominated by yellow-cedar, mountain hemlock and Pacific silver fir, and the warm matrix produces a forest dominated by western hemlock and Pacific silver fir with the two high elevation species virtually excluded. In general, times to equilibrium with the alternative matrices are longer than with the empirical matrix, reflecting the difference between the current species composition and the equilibria for either alternative matrix (Figure 5.2). The forest produced by the cold matrix resembles the current species composition at higher elevations in the Mountain Hemlock Zone, where the forest grades into sub-alpine parkland, except that western hemlock is normally excluded in such areas altogether. Although yellow-cedar is the most common species at equilibrium under the cold matrix, it takes an exceptionally long time for yellow-cedar to reach this abundance. Over most of the first 1,000 years there is a more equitable split between Pacific silver fir, mountain hemlock, and yellow-cedar. During much of this time it resembles the current species composition in stand HSC2 fairly closely (Table 2.3).  The effects of changing climate and climate-differential mortality interactions. Figure 5.3 shows the results of simulations with equal mortality rates under a regularly varying climate. The species' fluctuations are always either in phase or 180° out of phase with each other. When the duration of each climate state is long enough (>100 years), yellow-cedar is eliminated from the stand during warm periods. Though  135 species fluctuate in abundance, only with the longest periods is there a reversal in ordinal rank of species abundance, and Pacific silver fir is always the most abundant. In these simulations and those that follow, Pacific silver fir has two periods for every one of the other species. This is because the climate only passes through the optimum for the other species once per period, whereas, since the optimum for Pacific silver fir is in intermediate states, the climate passes through its optimum on both the ascending and descending phases. With equal mortality rates, the height of the two peaks for Pacific silver fir are the same. With differential mortality under a regularly varying climate, there are some important differences, which become increasingly clear when the duration of climate states increases (Figure 5.4). First, the species are not in phase; lags develop which are proportional to the differential in mortality rate. This is most clearly seen for yellowcedar and mountain hemlock where the lagged response of yellow-cedar can cause a reversal in relative abundance twice per period. Second, Pacific silver fir and western hemlock also show reversals of relative dominance. Third, the two peaks of Pacific silver fir abundance per climatic period are unequal in height: silver fir does better when it is growing cooler after a warm period than when it is growing warmer after a cool period. This is likely because the western hemlock population tracks the worsening climate (from its perspective) faster than does yellow-cedar, creating more gaps for the firs to fill during the transient periods when it is most competitive. Finally, yellow-cedar is not eliminated from the community under any of the scenarios examined. Its longevity allows it to persist through unfavourable periods of 900 years duration (warm+warmer+warm; 300 years each), even though for the middle 300 years it does not successfully fill any gaps (Table 5.3). Under random climate sequences, the dynamics were similar, except that the fluctuations were erratic rather than regular. Little can be concluded from one realization of a random process, so the discussion below is based on the results of a large number of such runs.  136 For all these models in a changing environment, the system state observed in any given window in time is transient. Two ways of summarizing the behaviour of the forest community under each set of conditions are the mean abundance of each species overtime and the variability about the mean. The mean abundances of each species are very similar under random and cyclic climate sequences, though the Pacific silver fir does slightly less well, and yellow-cedar slightly better under random climates (Table 5.4). Within a given pattern of climatic fluctuation, mean abundances varied negligibly between simulations with different durations of climate states. For both the cyclic and random climates, the realized mean species abundances differ from what is expected from the mean climate state. The intermediate empirical transition matrix is the average climate for all scenarios, but relative to the equilibrium composition for this matrix (Table 5.4), varying climates result in a decrease in the abundance of silver fir (0.457 to 0.385, cyclic climate mean; Table 5.4) and an increase in the abundance of yellow-cedar (0.029 to 0.107, cyclic climate mean; Table 5.4). While this represents only a 16% change in the fir population, it represents a 279% increase in yellow-cedar abundance (expressed in terms of their own abundances). These differences can be attributed to a combination of the system tracking a constantly changing climate, differential mortality among species, and the interaction between the two. The component due to the system tracking a changing climate alone can be seen in the simulations with changing climate and equal mortality rates. The reduction in Pacific silver fir and increase in yellow-cedar is seen to a lesser extent here. The mean abundances for the 100 year duration simulation with equal mortality are: PSF, 0.592; WHEM 0.275; MHEM, 0.093; AYC, 0.040. These are intermediate between the results from the equal mortality rate simulations without changing climate and the simulations with differential mortality and changing climate (Table 5.4.). Variability increases with increasing duration of climate states under both random and cyclic climate sequences (Figure 5.5): when the system has more time to  137 equilibrate to one climate, it is farther from the equilibrium for another and the magnitude of the oscillations increases. Several patterns among species are apparent in Figure 5.5. Though western hemlock consistently has the highest variability (standard deviation), yellow-cedar has the largest coefficients of variation because it has the smallest mean population size. At low numbers of individuals, the absolute magnitude of fluctuations in population size may be less important than the proportion they represent of the mean population size (see discussion of yellow-cedar below). It is interesting to note that the biggest difference in summary statistics between cyclic climate simulations and random climate simulations is the increase in the coefficients of variation for the two least common species (mountain hemlock and yellow-cedar) under random climates.  5.4 Discussion. These simulations support the idea that differential longevity among species can be a critical component in assessing the equilibrium status of a gap-phase regenerating forest (White et al. 1985; Veblen 1986). In these models, differences in life histories (longevity) and environmental variation both moderate the basic trend in the Markov models of succession to an equilibrium dominated by Pacific silver fir. Further, differential longevity interacts with a fluctuating climate in ways which increase the average abundance of rare, long-lived species at the expense of common, shorter-lived species. This can be seen as a special case of Chesson and Warner's lottery model where environmental fluctuation promotes coexistence through recruitment fluctuations (Chesson and Warner 1981; Warner and Chesson 1985; Chesson 1986). The changes in the transition matrices explicitly represent variation in recruitment probabilities with a changing environment. Species are long-lived, with overlapping generations, and yellow-cedar, especially, benefits from the "storage effect". In particular, the  138 persistence of yellow-cedar in the community during the 300 year duration fluctuations of Figure 5.4, when it is excluded in those of Figure 5.3, is an example of the "storage" by long-lived individuals allowing coexistence.  Long-term dynamics of the Cypress Park system: Assessing Change in the Field Without estimates of the actual mortality rates at Cypress Park, it is difficult to assess with certainty how close the forest is to equilibrium. If the real mortality rates are similar to those in Differential Mortality 1, then the forest is near equilibrium, and the replacement processes occurring today are more-or-less maintaining the current canopy composition. However, the modest change in species proportions necessary to reach equilibrium under Differential Mortality 2 will take a very long time. As well, the simulations suggest that conditions similar to those present today could result from a sequence of changing climates, during which the forest was never in equilibrium. What can be concluded, is that transition probabilities alone are inadequate to assess the compositional stability of the forest. Even without changing climate, processes affecting canopy tenure at long time scales can interact with patterns of recruitment to create different dynamics than would be inferred from short term transitions alone (see White et al. 1985). Adding variation in climate increases the problem of inferring long-term trends from short-term data. The patterns of fluctuation in species abundances shown in Figure 5.4 are readily explained, given knowledge of the model structure and the climatic forcing function. However, if one had only field data similar to Figure 5.4, explaining the patterns of change would be enormously more difficult, especially if it were only from an arbitrarily chosen 10-20 year period. For instance, we might sample a period when mountain hemlock had begun to respond to a favorable climate, but yellow-cedar had not. Similarly, the longer periods of the other species compared to Pacific silver fir are an obvious consequence of the transition matrices and the pattern of climatic fluctuation. But without much more detailed knowledge of species  139 interactions and climate history than is usual, would one be likely to tease out the causal pattern from an empirical time series? Beyond any complexities in the pattern of change, the slow rates of change exhibited in the model output suggest that assessing the stability of forest communities with time series of the lengths usually available may be very difficult. Forest composition may appear constant over the life of a researcher and yet be in a state of substantial, slow change. To summarize the implications of these results for the Cypress Park forest, the numerical dominance of Pacific silver fir among smaller canopy trees, suppressed, saplings, and gapfillers (Chapters 2 and 4) will not necessarily lead to an equilibrium where that species is more abundant than it is today. Similarly, despite its poor representation among the younger age classes, yellow-cedar will not necessarily decrease in abundance, and cannot be expected to be excluded from the stands. The forest is probably close to the equilibrium species composition for the current set of transition probabilities, but cannot be assumed to be unchanging. Though the stand itself is old enough that species composition might have been able to come to equilibrium after the last major fire (Lertzman and Brubaker, in prep.), the current species composition could be a transient state that is a consequence of a sequence of climatic fluctuations, rather than the slow approach to equilibrium under constant conditions.  Applicability of Results to Other Systems Differential longevity among competing tree species is an important factor promoting coexistence in a variety of forests (Acevedo 1981; White et al. 1985; Veblen 1986; Aplet et al. 1988). There are, however, some important differences between the system modelled here and many other forests. Coniferous trees of western North America are longer lived than most, and those at Cypress Park are among the longer lived of the trees in this region (Fowells 1965; Franklin 1988). Time scales of gap  140 creation and filling are proportionately slower than in many other forests (Chapters 2 and 3; Spies and Franklin 1989b). However, the overall differences in longevity among species may not be as unusual as the lifespans themselves. Early successional gap colonists in eastern hardwood and tropical forests can be substantially shorter lived than their shade tolerant associates (Fowells 1965; White et al. 1985; Hubbell and Foster 1986a; Whitmore 1989). The time scales over which the patterns described here are expressed may vary in other systems, but the kind of roles played by differential mortality rates and climatic variability should be similar. The models presented here suggest that many old-growth, gap-regenerating forests are not likely to be in compositional equilibrium, given a realistic assessment of changing climate. In fact, few old-growth forests have been found to be compositionally stable (Stephens and Waggoner 1980; Harcombe and Marks 1983; Lang and Knight 1983; Whitney 1984; Glitzenstein et al. 1986; Steijlen and Zackrisson 1987; Foster 1988; Franklin and DeBell 1988). It seems likely that detailed reconstructions of stand histories combined with palaeoecological and palaeoclimatic data will be required to arrive at any general conclusions about long term forest dynamics (Davis 1981, 1986a; Clark 1986a,b; Delcourt and Delcourt 1987; Bradshaw and Miller 1988; Payette et al. 1989; Lertzman and Brubaker, in prep.). The long time taken by the forest to equilibrate to the "warm climate" transition matrix from current conditions (500 years to reach within 10%, summed over the whole community; Figure 5.2) has interesting implications for the consequences of current trends in global climate. If current estimates of anthropogenic climate change over the next 100 years are at all accurate, then we can anticipate changes in the climatic variables driving change in stands such as at Cypress Park that are greater than any in their recent history (Dickinson and Cicerone 1986; Brown and Flavin 1988; Jones et al. 1988; Ramanathan et al. 1989). In that case, Figure 5.2 represents a conservative estimate of the degree and duration of disequilibrium to be expected. Of course, if the  141 frequency of catastrophic disturbances such as fires changes as well (Clark 1988), then these models will not apply.  Factors Not Included in the Model The models I have presented here are simple, incorporating only the biology implicit in the transition matrices and differential mortality rates. In reality, a variety of other biological and physical processes will be important. The effect of most of them will be to further delay a stable composition. While differential longevity is represented through species-specific mortality rates, age-structure as such is absent from the model, as is the time required for filling a gap (which may be substantial; Chapters 2 and 3). Both of these could increase the possibility of lags between tree populations and climate. Though the survival of adult trees is less sensitive to climatic fluctuations than recruitment (Ritchie 1986), at some level of climatic fluctuation, the mortality rates of canopy trees will be affected. This will increase the ability of tree populations to track fluctuations in climate, but will likely act in a highly non-linear way (such as catastrophic mortality after some threshold is reached). Fire frequency will also be affected by climate (Green 1982; Cwynar 1987; Clark 1988). Fires can act to overcome the inertia caused by persistent individuals established under past climatic conditions. When adults are removed, the new forest may reflect current climate much more accurately than did the old one (Ritchie 1986; Steijlen and Zackrisson 1987; Kullman 1983, 1987; Payette and Gagnon 1985; Payette et al. 1985). Other factors than fires may alter the disturbance regime in ways that will drastically alter the transition probabilities. Periodic insect or pathogen outbreaks could alter competitive asymmetries or the distribution of gap sizes. These models do not account for changes in disturbance regime other than simple fluctuations in climate. They could not, for instance, predict the community response to the removal of a  142 species by pathogens, such as occurred for eastern hemlock (Tsuga canadensis) in the mid-Holocene over much of eastern North America (Allison et al. 1986). Finally, there is no frequency dependence in the transition probabilities in these models (Acevedo 1981). At low densities, populations may not be self maintaining, and recolonization after local extinctions may be limited by seed sources. If I had included these factors, yellow-cedar would not have been able to recover from some of its low periods in the simulations with equal mortality rates and changing climate (Figure 5.3). True local extinctions were not possible in these models, as they are in the real world.  143  Table 5.1. Annual mortality rates for each species in varying models. Differential Mortality 1 uses data from Franklin and DeBell (1988) combined with each species' proportions among gapmakers to determine species-specific mortality rates. Equal Mortality Rate Model 1 uses the maximum of the mortality rates from Differential Mortality 1 for all species, and Equal Mortality Rate Model 2, the mean. Differential Mortality 2 uses mortality rates based on the ages of canopy trees at Cypress Park as explained in the text. See text for further explanation of each model. The inverse of these rates is an estimate of canopy turnover by species.  Pacific Silver Fir  Western Hemlock  Mountain Hemlock  YellowCedar  Equal MR 1  0.00842  0.00842  0.00842  0.00842  Equal MR 2  0.00447  0.00447  0.00447  0.00447  Differential Mortality 1  0.00842  0.00389  0.00388  0.00170  Differential Mortality 2  0.00184  0.00125  0.00125  0.00091  144  Table 5.2. Proportions of each species among primary and secondary gapmakers combined, canopy trees, and definitive gapfillers. Data are modified from Tables 2.2, 2.8 and Figure 4.1.  Pacific Silver Fir  Western Hemlock  Mountain Hemlock  YellowCedar  N  Gapmakers  0.640  0.240  0.100  0.019  147  Canopy  0.432  0.355  0.150  0.063  602  Definitive Gapfillers  0.670  0.195  0.090  0.045  221  145  Table 5.3 Matrices of transitions between species of gapmakers and gapfillers for 3 alternative climate states. "Warmer Climate" Matrix GAPFILLERS GAPMAKERS  PSF  WHEM  MHEM  AYC  PSF WHEM MHEM AYC  .504 .600 .680 .680  .496 .400 .260 .260  0 0 .060 .060  0 0 0 0  Empirical Matrix: Current Conditions GAPFILLERS GAPMAKERS  PSF  WHEM  MHEM  PSF WHEM MHEM AYC  .605 .766 .692 .500  .312 .213 0 0  .083 .021 .231 .250  AYC 0 0 .077 .250  "Colder Climate" Matrix GAPFILLERS GAPMAKERS  PSF  WHEM  MHEM  PSF WHEM MHEM AYC  .504 .670 .492 .300  .212 .110 0 0  .183 .120 .331 .350  AYC .101 .100 .177 .350  146  Table 5.4. Equilibrium species compositions for several versions of the Markov model. See Table 5.1 for explanation of the models. For the first six models, values are the proportion represented by each species after 5,000 years of model time. The "Warm Matrix" and "Cold Matrix" results represent the equilibrium obtained for Differential Mortality 1 using the two "alternative climate" transition matrices shown in Table 5.3. The values for Random and Cyclic climates are the overall mean proportions of each species for simulations with all durations of climate states (i.e. 50,100, 200, 300 years; see Figures 3, 4, and 5). The mean proportions varied negligibly among simulations with different durations of climate states. For the random climate simulations, the data represent samples drawn every 200 years from 100 replicates of 5,000 year model runs.  Pacific Silver Fir  Western Hemlock  Mountain Hemlock  YellowCedar  Current Canopy  0.432  0.355  0.150  0.063  Equal MR 1  0.653  0.259  0.080  0.008  Equal MR 2  0.653  0.259  0.080  0.008  Differential Mortality 1  0.457  0.392  0.122  0.029  Differential Mortality 2  0.558  0.324  0.102  0.016  "Cold" Matrix  0.239  0.124  0.253  0.385  "Warm" Matrix  0.359  0.642  <0.001  <0.001  Cyclic Climate  0.385  0.382  0.126  0.107  0.373  0.382  0.125  0.121  MODEL  Random Climate  *  147 re 5.1 Total community difference from equilibrium composition: models with empirical transition matrix. Data points represent the summed deviations of each species from its equilibrium value. Units are percent of total community. Squares = Differential Mortality 1. Circles = Differential Mortality 2. Point-up triangles = Equal Mortality 1. Point-down triangles = Equal Mortality 2.  YEARS  148 Figure 5.2 Total community difference from equilibrium composition: cold climate and warm climate transition matrices. Units are percent of total community. Both simulations used Differential mortality 1 mortality rates. Data points represent the summed deviations of each species from its equilibrium value. Squares = Cold climate transition matrix. Circles = Warm climate transition matrix.  O  200  400  600  YEARS  800  1000  149 Figure 5.3 Forest change with equal mortality rates and cyclic climate. In each case, the total individuals in the community is 1,000, so that the vertical axis value divided by 10 is percent composition. Different graphs represent different durations of climate states; 50, 100, 200, and 300 years. Upper solid line is Pacific silver fir; Upper broken line is western hemlock; Lower solid line is mountain hemlock; Lower broken line is yellow-cedar.  150 Figure 5.4 Forest change with differential mortality rates among species and cyclic climate. Upper solid line is Pacific silver fir; Upper broken line is western hemlock; Lower solid line is mountain hemlock; Lower broken line is yellowcedar.  600  50 YR  «J < Q  >  100 YR  5  Z  UL  O  a m  co 5  200 YR  300 YR  151 Figure 5.5 Change in coefficient of variation, by species, as the duration of each climate state increases. Data are the same as shown in Figures 5.3 and 5.4. Means corresponding to these CV's are shown in Table 5.4. a. Climate Cyclic, b. Climate Random.  i —  1  1-  i  1  y  300 y  ^ *  200 y  a.  2  0.4 H Z LU  O  LI!  - 100 y  •  0.2  -  50 y"  LL LU  o u •  i  i  PSF  WHEM  MHEM  0.0  AYC  SMOGS  1  z 2  -  LU  U. LL LU  1  1  b.  300 y-  y  o.a  200 y  •  £  u. 3 o z  1  0.4  -  / /  - 100 y"  /  50 y  0.2  O O  0.0  •  i  PSF  WHEM  MHEM  SPECIES  AYC  152 6. CONCLUSIONS  The main theme in the empirical chapters (Chapters 2 and 4) is that Pacific silver fir dominates the population of saplings at Cypress Park, both in the understory of closed canopy areas and among the individuals filling gaps. The overwhelming indication from this snapshot of forest dynamics is that in the future, all other things being equal, the firs will increase at the expense of the other species. The main theme of the modelling chapter (Chapter 5) is that, under a variety of conditions, the above snapshot is a poor indicator of long term trends: all other things are not equal. Despite the dominance of firs among gapfillers, there is a reasonable chance that the forest community is close to its steady state proportions. It is a good first approximation to suppose that the distribution of species among gapmakers does represent differential patterns in mortality (Differential Mortality 1). However, there is no reason to suppose that the distribution of species among gapmakers seen today is in steady state. The important result of Chapter 5 is that forest communities are sensitive to our assumptions about steady state climate and differentials in mortality rates. The general result from the methodological chapter (Chapter 3) echoes this: the views we obtain of long term dynamics are very sensitive to the equilibrium assumptions on which our estimates are based. As population and community ecologists, we need to incorporate a broader range of the variables which drive population or community change in our assessments. We need to be more careful in assuming steady state climate, disturbance regimes, and stand structure. These problems are not unique to forest ecologists (for example, Wiens 1977), but are perhaps more difficult to resolve because of the time scales at which forests change. The modelling chapters do more than cast doubt on empirical estimates. They suggest temporal dynamics that are much more interesting than steady state assumptions would lead us to expect. Examples are the increase in mean yellow-cedar  153 abundance under varying climate scenarios, or the firs cycling at twice the rate of the other species (Chapter 5). Similarly, the sensitivity of stand structure to short fluctuations in gap creation rate (Chapter 3) is of biological as well as methodological interest. The forest at Cypress Park differs from many other forests where gaps have been studied (e.g. few canopy species, predominance of multiple gapmakers and standing dead among gapmakers), but there are some strong similarities to very different systems. For instance, Uhl et al. (1988) found gapfillers in an Amazonian rainforest to be dominated by advanced regeneration and consequently saw little relationship between gapfiller species or distribution and gap environment, Lawton and Putz (1988) demonstrated the importance of different rooting substrates in colonizing gaps in a cloud forest in Costa Rica, and Veblen (1986) emphasized the role of differential mortality rates between Abies lasiocarpa and Picea englemanii in moderating the effect of abundant Abies advanced regeneration. This suggests that there are general principles to be found regarding the roles of gap-phase processes in a variety of systems, despite the differences between them. One such principle may involve the relative importance of early and later lifehistory events in promoting coexistence. Though most gap research has focussed on the role of early life-history events (i.e. the regeneration niche; e.g. Denslow 1980, 1987), in a number of cases, coexistence is more likely to be maintained by processes affecting patterns of mortality (White et al. 1985; Veblen 1986; this study). In these cases, the "senescence niche" may be more important than the regeneration niche. Where gap-phase replacement favours one species at the expense of the others, coexistence may still be possible if: 1) the species less favoured among gapfillers are longer lived than the favoured species, and 2) the less favoured species recruit with a frequency inversely proportional to the difference in longevity. Markov models with differential longevity such as those presented here and in White et al. (1985), and the "storage effect" models of Chesson and Warner (1981; Warner and Chesson 1985;  154 Chesson 1986), demonstrate that it is not enough to consider only short-term differences among species in recruitment probabilities in assessing the effects of a gap regime on community change. Especially under a changing environment, patterns of mortality may be critical. How representative are the stands at Cypress Park of similar forests in the region? In looking for sites in which to conduct this study, I found many areas that were inappropriate because they were too young for gaps to have played a major role in shaping stand structure, or the canopy was too open to delineate clearly gap and closed canopy environments. Probably it is rare for forests in this region to experience major disturbances as infrequently as the one at Cypress Park, and thus for gap-phase processes to have had as much time to act on forest structure. As well, the lack of large, long-lived serai dominants such as Douglas-fir or Sitka spruce (Picea sitchensis) leads to substantial differences in canopy structure (e.g. Stewart 1986a,b; Spies and Franklin 1989b). The Cypress Park forests represent one end of a regional continuum in the importance of gap-phase processes which play a role in most older stands.  155 7. LITERATURE CITED.  Acevedo, M.F.L. 1981 On Horn's Markovian model of forest dynamics with particular reference to tropical forests. Theoretical Population Biology 19:230-250. Allison, T.D., R.E. Moeller, and M.B. Davis. 1986. Pollen in laminated sediments provides evidence for a mid-Holocene forest pathogen outbreak. Ecology 76:1101-1105. Antos, J.A. and D.Z. Zobel. 1986. Habitat relationships of Chamaecvparis nootkatensis in southern Washington, Oregon, and California. Canadian Journal of Botany. 64:1898-1909. Aplet, G.H., R.D. Laven, and F.W. Smith. 1988. Patterns of community dynamics in Colorado Englemann Spruce-subalpine fir forests. Ecology 69:312-319. Armesto, J.J. and E.R. Fuentes. 1988. Tree species regeneration in a mid-elevation, temperate rain forest in Isla de Chiloe, Chile. Vegetatio 74:151-159. Arno, S.F. and R.P. Hammerty. 1977. Northwest trees. The Mountaineers. Seattle. Aubreville, A.M.A. 1938. La foret coloniale: Les forets de I'Afrique occidentale, francaise. Annales Acadamie Sciences Coloniale, Paris 9:1-245. . Barden, L.S. 1979. Tree replacement in small canopy gaps of a Tsuga canadensis forest in the southern Appalachians, Tennessee. Oecologia 44:141-142. Barden, L.S. 1980. Tree replacement in a cove hardwood forest of the southern Appalachians. Oikos 35:16-19. Barden, L.S. 1989. Repeatability in forest gap research: studies in the Great Smoky Mountain. Ecology 70:558-559. Bazzaz, F.A. 1983. Dynamics of wet tropical forests and their species strategies, (in) Medina, E., H.A. Mooney, and C. Vasquez-Yanes (eds.). Physiological ecology of plants of the wet tropics. Boston: Junk. Bazzaz, F.A. and S.T.A. Pickett. 1980. The physiological ecology of tropical succession: a comparative review. Annual review of Ecology and Systematics. 11:287-310. Beatty, S.W. and E.L. Stone. 1986. The variety of microsites created by tree falls. Canadian Journal of Forest Research 16:539-548. Bongers, F., J. Popma, J. Meave del Castillo, and J. Carabias. 1988. Structure and floristic composition of the lowland rain forest of Los Tuxtlas, Mexico. Vegetatio 74:55-80. Bradshaw, H.W. and N.G. Miller. 1988. Recent successional processes investigated by pollen analysis of closed-canopy forest sites. Vegetatio 76:45-54. Bray, J.R. 1956. Gap phase replacement in a maple-basswood forest. Ecology 37:598-600.  156 Brewer, R. and P.G. Merritt. 1978. Wind throw and tree replacement in a climax beech-maple forest. Oikos 30:149-152. Brink, V.C. 1959. A directional change in the subalpine forest-heath ecotone in Garibaldi Park, British Columbia. Ecology 40:10-16. Brokaw, N.V. 1982a. Treefalls: frequency, timing, and consequences, in Leigh, E.G. Jr., A.S. Rand, and D.M. Windsor. The ecology of a tropical forest: seasonal rhythms and long term changes. Smithsonian Institution Press. Washington, D.C.pp.468. Brokaw, N.V. 1982b The definition of treefall gap and its effect on measures of forest dynamics. Biotropica 14:158-160. Brokaw, N.V. 1985a. Treefalls, regrowth, and community structure in tropical forests, in S.T.A. Pickett and P.S. White (eds.), Natural disturbance: an evolutionary perspective. Academic Press, New York. Brokaw, N.V. 1985b. Gap-phase regeneration in a tropical forest. Ecology 66:682687. Brokaw, N.V. 1987. Gap-phase regeneration of three pioneer tree species in a tropical forest. Journal of Ecology 75:9-19. Brooke, R . C , E.B. Peterson, and V.J. Krajina. 1970. The subalpine mountain hemlock zone. Ecology of western North America. 2(2):147-349. Brown, L R . and C. Flavin. 1988. The earth's vital signs, in Worldwatch Institute. State of the world 1988. pp237. Canham, C D . 1988a. Growth and canopy architecture of shade-tolerant trees: response to canopy gaps. Ecology 69:786-795. Canham, C D . 1988b. An index for understory light levels in and around canopy gaps. Ecology 69:1634-1637. Canham, C D . 1989. Different responses to gaps among shade-tolerant tree species. Ecology 70:548-550. Chambers, J.M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphical methods for data analysis. Belmont, CA. Wadsworth International Group. Chazdon, R.L. and N. Fletcher. 1984. Photosynthetic light environments in a lowland tropical rain forest in Costa Rica. J. Ecology 72:553-564. Chesson, P.L. 1986. Environmental variation and the coexistence of species, in Diamond, J. and T.J. Case. Community Ecology. Harper and Row. New York. 269-284. Chesson, P.L. and T.J. Case. 1986. Overview: nonequilibrium community theories: chance, variability, history, and coexistence, in Diamond, J. and T.J. Case. Community Ecology. Harper and Row. New York. 229-239. Chesson, P.L. and R.R. Warner. 1981. Environmental variability promotes coexistence in lottery competitive systems. American Naturalist 117:923-943.  157 Christy, J.E. 1986. Effect of root competition and shading on growth of suppressed western hemlock (Tsuga heterophylla). Vegetatio 65:21-28. Christy, J.E. and R.N. Mack. 1984. Variation in demography of juvenile Tsuga heterophylla across the substratum mosaic. Journal of Ecology 72:75-91. Clark, J.S. 1986a. Dynamism in the barrier-beach vegetation of Great South Beach, New York. Ecological Monographs 56:97-126. Clark, J.S. 1986b. Coastal tree populations in a changing environment, southeastern Long Island, New York. Ecological Monographs 56:259-277. Clark, J.S. 1988. Effect of climate change on fire regimes in northwestern Minnesota. Nature 334:233-235. Clements, F.E. 1916. Plant succession:an analysis of the development of vegetation. Carnegie Inst. Washington Publ. No. 242:1-512. Clements, F.E. 1936. Nature and structure of the climax. Journal of Ecology 24:252284. Coale, A.J. 1972. The growth and structure of human populations. Princeton University Press. Princeton, pp 227. Cohen, J. 1987. Statistical power analysis for the behavioral sciences. Seconded. Lawrence Erlbaum Associates, Publishers, New Jersey, pp 474. Colinvaux, P. 1987. Amazon diversity in the light of the palaeoecological record. Quaternary Science Reviews 6:93-114. Collins, B.S., K.P. Dunne and S.T.A. Pickett. 1985. Responses of forest herbs to canopy gaps, in The ecology of natural disturbance and patch dynamics. Pickett, S. T. A. and P. S. White (eds.). Academic Press, New York. Connell, J.H. 1978. Diversity in tropical rainforests and coral reefs. Science 199:13021310. Connell, J.H. and R.O. Slatyer. 1977. Mechanisms of succession in natural communities and their role in community stability and organization. American Naturalist 111:1119-1144. Crome, F.H.J, and G.C. Richards. 1988. Bats and gaps: Microchiropteran community structure in a Queensland rain forest. Ecology 69:1960-1969. Cox, D.R. and H.D. Miller. 1965. The theory of stochastic processes. Chapman and Hall, pp.398. Cullen, P.J. 1987. Regeneration patterns in populations of Athrotaxis selaoinoides D. Don. from Tasmania. Journal of Biogeography 14:39-51. Cwynar, L C . 1987. Fire and the forest history of the North Cascade Range. Ecology 68:791-802. Davis, M.B. 1981. Quaternary history and the stability of forest communities. jriD.C. West, H.H. Shugart and D.B. Botkin, (eds.), Forest Succession: Concepts and Application. Springer Verlag.  158 Davis, M.B. 1986a. Climatic instability, time lags, and community disequilibrium, in Diamond, J. and T.J. Case. Community Ecology. Harper and Row. New York. 269-284. Dayton, P.K. 1971. Competition, disturbance, and community organization: the provision and subsequent utilization of space in a rocky intertidal community. Ecological Monographs 41:351 -389. Delcourt, P.A. and H.R. Delcourt. 1987. Long-term forest dynamics of the temperate zone. Springer-Verlag. pp439. Denslow, J.S. 1980. Gap partitioning among tropical rainforest trees. Biotropica, Supplement: Tropical Succession 12:47-55. Denslow, J.S. 1985. Disturbance mediated coexistence of species, in The ecology of natural disturbance and patch dynamics. Pickett, S. T. A. and P. S. White (eds.). Academic Press, New York. Denslow, J.S. 1987. Tropical rainforest gaps and tree species diversity. Ann. Rev. Ecology and Syst. 18:431 -451. Dickinson, R.E. and R.J. Cicerone. 1986. Future global warming from atmospheric trace gases. NAture:319:109-114. Drury, W.H. and I.T.C. Nisbet. 1973. Succession. Journal of the Arnold Arboretum 54:331-363. Dunwiddie, P.W. 1986. A 6,000 year record of forest history on Mount Ranier, Washington. Ecology 76:58-68. Eis, S. 1962. Statistical anlysis of several methods for estimation of forest habitats and tree growth near Vancouver, B.C. University of British Columbia, Faculty of Forestry, Forestry Bulletin No. 4, Vancouver. 76 p. Feinsinger, P., W.H. Busby, K.G. Murray, J.H. Beach, W.Z. Pounds, and Y.B. Linhart. 1988. Mixed support for spatial heterogeneity in species interactions: hummingbirds in a tropical disturbance mosaic. American Naturalist 131:33-57. Feinsinger, P. 1989. A framework for comparative studies in forest dynamics: the gapsnap-drap-sap-nap-crap model. Bulletin of the Ecological Society of America 70:25-27. Foster, D.R. 1988. Disturbance history, community organization and vegetation dynamics of the old-growth Pisgah forest, south-western New Hampshire, U.S.A. Journal of Ecology 76:105-134. Foster, J.R. and W.A. Reiners. 1983. Vegetation patterns in a virgin sub-alpine forest at Crawford Notch, White Mountains, New Hampshire. Bulletin of the Torrey Botanical Club 110:141-269. Foster, J.R. and W.A. Reiners. 1986. Size distribution and expansion of canopy gaps in a northern Appalachian spruce-fir forest. Vegetatio 68:109-114. Foster, R.B. and N.V.L. Brokaw. 1982. Structure and history of the vegetation of Barro Colorado Island, in Leigh, E.G. Jr., A.S. Rand, and D.M. Windsor. The ecology  159 of a tropical forest: seasonal rhythms and long term changes. Smithsonian Institution Press. Washington, D.C.pp.468. Fowells, H.A. 1965. Silvics of forest trees of the United States. U.S.D.A. Handbook No. 271. U.S. Govt. Printing Office. Washington, D.C. USDA Handbook 271. Fox, J.F. Alternation and coexistence of tree species. 1977. American Naturalist 111:69-89. Franklin, J.F. 1988. Pacific northwest forests, in Barbour, M.G. and W.D. Billings (eds.). North American Terrestrial Vegetation. Cambridge University Press, pp. 103-130. Franklin, J.F., W.H. Moir, G.W. Douglas, and C. Wiberg. 1971. Invasion of subalpine meadows by trees in the Cascade Range, Washington and Oregon. Arctic and Alpine Research 3:215-224. Franklin, J.F. and C.T. Dyrness. 1973. Natural vegetation of Oregon and Washington. USDA Forest Service Gen. Tech. Rep. PNW-8. Franklin, J.F. and R.H. Waring. 1980. Distinctive features of the northwestern coniferous forest, pp. 59-86. in R.H. Waring (ed.) Forests: fresh perspectives from ecosystem analysis. Proc. 40th Biol. Colloq. Oregon State University Press. Corvallis, Oregon. Franklin, J.F. and M.A. Hemstrom. 1981. Aspects of succession in coniferous forests of the Pacific Northwest, in D.C. West, H.H. Shugart and D.B. Botkin, (eds.), Forest Succession: Concepts and Application. Springer Verlag. Franklin, J.F., H.H. Shugart, and M.E. Harmon. 1987. Tree death as an ecological process: the causes, consequences and variability of tree mortality. Bioscience 37:550-556. Franklin, J.F. and DeBell. 1988. Thirty-six years of tree population change in an oldgrowth Pseudotsuga-Tsuga forest. Canadian Journal of Forest Research 18:633-639. Franklin, J.F., W.H. Moir, M.A. Hemstrom, S.E. Greene, and B.G. Smith. 1988. The forest communities of Mount Ranier National Park. Scientific Monograph Series No. 19. U.S. Department of the Interior. National Park Service. Washington, D.C. pp. 194. Garwood, N.C. 1983. Seed germination in a seasonal tropical forest in Panama: a community study. Ecological Monographs 53:159-181. Gleason, H.A. 1917. The structure and development of the plant association. Bulliten of the Torrey Botanical Club 44:463-481. Gleason, H.A. 1926. The individualistic concept of the plant association. Bulletin of the Torrey Botanical Club 53:7-26. Glitzenstein, J.S., P. A. Harcombe and D.R. Streng. 1986. Disturbance, succession, and maintenance of species diversity in an east Texas forest. Ecological Monographs 56:243-258.  160 Graham, R.L and K. Cromack, Jr. 1982. Mass, nutrient content and decay rate of dead boles in rain forests of Olympic National Park. Canadian Journal of Forest Research. 12:511-521. Grant, J.C. 1980. Growth comparisons between Pacific silver-fir and western hemlock on the western slopes of the northern Washington Cascades. M.Sc. Thesis. University of Washington, Seattle. Graumlich, L.J. and L B . Brubaker. 1986. Reconstruction of annual temperature (15901979) for Longmire, Washington, derived from tree rings. Quaternary Research 25:223-234. Graumlich, L.J., L B . Brubaker and C.C. Grier. 1989. Long-term trends in forest net primary productivity: Cascade Mountains, Washington. Ecology 70:405-410. Green, D.G. 1982. Fire and stability in the postglacial forests of southwest Nova Scotia. J. Biogeography 9:29-40. Green, R.N., P.J. Courtin, K. Klinka, R.J. Slaco, and C A . Ray. 1984. Site diagnosis, tree species slection and slashburning guidelines for the Vancouver Forest Region. Land Management Handbook 8, B.C. Ministry of Forests, Victoria, B.C., 143 pp. Grier, C.C. and R.S. Logan 1977. Old-growth Pseudotsuga menziesii communities of a western Oregon watershed: Biomass distribution and production budgets. Ecological Monographs 47:373-400. Grubb, P.J. 1977. The maintenance of species-richness in plant communities: The importance of the regeneration niche. Biol. Rev. 52:107-145. Harcombe, P.A. and P.L. Marks. 1978. Tree diameter distributions and replacement processes in southeast Texas forests. Forest. Sci. 24:153-166. Harcombe, P.A. and P.L. Marks. 1983. Five years of tree death in a Faqus-Maanolia forest, Southeast Texas (USA). Oecologia 57:49-54. Harmon, M.E. and J.F. Franklin. 1989. Tree seedlings on logs in Picea-Tsuga forests of Oregon and Washington. Ecology 70:48-59. Hartshorn, G.S. 1978. Tree falls and tropical forest dynamics, in P.B. Tomlinson and M.H. Zimmerman, (eds.), Tropical trees as living systems. Cambridge University Press. Cambridge. Heinselman, M.L. 1973. Fire in the virgin forests of the Boundary Waters Canoe Area, Minnesota. Quat. Res. 3:329-382. Hemstrom, M.A. and J.F. Franklin. 1982. Fire and other disturbances of the forests in Mount Ranier National Park. Quat. Res. 18:32-51. Henry, J.D. and J.M.A. Swan. 1974. Reconstructing forest history from live and dead plant material-an approach to the study of forest succession in southwest New Hampshire. Ecology 55:772-783. Herring, L.J. and D.E. Etheridge. 1976. Advance amabilis-fir regeneration in the Vancouver forest district. BCFS/CFS Joint report no. 5.  161 Hibbs, D.A. 1982. Gap dynamics in a hemlock-hardwood forest. Canadian Journal of Forest Research 12:522-527. Hobbs, R.J. 1983. Markov models in the study of post-fire succession in heathland communities. Vegetatio 56:17-30. Hobbs, R.J. and C.J. Legg. 1983. Markov models and initial floristic composition in heathland vegetational change. Vegetatio 56:31 -43. Horn, H.S. 1971. The adaptive geometry of trees. Princeton Univ. Press. Princeton.pp.144. Horn, H.S. 1975. Markovian properties of forest succession, in M.L. Cody and J. Diamond (eds.), Ecology and evolution of communities. Belknap Press. Cambridge, Mass. Horn, H.S. 1981. Succession, in R.M. May (ed.), Theoretical ecology (2cd ed.). Blackwell, Oxford. Hubbell, S.P. 1979. Tree dispersion, abundance, and diversity in a tropical dry forest. Science 203:1299-1309. Hubbell, S.P. and R.B. Foster. 1986a. Biology, chance, and history and the structure of tropical rainforest tree communities, in Diamond, J. and T.J. Case. Community Ecology. Harper and Row. New York. 314-329. Hubbell, S.P. and R.B. Foster. 1986b. Canopy Gaps and the dynamics of a neotropical forest, in M.J. Crawley (ed.) Plant Ecology. Blackwell Scientific Publications. Oxford, pp. 77-96. Hubbell, S.P. and R.B. Foster. 1986c. Commonness and rarity in a neotropical forest: implications for tropical tree conservation, in Soule, M.E. Conservation Biology. Sinauer Associates, Inc. Massachusetts, pp. 205-231. Hulst, R. van. 1979a. On the dynamics of vegetation: succession in model communities. Vegetatio 39:85-96. Hulst, R. van. 1979b. On the dynamics of vegetation: Markov chains as models of succession. Vegetatio 40:3-14. Jones, E.W. 1945. The structure and reproduction of the virgin forest of the north temperate zone. New Phytol. 44:130-148. Jones, P.D., T.M.L. Wigley, C.K. Folland, D.E. Parker, J.K. Angell, S. Lebedeff, and J.E. Hansen. 1988. Evidence for global warming in the past decade. Nature 322:790. Kemeny, J.G., and J.L. Snell. 1960. Finite Markov chains. Van Nostrand Company, Inc. Princeton, pp 210. Koyama, T. 1983. Seedling stage of two subalpine Abies species in distinction from sapling stage: a matter economic analysis. Botanical Magazine of Tokyo 96:4965. Korstian, O F and T.S. Coile. 1938. Plant competition in forest stands. Duke University School of Forestry Bull. 3. pp 125.  162 Kotar, J. 1972. Ecology of Abies amabilis in relation to its altitudinal distribution and in contrast to its common associate Tsuga heterophylla. Dissertation. University of Washington, Seattle. Krajina, V.J. 1970. Ecology of forest trees in British Columbia. Ecology of western North America. 2(1):1-146. Krebs, C.J. 1989. Ecological methodology. Harper and Row. pp 654. Kullmann, L. 1983. Past and present tree-lines of different species in the Handolan Valley, central Sweden, in P. Morisset and S. Payette (eds.), Tree-line ecology. Collection Nordicana, No. 47. Universite Laval, Quebec. Kullman, L. 1987. Long-term dynamics of high-altitude populations of Pinus sylvestris in the Swedish Scandes. Journal of Biogeography 14:1-8. Lamb, H.H. 1982. Climate, history, and the modern world. Methuen, London. Lang, G.E. and D.H. Knight. 1983. Tree growth, mortality, recruitment, and canopy gap formation during a 10-year period in a tropical moist forest. Ecology 64:1075-1080. Lassoie, J.P., T.M. Hinckley, and C.C. Grier. 1986. Coniferous forests of the Pacific Northwest, in Chabot, B.F. and H.A. Mooney, Physiological ecology of North American plant communities. Chapman and Hall. Lawton, R.O. and F.E. Putz. 1988. Natural disturbance and gap-phase regeneration in a wind-exposed tropical cloud forest. Ecology 69:764-777. Lertzman, K.P. and L.B. Brubaker. (in prep.) Climate, fires, and forest history: the late Quaternary development of a sub-alpine old growth forest. Levey, D.J. 1988. Tropical wet forest treefall gaps and distributions of understory birds and plants. Ecology 69:1076-1089. Lieberman, M, D. Lieberman, R. Peralta and G. Hartshorn. 1985. Mortality patterns and stand turnover rates in a wet tropical forest in Costa Rica. Journal of Ecology 73:915-924. Lippe, E., J.T. de Smidt and D.C. Glenn-Lewin. 1985. Markov models and succession: a test from a heathland in the Netherlands. Journal of Ecology 73:775-791. Long, J.N. 1976. Forest vegetation dynamics within the Abies amabilis Zone of a western Cascades watershed. Dissertation. University of Washington, Seattle. Lorimer, C G . 1980. Age structure and disturbance history of a southern Appalachian virgin forest. Ecology 6:1169-1184. Ludwig, D. 1989. Small models are beautiful: efficient estimators are even more beautiful. Mathematical approaches to environmental and ecological problems. (Conference held at Cornell University) Springer-Verlag, New York, N.Y. (in press) Ludwig, D. and C J . Walters. 1989. A robust method for parameter estimation from catch and effort data. Canadian Journal of Fisheries and Aquatic Sciences 46:137-144.  163 Lutz, H.J. 1940. Disturbance of forest soil resulting from uprooting of trees. Yale University School of Forestry Bull. No. 45. pp. 37. Martinez-Ramos, M., E. Alvarez-Buylla, J. Sarukhan, and D. Pinero. 1988. Treefall age determination and gap dynamics in a tropical forest. Journal of Ecology 76:700-716. McAuliffe, J.R. 1988. Markovian dynamics of simple and complex desert plant communities. American Naturalist 131:459-490. McKee, A., G. LaRoi, and J.F. Franklin. 1980. Structure, composition, and reproductive behavior of terrace forests, South Fork Hoh River, Olympic National Park. Proc, Second Conference on research in the National Parks, 1980. National Park Service, Washington, DC. Means, J.E., K. Cromack, Jr., and P.C. MacMillan. 1985. Comparison of decomposition models using wood density of Douglas-fir logs. Canadian Journal of Forest Research 15:1092-1098. Minore D. 1979. Comparative autecological characteristics of north-western tree species. United States Department of Agriculture Pacific Northwest Forest and Range Experiment Station Technical Report 87. 72 pp. Mladenhoff, D.J. 1987. Dynamics of nitrogen mineralization and nitrification in hemlock and hardwood treefall gaps. Ecology 68:1171-1180. Naka, K. 1982. Community dynamics of evergreen broadleaf forests in southwestern Japan. 1. Wind damaged trees and canopy gaps in an evergreen oak forest. Bot. Mag. Tokyo 95:385-399. Nakashizuka, T. 1983. Regeneration process of climax beech (Fagus crenata Blume) forests. III. Structure and development processes of spaling populations in different aged gaps. Japanese Journal of Ecology 33:409-418. Odum, E.P. 1969. The strategy of ecosystem development. Science 164:262-270. Oliver, C D . 1976. Growth response of suppressed hemlocks after release, in Atkinson, W.A. and R.J. Zasoski (eds.), Western hemlock management. University of Washington College of Forest Resources, Institute of Forests Products Contribution Number 34. Oliver, C D . and E.P. Stephens. 1977. Reconstruction of a mixed-species forest in central New England. Ecology 58:562-572. Orians, G.H. 1982. The influence of tree-falls in tropical forests in tree species richness. Tropical Ecology 23:255-279. Packee, E . C , C D . Oliver, and P.D. Crawford. 1982. Ecology of pacific silver fir. in Oliver, C D . and R.M. Kenady (eds.), Proc. of the Biology and management of true fir in the Pacific Northwest symposium. University of Washington, College of Forest Resources. Paine, R.T. and S.A. Levin. 1981. Intertidal landscapes: disturbance and the dynamics of form. Ecological Monographs 51:145-178.  164 Panshin, A.J. and C. de Zeeuw. 1980. Textbook of wood technology. McGraw Hill, New York. Payette, S., L. Filion, L. Gauthier, and Y. Boutin. 1985. Secular climate change in oldgrowth tree-line vegetation of northern Quebec. Nature 315:135-138. Payette, S. and R. Gagnon. 1985. Late Holocene deforestation and tree regeneration in the forest-tundra of Quebec. Nature 313:570-572. Payette, S., C. Morneau, L. Sirois, and M. Desponts. 1989. Recent fire history of the Northern Quebec biomes. Ecology 70:656-673. Peet, R.K., and N.L Christenson. 1980. Succession: a population process. Vegetatio 43:131-140. Pickett, S.T.A. 1982. Population patterns through twenty years of oldfield succession. Vegetatio 49:45-59. Pickett S.T.A. and P.S. White. 1985. The ecology of natural disturbance and patch dynamics. Academic Press, Inc. pp 472. Pojar, J., K. Klinka, and D.V. Meidinger. 1987. Biogeoclimatic ecosystem classification in British Columbia. Forest Ecology and Management 22:119-154. Pojar, J., K. Klinka, and D.V. Meidinger. 1987. Biogeoclimatic ecosystem classification in British Columbia. Forest Ecology and Management 22:119-154. Poore, M.E.D. 1968. Studies in Malaysian rainforest. Journal of Ecology 56:143-189. Poulson, T.L. and W.J. Piatt. 1989. Gap light regimes influence canopy tree diversity. Ecology 70:553-555. Putz, F.E. 1983. Treefall pits and mounds, buried seeds, and the importance of soil disturbance to pioneer trees on Barro Colorado Island, Panama. Ecology 64:1069-1074. Putz, F.E. and K. Milton. 1982. Tree mortality rates on Barro Colorado Island, in Leigh, E.G. Jr., A.S. Rand, and D.M. Windsor. The ecology of a tropical forest: seasonal rhythms and long term changes. Smithsonian Institution Press. Washington, D.C.pp.468. Putz, F.E., P.D. Coley, K. Lu, A. Montalvo, and A. Aiello. 1983. Uprooting and snapping of trees: structural determinants and ecological consequences. Canadian Journal of Forest Research 13:1011-1020. Putz, F.E. and N.V.L. Brokaw. 1989. Sprouting of broken trees on Barro Colorado Island, Panama. Ecology 70:508-512. Ramanathan, V., R.D. Cess, E.F. Harrison, P. Minnis, B.R. Barkstrom, E. Ahmad, and D. Hartmann. 1989. Cloud-radiative forcing and climate: results from the earth radiation budget experiment. Science 243:57-63. Richards, P.W. 1952. The tropical rainforest. Cambridge University Press. Cambridge. Ritchie, J.C. 1986. Climate change and vegetation response. Vegetatio 67:65-74.  165 Romme, W.H. and W.H. Martin. 1982. Natural disturbance by treefalls in old-growth mixed mesophytic forest: Lilley Cornett Woods, Kentucky, in R.N. Muller (ed.), Central hardwoods forest conference IV. Proceedings. Univ. of Kentucky, pp. 367-383. Runkle, J.R. 1981. Gap regeneration in some old-growth forests of the eastern United States. Ecology 62:1041-1051. Runkle, J.R. 1982. Patterns of disturbance in some old-growth mesic forests of eastern North America. Ecology 63:1533-1546. Runkle J.R. 1985. Disturbance regimes in temperate forests, in Pickett S.T.A. and P.S. White. The ecology of natural disturbance and patch dynamics. Academic Press, Inc. pp 472. Runkle, J.R. 1987. A protocol for sampling forest gaps. Unpublished MS. Runkle, J.R., and T.C. Yetter. 1987. Treefalls revisited: gap dynamics in the southern Appalachians. Ecology 68:417-424. Schemske D.W. and N. Brokaw. 1981. Treefalls and the distribution of understory birds in a tropical forest. Ecology 62:938-945 Schupp, E.W., H.F. Howe, C.K. Augspurger, and D.J. Levey. 1989. Arrival and survival in tropical treefall gaps. Ecology 70:562-564. Shelly, T.E. 1988. Relative abundance of day flying insects in treefall gaps vs shaded understory in a Neotropical forest. Biotropica 20:114-119. Sollins, p. 1982. Input and decay of coarse woody debris in coniferous stands in western Oregon and Washington. Canadian Journal of Forest Research. 12:1828. Sousa, W.P. 1984. Intertidal mosaics: propagule availability, and spatially variable patterns of succession. Ecology 65:1918-1935. Spies, T.A., J.F. Franklin and T.B. Thomas. 1988. Coarse woody debris in Douglas-fir forests of western Oregon and Washington. Ecology 69:1689-1702. Spies, T.A., J.F. Franklin. 1989a. Gap characteristic and vegetation response in coniferous forests of the Pacific Northwest. Ecology 70:543-545. Spies, T. and J.F. Franklin. 1989b. (in press, Canadian Journal of Forest Research). Gap-phase dynamics of an old-growth Douglas-fir forest in Oregon. Canadian Journal of Forest Research. Steijlen, I. and O. Zackrisson. 1987. Long-term regeneration dynamics and successional trends in a northern Swedish coniferous forest stand. Canadian Journal of Forest Research 65:839-848. Stephens, E.P. 1956. The uprooting of trees: a forest process. Soil Science Society of America, Proceedings, 20:113-116. Stephens, G.R. and P.E, Waggoner. 1980. A half century of natural transitions in mixed hardwood forests. Bulletin of the Connecticut Agricultural Experiment Station, New Haven, No. 783.  166 Stewart, G.H. 1986a. Population dynamics of a montane conifer forest, western Cascade Range, Oregon, U.S.A. Ecology 67:534-544. Stewart, G.H. 1986b. Forest development in canopy openings in old-growth Pseudotsuga forests of the western Cascade Range, Oregon. Canadian Journal of Forest Research 16:558-568. Taylor, A.H. and Q. Zisheng. 1988a. Regeneration patterns in old-growth Abies-Betula forests in the Wolong Natural Reserve, Sichuan, China. Journal of Ecology 76:1204-1218. Taylor, A.H. and Q. Zisheng. 1988b. Tree repacement patterns in subalpine AbiesBetula forests. Wolong Natural Reserve. China. Vegetatio 78:141-149. Thoreau, H.D. 1860. On the succession of forest trees. Reprinted in 1980: Natural History Essays. Literature of the American Wilderness. Peregrine Smith Inc. Thornburgh, D.A. 1969. Dynamics of the true fir-hemlock forests of the west slope of the Washington Cascade Range. Dissertation. University of Washington, Seattle. Toft, C A . and P.J. Shea. 1983. Detecting community-wide patterns: estimating power strengthens statistical inference. 122:618-625. Tourney, J.W. and R. Kienholz. 1931. Trenched plots under forest canopies. Yale University School of Forestry Bull. 30:1-31. Tucker, G.F., T.M. Hinckley, J. Leverenz, and S.M. Jiang. 1987. Adjustments of foliar morphology in the acclimation of understory Pacific silver fir following clearcutting. Forest Ecology and Management 21:249-268. Triska, F.J. and K. Cromack, Jr. 1979. The role of wood debris in forests and streams, in R.H. Waring (ed.) Forests: fresh perspectives from ecosystem analysis. Proc. 40th Biol. Colloq. Oregon State University Press. Corvallis, Oregon. Ugolini, F . C 1982. Soil development in the Abies amabilis Zone of the central Cascades, Washington, in Oliver, C.D. and R.M. Kenady (eds.), Proc. of the Biology and management of true fir in the Pacific Northwest symposium. University of Washington, College of Forest Resources. Uhl, C , K. Clark, N. Dezzeo, and P. Maquirino. 1988. Vegetation dynamics in Amazonian treefall gaps. Ecology 69:751-763. USDA Forest Service. 1974. Seeds of woody plants of the United States. Agriculture handbook No.450. U.S. Govt. Printing Office. Washington, D.C Usher, M.B. 1981. Modelling ecological succession, with particular reference to Markovian models. Vegetatio 46:11-18. van Hulst, R. 1979a. The dynamics of vegetation: Markov chains as models of succession. Vegetatio 40:3-14. van Hulst, R. 1979b. The dynamics of vegetation: succession in model communities. Vegetatio 39:85-96.  167 Veblen, T.T. 1985. Forest development in tree-fall gaps in the temperate rain forests of Chile. Nat. Geog. Res. 1:162-183. Veblen, T.T. 1986. Treefalls and coexistence of conifers in subalpine forests of the central Rockies. Ecology 67:644-649. Veblen, T.T. 1989. Tree regeneration responses to gaps along a transandean gradient. Ecology 70:541-543. Vitousek, P.M. and J.S. Denslow. 1986. Nitrogen and phosphorus availability in treefall gaps of a lowland tropical rainforest. Journal of Ecology 74:1167-1178. Wagner, R.G. 1980. Natural regeneration at the edge of an Abies amabilis Zone clearcut on the west slope of the central Washington Cascades. M.Sc. Thesis. University of Washington, Seattle. Walters, B.W. and K. P. Lertzman. (in prep.) Differential use of canopy gaps by small mammals in an old-growth forest. Waring, R.H. and J.F. Franklin 1979. Evergreen coniferous forests of the Pacific Northwest. Science 204:1380-1386 Warner, R.R. and P.L. Chesson. 1985. Coexistence mediated by recruitment fluctuations: a field guide to the storage effect. American Naturalist 125:769-787. Watt, A.S. 1947. Pattern and process in the plant community. Journal of Ecology 35:1-22. West, D.C, H.H. Shugart and D.B. Botkin. 1981. Forest succession: concepts and application. Springer-Verlag. New York. pp. 517. White, P.S. 1979. Pattern, process, and natural disturbance in vegetation. Bot. Rev. 45:229-299. White, P.S., M. MacKenzie, and R. Busing. 1985. A critique on overstory/understory comparisons based on transition probability analysis of an old growth spruce-fir stand in the Appalachians. Vegetatio 64:37-45. Whitmore, T . C 1978. Gaps in the canopy, in P.B. Tomlinson and M.H. Zimmerman, (eds.), Tropical trees as living systems. Cambridge University Press. Cambridge. Whitmore, T.C. 1989. Canopy gaps and two major groups of forest trees. Ecology 70:536-538. Whitney, G.G. 1984. Fifty years of change in the arboreal vegetation of Heart's Content, an old-growth hemlock-white pine-northern hardwood stand. Ecology 65:403-408. Whittaker, R.H. 1953. A consideration of the climax theory: the climax as population and pattern. Ecological Monographs 23:41-78. Wiens, J.A. 1977. On competition and variable environments. American Scientist 65:590-597. Wilkinson, L. 1988a. SYGRAPH. Evanston, IL: SYSTAT, Inc.  168 Wilkinson, L. 1988b. SYSTAT: The system for statistics. Evanston, IL: SYSTAT, Inc. Woods, K.D. 1979. Reciprocal replacement and the maintenance of codominance in a beech-maple forest. Oikos 33:31-39. Woods, K.D. 1984. Patterns of tree replacement: canopy effects on understory pattern in hemlock-northern hardwood forests. Vegetatio 56:87-107.  

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0098316/manifest

Comment

Related Items