L A N D M A N A G E M E N T H A N D B O O K

40

Field Studies of Seed Biology

1997

Ministry of Forests Research Program Field Studies of Seed Biology

Carole L. Leadem, Sharon L. Gillies, H. Karen Yearsley, Vera Sit, David L. Spittlehouse, and Philip J. Burton

Ministry of Forests Research Program Canadian Cataloguing in Publication Data Main entry under title: Field studies of seed biology

(Land management handbook ; )

“Tree seed biology.”--Cover.  ---

. Trees – British Columbia – Seeds – Experiments. . Trees – Seeds – Experiments. . Reforestation – British Columbia – Experiments. . Reforestation – British Columbia – Experiments. I. Leadem, Carole Louise Scheuplein, – II. British Columbia. Ministry of Forests. Research Branch. III. Series

..  .’’ --

©  Province of British Columbia Published by the Research Branch B.C. Ministry of Forests  Bastion Square Victoria, BC  

Copies of this and other Ministry of Forests titles are available from Crown Publications Inc.  Fort Street Victoria, BC  

Please address any comments or suggestions to the senior author: Carole L. Leadem Glyn Road Research Station B.C. Ministry of Forests PO Box  Stn Prov Govt Victoria, BC  

ii CREDITS

Carole L. Leadem B.C. Ministry of Forests, Research Branch, Glyn Road Research Station, PO Box  Stn Prov Govt, Victoria, BC  

Sharon L. Gillies University College Fraser Valley,  King Road, Abbotsford, BC  

H. Karen Yearsley B.C. Ministry of Forests, Research Branch, PO Box  Stn Prov Govt, Victoria, BC  

Vera Sit B.C. Ministry of Forests, Research Branch, PO Box  Stn Prov Govt, Victoria, BC vw c

David L. Spittlehouse B.C. Ministry of Forests, Research Branch, PO Box  Stn Prov Govt, Victoria, BC vw c

Philip J. Burton Symbios Research and Restoration, PO Box , Smithers, BC  

Editor, indexer: Fran Aitkens Typesetting: Dynamic Typesetting Graphic production: Lyle Ottenbreit Proofreading: Rosalind C. Penty Steve Smith Publication design: Anna Gamble

Original photos and illustrations: Figures ., ., . Dr. David Spittlehouse Figure . Dr. John Owens, Dep. Biol., Univ., Victoria, B.C. Figures . and . Paul Nystedt, B.C. Min. For. Figures ., ., . H. Karen Yearsley Figures . and . Dr. D.G.W. Edwards, Can. For. Serv., (retired) Pac. For. Cent., Victoria, B.C.

iii INTRODUCTION

I like trees because they seem more resigned to the way they have to live than other things do. (Willa Cather “O Pioneers!”)

Except in limited areas where there is enough methods for conducting field studies of tree seeds. A advance regeneration, establishment of forest cover recent assessment of ecosystem management needs on harvested lands continues to depend on seedling stressed the importance of standardized sampling planting programs or on natural regeneration by and monitoring techniques, and the lack of consist- seeds. Whereas successful plantation programs ent methods for archiving, accessing, and updating depend primarily on plant competition and site databases (U.S. Dep. Agric. For. Serv. a). Tech- variables at the time of planting, successful natural niques gleaned from agriculture literature are generally regeneration depends not only on the availability not applicable, and traditional ecological studies of seeds, but on favourable environmental condi- (e.g., of seed banks) tend to be primarily descriptive tions throughout the processes of seed production, with little emphasis on experimental approaches. dispersal, germination, and seedling establishment. The primary objective of this manual is to detail Site preparation and other silvicultural treat- methods that have been gleaned from the literature ments can improve the suitability of the seedbed and from personal experience of the authors. It is a and its micro-environment, but there is still much manual of methods with some general guidelines we do not understand about how various factors and interpretation. Relevant background papers are contribute to successful forest establishment. We cited where appropriate, but it is not a literature have gained some insights, under controlled condi- review. The manual is intended for use by researchers tions, about the influence of major factors such as in public and private forest resource management light and temperature, but we have limited experi- agencies, universities, and colleges. Although specifi- ence with biological responses under actual cally directed to tree seed research in forested conditions in the field. ecosystems, many of the methods described can Anyone who has conducted research in the field be used to study seeds of graminoid, herb, and quickly comes to realize the complexity of the sys- shrub species in both forest and non-forest plant tems chosen for study. An immense number of communities. The extensive background informa- external and internal factors that affect living tion included in the text also provides valuable organisms must be taken into account—with lim- reference material for many who have an interest ited possibilities to control these factors. A major in tree seeds, but who are not directly involved in constraint, particularly in a forest environment, is research activities. The detailed examples from the difficulty inherent in conducting field studies previous studies are included, not to prescribe how involving seeds. Infrequent seed production, preda- such studies should be done, but to assist in plan- tion by animals, difficulty locating small seeds, ning by providing reference values on which to estimating the numbers of buried seeds, measuring base measurements, sample sizes, and other germination, and monitoring survival pose myriad experimental details. challenges for the field researcher. Added to these Since the manual is directed primarily to re- difficulties is the lack of information about effective searchers working in the province of British

v Columbia (B.C.), Canada, many examples (forest types, species, research topics), procedures (the biogeoclimatic ecosystem classification system), and regulatory policies are specific to this geographic and political jurisdiction. Nevertheless, it is hoped that the underlying principles are self-evident and will be generally applicable to the conduct of field research elsewhere.

vi ORGANIZATION OF THE HANDBOOK

Following a discussion of planning and organizing a H. Karen Yearsley, M.Sc., R.P.Bio., earned her field study (Section ) and setting up an environ- graduate degree from the Faculty of Forestry, Uni- mental monitoring program for the experimental versity of British Columbia, and is a member of the site (Section ), the manual is arranged by subject Association of Professional Biologists of British areas most often associated with field studies of tree Columbia. Her  years of research experience in seeds: natural seed production (Section ), seed dis- B.C. include work on ecosystem classification, persal (Section ), seed predation (Section ), seed forest succession, and forest soil seed banks. Karen banks (Section ), assessing seed quality and viabil- Yearsley wrote the sections on seed predation ity (Section ), and effects of silvicultural practices (Section ) and soil seed banks (Section ), and on emergence (Section ). Each section was written contributed to the sections on planning field by one or more experts as follows: studies (Table .) and seed dispersal (Section ). Carole Leadem, Ph.D., R.P.Bio., earned her Vera Sit, M.Sc., earned her graduate degree from degree in plant physiology from the Botany Depart- the Statistics Department, Dalhousie University, and ment, University of British Columbia, and is a is a member of the Statistical Society of Canada. She member of the Association of Professional Biolo- has been with Biometrics Section, Research Branch, gists of British Columbia. She has been in charge of B.C. Ministry of Forests, since . Vera Sit wrote the tree seed biology research program with the B.C. the sections on experimental design and data analy- Ministry of Forests in Victoria since . Carole sis (Sections ., ., ., ., ., ., .) and the Leadem wrote the sections on planning and case studies (Section .), and reviewed and contrib- organizing field studies (Section ), natural seed uted to all the statistical sections. production (Section ), seed responses to the envi- David Spittlehouse, Ph.D., P.Ag., earned his ronment (Section .), seed testing in the laboratory graduate degree in forest climatology from the De- (Section .), seedbed preferences (Section ..), partment of Soil Science, University of British and contributed to the sections on seed dispersal Columbia. His research includes modifying site and silvicultural practices. microclimate to improve seedling regeneration, and Sharon Gillies, Ph.D., earned her degree in plant determining how forest harvesting and regrowth of physiology from the Department of Biological Sci- the forest affects forest hydrology. He has worked ences, Simon Fraser University. She has been a for the B.C. Ministry of Forests in Victoria since biology instructor at the University College Fraser . Dave Spittlehouse wrote most of the section Valley since . Sharon Gillies coordinated compi- on designing an environmental monitoring program lation of the original manuscript, was responsible (Section ). for creating the handbook structure and adhering Philip Burton, Ph.D., R.P.Bio., earned his degree to Ministry of Forests style manual, edited author in plant biology from the University of Illinois at submissions for the first complete draft, wrote the Urbana-Champaign. An independent researcher and section on seed dispersal (Section ), and provided consultant, he has been investigating seed biology, environmental monitoring material for Section , forest regeneration, and vegetation dynamics since and Table . on seedbed suitability. . Phil Burton contributed material for the

vii sections on seed dispersal (Section ), field germina- contemplating new research projects to avoid some tion studies (Section .), and effects of silvicultural of the pitfalls associated with studies in the field. We practices (Section ). anticipate other benefits: that this handbook will help Each section contains background material on standardize field methods and enable comparisons the subject and descriptions of some of the methods between studies, will increase cooperation between and approaches that have been used. There is also investigators, and will promote more efficient use of advice on experimental design and analysis of the resources (equipment, finances, personnel). All of data. Some laboratory procedures have been in- these efforts will help broaden the forest resource cluded to serve as controls for experiments database and increase our understanding of the conducted in the field. Laboratory experiments can multiplicity of factors involved in forest provide valuable data to supplement field measure- regeneration. ments because the results are generally reproducible We anticipate that methods documented in this and environmental variables can be controlled. handbook will be improved once they undergo Many terms are discussed in a comprehensive glos- more extensive field testing, and we invite com- sary, and the main subject areas have been indexed. ments about the information and methods The logistics of field research are difficult enough suggested here, and about your own field experi- in their own right. We hope the information con- ences. Please direct your suggestions to the senior tained in this handbook will help those author at the address inside the front cover.

viii ACKNOWLEDGEMENTS

Very special thanks are due to Dr. John Zasada, Edith Camm, Ph.D., University College Fraser USDA Forest Service, North Central Experiment Valley, Abbotsford, B.C. (Glossary terms). Station, Rhinelander, Wisconsin, who reviewed the Andrea Eastham, M.Sc., Industrial Forest Service, entire handbook twice. From the time John reviewed Regeneration and Research Specialist, Industrial the first draft, his interest, enthusiasm, and encour- Forestry Service Ltd.,  Fifth Ave., Prince George, agement have helped immeasurably to sustain our B.C. (Section : Silvicultural Practices and Tree Seed own commitment to the project. His deep under- Biology). standing of the subject matter emanating from his D. George W. Edwards, Ph.D., Canadian Forest years of experience in the field has substantially Service, Pacific Forestry Centre, Victoria, B.C. broadened the context and increased the value of (Section : Seed Quality and Viability). the field manual. We thank him, too, for insisting David F. Greene, Ph.D., Departments of Geogra- that we include more coverage on hardwood species, phy and Biology, Concordia University, Montreal, forcing us to go beyond the traditional tendency to Quebec (Section : Seed Dispersal). regard conifers as the only trees in the forest. Robert (Bob) Karrfalt, Ph.D., USDA Forest Serv- Next, we want to thank Fran Aitkens, who has ice, National Tree Seed Laboratory, Dry Branch, given us an understanding of the importance of an Georgia (Section : Natural Seed Production and editor. It’s not an easy task to try to meld all the Section : Seed Quality and Viability). styles and approaches of a group of authors with Gina Mohammed, Ph.D., Ontario Ministry of very different backgrounds. Fran’s thoroughness and Natural Resources, Ontario Forest Research Insti- attention to detail, understanding of the technical tute, Sault Ste. Marie, Ontario (Section : Seed aspects of the text, and ability to craft disparate text Quality and Viability). into a cohesive manual have vastly improved the Peter Ott, M.Sc., B.C. Ministry of Forests, material she was given by the authors. Readers will Research Branch, Victoria, B.C. (Biometrics in all also appreciate the clarity and organizational struc- sections). ture she has introduced into the manual. George Powell, M.Sc., Agriculture Canada Performing a technical review is not an especially Research Station (Range), Kamloops, B.C. rewarding task, but it is essential to the review (Section : Seed Banks). process and to gain the benefits of a range of Michael Stoehr, Ph.D., B.C. Ministry of Forests, perspectives. Thus, we extend our sincere apprecia- Glyn Road Research Station, Victoria, B.C. tion to the reviewers of the various drafts of the (Section : Natural Seed Production and Section : manuscript: Silvicultural Practices and Tree Seed Biology). William Archibold, Ph.D., Department of Geog- Tom Sullivan, Ph.D., Applied Mammal Research, raphy, University of Saskatchewan, Saskatoon, Sask. Summerland, B.C. (Section : Seed Predation). (Section : Seed Banks). The following, all of the B.C. Ministry of Forests, Wendy Bergerud, M.Sc., B.C. Ministry of Forests, reviewed and commented on administrative details Research Branch, Victoria, B.C. (Biometrics in all associated with Section  Planning Tree Seed Research sections). in the Field: Brian Barber, Seed Policy Officer, Forest

ix Practices Branch, Victoria; Ken Bowen, Special Pro- information on site index; Dr. John Owens, Univer- jects and Boundaries Supervisor, Resource Tenures sity of Victoria, for the pollen micrographs (Figure and Engineering Branch, Victoria; Dave Cooper- .); and Heather Rooke, Tree Seed Centre, Surrey, smith, Research Silviculturist, Prince George Forest for valuable information on cone and seed charac- Region, Prince George; Ann Cummings, Records teristics of B.C. conifers. Management Analyst, Technical and Administration Branch, Victoria; Brian D’Anjou, Research Silvi- Carole L. Leadem culturist, Vancouver Forest Region, Nanaimo; Sharon L. Gillies Dr. Suzanne Simard, Research Silviculturist, H. Karen Yearsley Kamloops Forest Region, Kamloops; and Alan Vyse, Vera Sit Research Group Leader, Kamloops Forest Region, David L. Spittlehouse Kamloops. Philip J. Burton In addition, we thank Dave Kolotelo, Tree Seed Centre, Surrey, for the germination value data Victoria, B.C. (Table .); Gordon Nigh, Research Branch, for September 

x CONTENTS

credits ...... iii introduction ...... v organization of the handbook ...... vii acknowledgements ...... ix section 1 planning tree seed research in the field ......  . Overview ......  . General Structure of Successful Field Studies ......  . Designing a Field Study ......  .. Formulating the hypothesis ......  .. Stating the objectives ......  .. Selecting the factors to study ......  .. Selecting the methods ......  .. Setting the time frame and determining a schedule ......  .. Choosing the test conditions ......  . Experimental Design ......  .. Basic concepts ......  .. Determining the sample size ......  . Data Management ......  .. Establishing a coding scheme ......  .. Creating a permanent file ......  .. Preparing to collect data and samples ......  .. Collecting the samples and recording the data ......  .. Reporting ......  . Selecting and Describing the Study Site ......  .. Selecting the study site ......  .. Deciding on temporary or permanent plots ......  .. Determining size and shape of the plots ......  .. Installing, marking, and relocating the plots ......  .. Describing the site ......  .. Site index ......  . Analyzing and Interpreting the Data ......  . Administration of the Research Site ......  .. Obtaining site approvals ......  .. Registering field installations ......  .. Security ......  .. Safety ......  .. Using registered seeds ......  .. Making seed collections ...... 

xi . Ecosystem Management ......  . Summary ...... 

section 2 designing an environmental monitoring program ......  . Background ......  . Designing an Environmental Monitoring Program ......  . Methods for Measuring Environmental Factors ......  .. Soil temperature ......  .. Soil moisture ......  .. Solar radiation (light) ......  .. Wind speed and wind direction ......  .. Precipitation ......  .. Air temperature and humidity ......  .. Plant temperature ......  .. Canopy cover ......  .. Soil variables ...... 

section 3 natural seed production ......  . Background ......  .. Collecting stand and study plot information ......  .. Determining sample size ......  . Predicting Natural Seed Yields ......  .. Correlation with weather variables ......  .. Correlation with aspect and slope ......  .. Correlation with crown size and crown class ......  .. Sampling methods using bud counts ......  .. Scales for rating cone crops ......  .. Monitoring the seed crop ......  . Determining Fruit and Seed Maturity and Quality ......  .. Description of conifer and hardwood fruits ......  .. Assessing embryo development ......  .. Assessing seed colour ......  .. Measuring cone and seed dimensions ......  .. Estimating seed weight and volume ......  . Collecting and Processing Seeds ......  .. Conifer seeds ......  .. Hardwood seeds ......  . Assessing Factors that Reduce Seed Yields ......  .. Assessing serotiny ......  .. Assessing predation ......  .. Using X-ray analysis to determine causes of loss ......  . Experimental Design ......  .. Estimation studies ......  .. Modelling studies ......  .. Comparative studies ......  . Data Analysis ......  .. Estimation studies ......  .. Modelling studies ......  .. Comparative studies ......  . Seed Production Case Studies ...... 

xii section 4 seed dispersal ......  . Background ......  .. Why study seed dispersal? ......  .. Mechanisms of seed dispersal ......  .. Timing of seed release ......  .. Dispersal distance ......  .. Quantity and quality of dispersed seeds ......  .. Climatic conditions ......  .. Dispersal patterns ......  .. Dynamics of seedfall ......  .. Approaches to Studying Seed Dispersal ......  . Measurements and Methods ......  .. Basic considerations ......  .. Seed trap design ......  . Experimental and Sampling Design ......  .. Estimating seed rain density ......  . Data Analysis ......  .. Descriptive analysis ......  .. Comparative analysis ......  .. Regression analysis ......  .. Spatial analysis ......  .. Mechanistic modelling ......  section 5 seed predation ......  . Background ......  . Seed Predators ......  . Approaches to Studying Seed Predation ......  .. Natural seed crops versus artificially introduced seeds ......  .. Predation on natural seed crops ......  .. Predation on artificially introduced seeds ......  .. Quantifying seed predation ......  . Methods, Techniques, and Equipment ......  .. Distributing seeds ......  .. Excluding seed predators ......  .. Marking and recovering seeds ......  . Data Analysis ......  section 6 seed banks ......  . Background ......  . Approaches to Studying Soil Seed Banks ......  .. Seed separation versus direct counts ......  .. Assessing vertical distribution ......  .. Monitoring germination in the field ......  .. Seed burial experiments ......  . Methods, Techniques, and Equipment ......  .. Collecting and preparing soil samples ......  .. Seed separation and direct counts ......  .. Germinating sseds in samples ......  .. Monitoring germination in the field ......  .. Seed burial experiments ...... 

xiii . Experimental and Sampling Designs ......  .. Seed bank inventory studies ......  .. Comparison studies ......  . Data Analysis ...... 

section 7 seed quality and viability ......  .. Factors Affecting Seed Biology ......  .. Factors affecting dormancy and emergence ......  .. Factors affecting germination ......  . Seed Testing in the Laboratory ......  .. Sampling methods ......  .. Seed purity, seed weight, and moisture content ......  .. Preparing seeds for testing ......  .. Dormancy-breaking procedures ......  .. Laboratory germination tests ......  ... Quick tests and other viability tests ......  . Field Tests of Tree Seed Germination ......  .. Experimental design and analysis ......  .. Delimiting the site ......  .. Excluding other seeds ......  .. Preparing the seeds ......  .. Excluding predators ......  .. Monitoring germinants ......  . Experimental Design for Germination Studies ......  .. Experimental factors ......  .. Experimental designs ......  .. Replication and randomization ......  . Data Analysis in Germination Studies ......  .. anova ......  .. Categorical data analysis ......  .. Regression ...... 

section 8 silvicultural practices and tree seed biology ......  . Background ......  .. Principles of forest stand manipulation ......  .. Standard silvicultural practices ......  . Effects of Canopy Manipulation ......  .. Light ......  .. Temperature ......  .. Moisture ......  .. Suggested questions and approaches ......  . Effects of Seedbed Manipulation ......  .. Seedbed preferences ......  .. Site preparation ......  .. Suggested questions for seedbed studies ......  .. Methods for seedbed research ......  . Combined Studies ......  . Summary ...... 

index ...... 

xiv appendices

a Tree Species Occurring in British Columbia ...... 

b Conversion Factors ...... 

c Resources for Tree Seed Studies ......  glossary ......  references cited ......  figures

. Framework for evaluating the seed reproduction process in boreal forest trees ...... 

. Maps giving details of the research site should be included in the permanent file ......  . Sustainability can only be achieved when the needs of society and the potential capacity of the earth we live in overlap ...... 

. Climate station with wind direction and wind speed sensor, a rain gauge, and tower with solar radiation, air temperature, and humidity sensors ......  . Electronic datalogger used to monitor and capture data from a series of environmental sensors ......  . Automatic tram system that moves back and forth over a  m span to determine the variation in short- and longwave radiation, and surface and air temperature under a forest canopy ......  . Influence of forest canopy on the intensity and spectral distribution of solar radiation reaching the forest floor ......  . A fine wire thermocouple is used to measure the temperature inside the leader of a young spruce tree ...... 

. Different spatial arrangements comprising % canopy cover ...... 

. Typical development and maturation cycles of British Columbia conifer seeds ...... 

. Climatic conditions required for cone crop production in Douglas-fir ......  . Percentages of black spruce trees (concentric circles) – years old from seed, growing on slight (–%) slopes and on various aspects ......  . Position of measurement for trunk diameters, the diameter of the base of a branch, and main axes for estimating of the number of cones ...... 

. Scanning electron micrographs showing whole pollen and details of the exine ......  . The oval, raised cone scars of Pinus albicaulis can be counted and aged by the nearby annual bud scars on twigs ...... 

. Longitudinal and transverse sectioning of cones ...... 

. Anatomy of a mature Douglas-fir seed ...... 

. Tree seed anatomy (longitudinal sections) ...... 

. Outline drawing of a typical seed of Thuja occidentalis ...... 

. Salix capsules at various stages of opening and the dispersal unit at various stages ...... 

. Partial life table for  jack pine conelet crop, Oneida County, Wisconsin ...... 

xv . X-rays of tree seeds ...... 

. Correlation coefficients for hypothetical relationships ...... 

. A descriptive model of eastern redcedar (Juniperus virginiana L.) cone-crop dispersal from June through May of the following year ...... 

. Seed dispersal curves for nine conifers of the Inland Mountain West ...... 

. Examples of dispersal mechanisms of winged seeds ...... 

. Seed trap designs ...... 

. Schematic of the distribution of seed-traps placed around a point-source ...... 

. Recommended seed-trap layouts at a forest edge for an area source ...... 

. Small mammal exclosures ...... 

. Using a soil auger to remove a soil core for seed bank studies ...... 

. Method for cutting a square forest floor sample ...... 

. Preparing square soil samples for greenhouse germination ...... 

. Dividing a soil core into layers ...... 

. Soil samples in the greenhouse ...... 

  . Absorption of far-red light converts the pigment phytochromefar-red (usually the  active form) back to phytochromered (the inactive form) ......

. Sampling seeds by hand (a) and with a grid (b) ......  . Germination of (a) lodgepole pine, (b) Sitka spruce, and (c) Douglas-fir at different temperatures and after stratification for , , , and  weeks ...... 

. Stages of germinant development in hypogeal and epigeal germination ......  . Stages of germination for Populus seeds, showing time period in which each stage usually occurs ...... 

. Cutting diagram for the tetrazolium test ...... 

. A recommended frame design for delimiting field germination plots ...... 

. Layouts for one factor with two levels ...... 

. The two stages of a split-plot design ...... 

. Effective natural regeneration depends on an adequate seed supply, a suitable seedbed, and an appropriate environment ......  . Illustration of stand structure resulting from five different silvicultural systems used in the Lucille Mountain Project in the Engelmann Spruce–Subalpine Fir () biogeoclimatic zone, Prince George Forest Region, British Columbia ......  . Measured levels of photosynthetic active radiation (par) available to seedlings above and below the shrub layer in various partial cut systems at the Lucille Mountain Project, Prince George Forest Region, British Columbia ...... 

xvi . Growing-season soil temperatures (at  cm depth, -year means on a north-facing slope) in clearcut, small patch cut, and group selection treatments ......  . Number of subalpine fir and Engelmann spruce germinants per hectare within three silvicultural treatments at the Lucille Mountain Project, Prince George Forest Region, British Columbia ......  tables

. Installing and marking the research plots ...... 

. Seed production characteristics of hardwoods native to British Columbia ...... 

. Seed production characteristics of conifers native to British Columbia ...... 

. Cone crop rating based on the relative number of cones on the trees ...... 

. Rating of seed crops by number of filled seeds per hectare ...... 

. Seed-bearing structures of trees occurring in British Columbia ...... 

. Seed sizes of tree species occurring in British Columbia ...... 

. Seed dispersal mechanisms of winged seeds ......  . Mean terminal velocities reported for seeds (with seed wings attached) of some British Columbia tree species ......  . Split-plot-in-time analysis of variance (anova) table for a hierarchical sampling design ...... 

. Summary of exclosure choices for seed predators ...... 

. Two-dimensional contingency table for analyzing seed losses to two predators ......  . Comparison of methods for calculating proportion of seeds lost to predation over time ...... 

. Moisture content guidelines for orthodox tree seeds ...... 

. Classification of mould infestation in seeds ...... 

. Stratification and incubation conditions for British Columbia conifer seeds ...... 

. Stratification and incubation conditions for British Columbia hardwood seeds ...... 

. Summary of stratification methods ...... 

. Dormancy release treatments for tree seeds ...... 

. Germination values for British Columbia conifers ...... 

. Comparative seedbed suitability of some northwestern tree species ......  . (a) The relative abundance of seedbed substrates in an interior Douglas-fir stand (b) Expected and observed seedling association with four forest floor substrates in an interior Douglas-fir stand ...... 

xvii SECTION 1 PLANNING TREE SEED RESEARCH IN THE FIELD

The road to chaos is paved with good assumptions. (Anon.)

It is much less expensive to learn from other people’s mistakes than your own. (McRae and Ryan )

. Overview . General Structure of Successful Field Studies

Forests cover only about % of the earth’s surface, A survey of long-term forest research programs con- yet they account for nearly half of its net primary ducted at many locations throughout the world by productivity, and about % of the net productivity Powers and Van Cleve () stressed the importance occurring on land (Whittaker and Likens ). For- of planning, commitment, and focus. They con- est ecosystems are complex and diverse, and develop cluded that successful long-term experiments shared relatively slowly. Thus, forest ecosystem research eight essential components. Not all field studies are projects may require years to produce meaningful conducted over long periods, nonetheless, consid- findings that can be widely applied. eration of the following principles is instructive to The duration of a study depends entirely on what anyone contemplating field research, regardless of you want to find out. Many of the biological, physical, duration. and chemical phenomena associated with forest eco- systems can be studied over relatively brief periods . Sustained commitment on temporary field plots or in the laboratory. To Fluctuations in philosophy, politics, and funding are document and compare natural processes, short-term the surest way to dampen scientific spirit and inspira- descriptive and baseline studies are essential. Short- tion. Field studies, once established, must have a fair term studies also are important to establish the certainty of continued support, at least to the level immediate impact of processes on a system, even required to maintain research sites and to collect though they are prone to being confounded by core data. This support should be free from political environmental fluctuations such as climate. interference. To enlist this level of commitment, On the other hand, many biological phenomena, researchers should present their arguments based such as plant succession, occur on time scales of primarily on the benefits that can be derived from decades or centuries. Long-term studies allow us investigating socially relevant issues (e.g., sustaining to evaluate interactions among the various factors wood production, providing clean water, protecting controlling ecosystem function that, on a short time soils). Proposals that are couched in terms of “under- scale, might seem inconsequential. Forest ecosystems standing how forest ecosystems work” are far less usually require many years for the effects of per- likely to be granted support by funding administrators. turbations to subside and for long-term trends to appear. This is especially true for communities not . Long-term dedication of a site in equilibrium, such as those recovering from fire Plots maintained after the original questions have or harvesting. been answered can continue to have demonstration

section 1 planning tree seed research in the field 1 value for professionals and the public, and can pay in continuity and maintaining the research site, substantial dividends well beyond the life of the interdisciplinary field studies inherently promote the original study. Again, the chances of having land stability of long-term projects. However, interdiscipli- dedicated for the site will be enhanced if the research nary studies only work if there is strong central has a central, timeless theme. Support is more likely planning and coordination. to continue if research results are disseminated rap- idly to administrators and land managers. . Extension of results Research must be designed to make data as portable . A guiding paradigm as possible so that results can be generalized to a A central focus is necessary to provide structure and variety of species, soils, and forest types. Research will maintain research objectives. As long-term objectives have the highest value if results can be incorporated may grow hazy with time and personnel changes, into a network of coordinated, but geographically periodic reference to the guiding paradigm will help separated, studies. Experiments should be sufficiently to refocus the research. comparable so that databases can be shared, and each research site should be instrumented so that a . A central hypothesis baseline of climatological data can be established. A clear statement of the principal scientific question that the research is designed to answer helps to clarify . Low red tape the research direction and stimulate development of Maintain the least amount of bureaucratic structure the experimental approaches. The central hypothesis needed to prevent chaos. Initially, a board of senior is tested through a number of individual studies with scientists from a variety of disciplines should review definite life spans that terminate once a particular all research proposals. Later, a research coordinator question has been addressed. can review projects to ensure that one study does not interfere with another, and that all collaborators are . Large plots and replication kept abreast of the overall research program. Depen- Plots should be large enough to simulate natural eco- ding on the size of the research site, a site manager system conditions as closely as possible. Large plots may be needed to facilitate day-to-day (or seasonal) not only minimize edge effects, but also increase the scheduling. flexibility of future studies on research sites. Options might include retaining extra control plots that could . Designing a Field Study later be converted to secondary treatments, or creat- ing split-plots for treatments supplemental to the Careful initial planning and organization are critical original design. Large plots facilitate replication of to the success of any field study. Considering the com- treatments, which is essential for statistical analysis plexity, the expense, and the duration of many field and setting confidence intervals. studies, the consequences of poor planning can be great. Most studies consist of three stages—planning . Interdisciplinary approach to research (Stage I), data gathering (Stage II), and data analysis Field installations should be made available to all and interpretation (Stage III). However, the frame- research collaborators, regardless of affiliation or work for all three stages is constructed during the specialty. This will attract excellent scientists and planning stage. The planning process can be articu- promote openness and synergy. Studies that attract lated as a series of steps that provide answers to the a broad array of scientific interests result in much questions: why, what, how, when, where, how much, greater understanding than can be achieved through and so what. isolated, independent efforts. In addition, program scientists benefit from exchanging ideas, cooperating .. Formulating the hypothesis on experimental work, and collaborating on profes- The first step before undertaking any field study is to sional papers. Because collaborators have an interest formulate a clear statement of the principal scientific

2 field studies of seed biology question or central hypothesis that the research is germination, the most suitable variables to study designed to answer—the why of the experiment. A would be (micro)climate, substrate, and species. research plan with a clear statement of the problem Treatments should be chosen to reflect major helps to identify the research direction and provide changes in ecosystem function. Viewing ecosystems valuable guidance if the project should run into under extreme conditions is most likely to reveal difficulty (such as loss of support, changes in per- how various ecosystem components function and sonnel, or environmental disaster). to demonstrate the capacity of these components to recover from change. Studies will have greater value .. Stating the objectives if the treatments have a generally continuous pattern Once the principal scientific problem has been (i.e., increasing or decreasing in size). Data obtained elucidated, research objectives provide the necessary from such treatments lend themselves to predictive structure for planning and executing the project. Ob- regression analysis. Changes in responses can then be jectives are succinct summary statements of what the correlated with changes in the magnitude of specific research is trying to achieve. Keeping research objec- factors, and results can be more readily extrapolated tives in focus during all stages of planning will stimu- to similar sites. Choosing treatments that span and late development of experimental approaches, guide extend slightly beyond the full range of expected re- complete and efficient collection of data, and keep sponses helps to define the end points and establish the research on track. the limits of the system under study. Particularly in long-term studies, it is advisable .. Selecting the factors to study to retain some flexibility in the original design by The factors to be studied—another aspect of what— incorporating ways the experiment can be changed are usually specifically identified in the statement if future circumstances should require it (Leigh et al. of objectives. The factors chosen will depend upon ). To ensure the longevity of the field site, it is whether the study is primarily descriptive or experi- best if only minimum changes are made to the treat- mental in nature. In experimental studies, factors are ments. Changing the experiment to obtain more the vehicles through which the objectives are information in the short term generally results in achieved, and they are generally identified as treat- sacrificing the longevity of the treatments. It is some- ments. Factors may consist of one or several levels. times advantageous to incorporate innovations in For example, suppose your objective is to determine forest management practices into the study, but this if light affects the survival of lodgepole pine should be done only if the major objectives can be germinants on open harvested sites. To investigate retained. Another possibility is to modify the original this objective experimentally, you would identify light objectives and continue the experiment in a different as the factor to be tested. You might also want to form, but again longevity will be lost. A final, but more specifically compare how different light levels generally less desirable option, is to set aside the site affect seedling survival; then you would expand the indefinitely or reserve it for future use. light treatment to include several levels, such as full sun, partial shade, and full shade. .. Selecting the methods Constructing a schematic diagram of the bio- Methods can be considered the how of experimental logical cycle (or other process) is an effective way studies. The techniques chosen for the study will to identify what variables affect the process being be governed by the study objectives, and the most investigated. Diagrams help to clarify relationships effective means of achieving those goals. Usually, and suggest the most appropriate factors to study more than one method will achieve a particular (Figure .). For example, if you want to study initia- purpose, and for this reason a variety of methods for tion of reproductive buds, you will want to include field studies of tree seeds has been included in this climate, plant condition, and resource availability as handbook. Ultimately, the choice of the most suitable major factors in the experiment. On the other hand, technique will depend on the available resources. if you want to examine the factors affecting field In most cases, the final decision will be based on

section 1 planning tree seed research in the field 3  . Framework for evaluating the seed reproduction process in boreal forest trees (adapted from Zasada et al. 1992). A schematic diagram can help to clarify processes and suggest factors for the study. balancing the trade-offs between the detail desired process be automated? The choice may also be affec- and the constraints of time and money. ted by the logistics of the experimental site—certain Most research projects will employ a variety of techniques may not be suitable for field use. For ex- methods. Often the distinction between different ample, precise and automated methods may be ideal methods is determined more by the purpose than by for making a particular measurement, but the instru- the type and intensity of the monitoring. Note that ments may not be robust enough for field conditions different objectives do not necessarily require distinct or may require an external source of reliable power. and independent data collection efforts. There may After choosing the methods, it should be deter- be some overlap in the data needs. As long as the re- mined whether the research data are continuous or search objectives are kept clearly in mind, taking categorical. The term continuous implies that the advantage of this overlap can result in substantial measurements, in theory, belong to a numerical scale cost savings. consisting of an infinite number of possible values. Answers to the following questions will help in However, sometimes you need to measure a quality choosing appropriate methods: What is the primary or condition that cannot be expressed on a continu- purpose of the measurement? How well does the chosen ous scale. This discrete, noncontinuous type of data method quantify the factor or characterize the inten- is called categorical because it generally consists of sity of the response? How precise or accurate is the the number of observations falling into prespecified method? How many measurements are required? If classifications, groups, or categories (e.g., number of many or frequent measurements are required, can the seeds that have or have not germinated).

4 field studies of seed biology Continuous data can usually be analyzed using monitoring, effectiveness monitoring, project moni- parametric methods such as anova, regression, and toring, validation monitoring, and compliance manova (Sections ., ..). However, nonparametric monitoring. analysis is sometimes required for continuous data if Visual data are another significant source of primary the assumptions of anova, for example, cannot be scientific information, although they are not gener- met. Categorical data, on the other hand, are usually ally considered as data. Visual representations may analyzed using methods such as contingency tables be the most effective way to present information that and log linear models (Sections ., ..) although otherwise would be too unwieldy or difficult to alternative nonparametric techniques are available, understand (e.g., site maps or structural diagrams). if necessary. Some information can only be captured visually (e.g., Data can also be categorized by the type of meas- seed X-rays or photomicrographs of plant structure). urement used for collection. Determining the type Although not quantitative, visual data represent an of measurement needed helps to identify the type important source of research information, and a of information required, the frequency of data valuable means of portraying certain characteristics. collection, and the most suitable means of analysis. Unfortunately, visual data are underutilized in most Types of measurement include assessment, inventory, research studies. monitoring, and visual data. An assessment is an estimation or evaluation of .. Setting the time frame and determining the significance, importance, or value of a quality or a schedule character. It generally implies a subjective judgement Before starting the study, a schedule should be pre- (e.g., maturity) to determine placement in a class. pared to outline the temporal distribution of the The classification scheme may be based on some ar- major components of the study. This is the when of bitrary characteristic or a ranked order. Assessment experimental studies. Many field studies are short in data are usually nonparametric. duration, but some studies can be very lengthy, such An inventory is an itemized list or catalog that as the ecological studies of the Carnation Creek wa- may or may not be organized into groups. Usually tershed on Vancouver Island, which have been under the number of items in a group are simply counted, way for over  years. with no additional judgement or interpretation. The experimental design and type of data analysis For example, for an inventory of seeds in a seed will direct how often to collect the data (e.g., daily, bank, a count is made of the number of seeds by weekly, monthly), but the timing of treatments must species present in the soil. An inventory is usually also be taken into consideration when designing field a one-time measurement, but it can be repeated studies. Treatments may be applied only once, repeated periodically (e.g., annually). Greater use often can at fixed periods (e.g., annually), or even rotated. The be made of inventory data if the samples are strati- timing will also depend on what type of information fied in some way, for example, by making separate is required—whether you are interested in the direct seed counts at various depths of a soil core, rather effects in the year the treatment is applied, the re- than performing a single count of the total core. De- sidual (or carry-over) effects in subsequent years, or pending on the manner in which samples are taken, the cumulative effects of repeated treatments. inventory data can be parametric or nonparametric. The length of the study periods should be clearly The term monitoring is used to describe a series of defined, especially when planning a long-term study. observations made over time. The repetition of meas- The most suitable period length is defined by how urements to detect change over time is the quality long plot management can be kept constant. Period that distinguishes monitoring from the related proc- length might also be governed by the time when the esses of inventory and assessment. The data obtained first full assessment can be made (e.g., at the end of can be parametric or nonparametric. MacDonald et al. the first growing season), or when treatment differ- (), in compiling guidelines for monitoring water ences might first be discernible. quality, recognized seven types of monitoring: base- In some instances, period length may be used to line monitoring, trend monitoring, implementation apportion temporal variation (McRae and Ryan ).

section 1 planning tree seed research in the field 5 In the same way that blocks are used to control .. Basic concepts spatial variation among plots within a site, changes It is assumed that readers of this handbook have over time can be partitioned into periods. Although some knowledge of statistics, and will know where to period lengths may sometimes differ because of obtain assistance for particular statistical problems. operational constraints, analysis is simpler when The statistical discussions included in various chap- all plots have study periods of the same length. ters are intended only to provide general background on important aspects of experimental design and .. Choosing the test conditions data analysis, and to raise awareness of some poten- The next step is to determine where the factors will be tial pitfalls or problems that may be encountered in tested. Depending on the research objectives, the test specific topic areas. Discussions relating to some conditions are sometimes considered to be a factor common statistical methods can be found in the fol- of the experiment. If this is the case, you should lowing sections: summary statistics (Sections .., choose test sites or conditions that follow some sort .); anova (Sections ., .., ., .); regression of progression or gradient (e.g., small to large open- (Sections .., ., .); correlation (Section .); ings, low to high elevation). This will allow the results and chi-square (Sections ., ..). to be more readily generalized to other sites (if the Careful study of the proposed designs can be in- requisite experimental design criteria have been met). valuable during the planning stage (McRae and Ryan Most details relating to test conditions will be ). Trial analyses will demonstrate whether the specific to the study and what you are trying to ascer- contrasts of interest can be estimated, and will point tain. For further details refer to the section of interest: out deficiencies in the design and analysis methods. seed production (Section ), dispersal (Section ), A postmortem of similar experiments often provides predation (Section ), germination of seed banks data sets and estimates of experimental errors that (Section ), laboratory and field germination tests can be used to evaluate the proposed design. As in (Section ), and silvicultural practices (Section ). the actual experiment, there should be sufficient rep- lication to achieve the degree of precision required . Experimental Design to detect the treatment differences. If trial analyses reveal it is unlikely that differences will be found, the study may not warrant the investment. This is If you don’t deal with each of these levels of variation, especially critical for long-term studies. your sampled population may not be representative The distribution of replicates in space and time of your target population, and in that case is the most critical element of experimental design. a statistician or a sharp lawyer can make you Randomization provides for estimates of the experi- and your data look pretty lame. mental errors, which should always be reported, (MacDonald and Stednick ) either as the standard error of the mean or the difference between means (McRae and Ryan ). Once the objectives are identified and the factors, meth- Replication is generally accomplished by applying ods, and test conditions are established, attention treatments to two or more plots within the site should be turned to experimental design. The experi- (often divided into blocks) and/or by repetitions mental design will prescribe how essential elements of of the experiments at other locations or times. data collection (Stage II) and data analysis and inter- The allocation of replicates will be largely a func- pretation (Stage III) are executed. Field studies require tion of the objectives and the expected variability substantial commitments of time, labour, money, (MacDonald and Stednick ). The more sources materials, and maintenance; inadequate attention of variability you address, the more reliable your re- to details such as experimental design and data man- sults will be. In general, it is not efficient to test for all agement can pose considerable risks to the resources sources of variability everywhere. However, if you do invested in the project. Losses due to errors in not repeat any of your measurements, you have an experimental design may severely damage a scientist’s unknown source of error that will weaken all your reputation and will reflect badly on collaborators. subsequent conclusions. Repeated measurements

6 field studies of seed biology can give a better estimate of a variable such as seed Stednick ). For example, in natural resource production or field germination, but when you are management, the significance level is typically set at replicating your measurements, you have to be clear α = ., meaning that there is a  in  chance that an about the level of variability with which you are deal- observed difference will be due to chance. A strong ing. Measurement variability is very different from level of significance combined with high variability the variability between experimental units (e.g., field means that usually you will not detect a statistically germination plots). significant change until damage to the resource has Each time you design a project, you need to identify occurred. However, given the high natural variability all these potential sources of variation, and deter- in natural systems, it may be better to use a less strin- mine how you want to deal with them. If you don’t gent significance level in exchange for a higher level deal with each source of variation, your sampled of resource protection. Another example is power, population may not be representative of your target which is usually designated as -β. When comparing population. Hurlburt () uses the term pseudo- two sample means, the quantity β is known as a replication to refer to the testing of treatment effects Type II error, which is the probability of incorrectly with an inappropriate error term for the hypotheses concluding that two populations are the same when being tested. You can think of it as a source of vari- in fact they are different. Again, if a resource is slow ability which is inherent in the data but cannot be to recover or is of high value, you probably want to defined because of the sampling strategy. In other increase the value of α. In natural resources manage- words, if you have some sources of error in the data ment α should probably be set at . (MacDonald that cannot be tested, then typically you are dealing and Stednick ). with pseudoreplication. Excellent discussions of pseudoreplication can be One of the examples Hurlburt () uses is a found in Stewart-Oaten et al. (), Bergerud (), study to determine the effect of water depth on the and MacDonald and Stednick (). Additional rate at which leaves rot in a lake. Although this is not discussion of randomization and replication in a seed-related example, it is worthwhile using here relation to field studies of tree seeds can be found because it is so clear and succinct. Four bags of leaves in Section ... were placed together at one location at a depth of Blocking of the plots reduces experimental error  m, and another four bags were placed together at by removing any gradient effects in site variation. another location at a depth of  m. After some time Choosing blocks that are arranged contrary to the the bags were retrieved, dried, and weighed. If there field gradient will increase the experimental error, was a significant difference between the two sets but this is difficult to avoid because the field gradient of bags, all that can be said is that there was some is often unknown. Appropriate variables on the site difference between the two locations. To make an (temperature, soil, moisture) can be measured and inference about the effect of a particular water depth the results used as a basis for blocking. If elevations on leaf rotting in this lake, the bags would have to be of the site vary considerably, blocking would most distributed at the same depth around the lake. To likely be parallel to contour lines, and not perpen- make a more general statement, replicated samples dicular. The effectiveness of a block arrangement would have to be distributed at different depths in can be assessed only after the experiment has been several lakes. The design depends on the question run. A good strategy is to have a robust blocking you want to answer, but placing all the bags in one structure that allows for an environmental gradient place is pseudoreplication. From Hurlburt’s point of in either or both directions in a rectangular site view, you must have replication on at least one level. layout (McRae and Ryan ). See also Section .. If you don’t have the ability to test for differences, it In forestry research the same unit or process is is not an experiment. usually measured on more than one occasion. For Often you need to make statistical compromises, example, in trials to compare several treatments, and if so, you should be explicit about the statistical data are typically collected before and after treat- trade-offs that you have made, rather than letting ments are applied. Such data tend to be serially them be set by neglect or default (MacDonald and correlated, or autocorrelated, which means that the

section 1 planning tree seed research in the field 7 most recent measurements are dependent on, or to dispersal, Section  for seed banks, and Section  for some extent predictable from, earlier observations. seed germination tests. A more general discussion of Because this violates the independence assumption various types of sampling can be found in Cochran on which many standard statistical methods are () or Thompson (). based, alternative methods are required for their analysis. Two broad classes of methods have been . Data Management developed for this purpose: repeated-measures analysis and time-series analysis. For additional Data management protocols should be established background and discussion of this topic, refer to when the study is initiated. The type of data, experi- Nemec (). mental design, and method of analysis will guide how Carry-over effects from previous treatments are the records and data are organized and managed. a hazard of long-term studies in which multiple This section provides a brief overview of the major treatments are applied. When analyzing by anova points of data management, as well as some special or multiple linear regression, estimates of the direct considerations required for long-term studies. effects of later treatments must be adjusted for any residual effects remaining from previous treatments .. Establishing a coding scheme (McRae and Ryan ). A consistent coding scheme should be established to correlate all data records with the research plots and .. Determining the sample size treatments. The coding scheme is best defined in a To ensure that enough samples are collected for a table assigning unique label codes to identify field study—the how much of the experiment—it is advis- plots, factor levels, treatments, and replications. The able to determine the appropriate sample size specific table should indicate the exact units in which the to the parameter being studied. data will be recorded (e.g., millimetres, kilograms, Sample sizes for each measurement must be deter- or watts per square metre). For categorical data mined independently, because variability may be (Section ..), a brief description should be given different for different characteristics. For example, of the significance of each classification code (e.g., in the sample size required for measuring cone charac- Section .., for cones,  = scales fully open;  = scales teristics may not be the same as that needed for seed partly open;  = scales completely closed). characteristics, even for the same species, because Allow for some flexibility in the coding scheme of differences in the variability of the data (Carlson so that labels can be added if new treatments are and Theroux ). Environmental changes may also incorporated, or if treatments change over time. result in year-to-year variations, but these differences If treatments are changed, ensure that the coding can sometimes be minimized by adjusting all values scheme is annotated to relate the new treatments to to be relative to those observed in a particular year the original treatments. (Ager and Stettler ). The same format should be established for field Sample size is usually determined by applying and computer records so that data can be easily statistical efficiency calculations to a preliminary accessed for future examination or analysis. Where set of measurements (Sokal and Rohlf ; Ager feasible, coordinate with other agencies or researchers and Stettler ). See also Stauffer (, ) for to use standard codes or data-entry protocols. This sample-size tables oriented to forestry applications. will facilitate exchange of data between programs. For In the absence of any other information, a sample example, the standard coding formats used for the size of  is often a good place to start (MacDonald biogeoclimatic ecosystem classification data should et al. ; MacDonald and Stednick ). be used for all site and vegetation data. Standard spe- The topic of sampling, both how to sample and cies names and codes for British Columbia can be how many samples to collect, is a critical aspect of found in both access . and excel . files at the all field studies of tree seeds. For more detailed B.C. Ministry of Forests Research Branch ftp site discussions relating to particular subject areas, see (see Appendix C) in the directory/pub/provspp. The Section  for seed production, Section  for seed files are regularly revised and updated. If you want

8 field studies of seed biology to collect vegetation data, then you should follow a) Describing Terrestrial and Aquatic Ecosystems in the Field (in preparation ) which will update Lutt- merding et al. (). This is also a useful reference for making site descriptions (see Section ..). A variety of computer data entry and reporting tools (e.g., venus) are also available (see Appendix C for more complete information).

.. Creating a permanent file The permanent file should include the initial plans and objectives and all parameters of the experiment. A statistical guide should be included in the perma- nent file giving full details of the experimental design(s), the proposed method of analysis, and all associated computer programs. Include the type and number of annual data sets, and a list of the different annual records. Special notes about the trial should be recorded and arranged by date or other logical se- quence. Include maps giving details of the research sites, the location of the plots, and the arrangement of the treatment blocks and replications (see Figure .). Provide room in the structure of the permanent file so that you can add data and field notes for the current year and update the parameters, if necessary. It is useful to link computer data files to a spreadsheet or graphics program to produce a series of graphs depicting the different responses over time for each treatment. Create a summary table of the cumulative effects for each variate, giving the relevant summary statistics. Plan the permanent file so that computer formats b) and files will remain compatible over changes in computers and software. To ensure efficient data entry, carefully design data sheets and format computer files. The spreadsheet software into which you plan to import your data should guide the data file format. For most data, a row/column format is best. If you are uncertain about the type of software that will be used, a simple ascii (text) file format is recommended. The permanent file should also contain detailed directions for finding the plots again after installation.  . Maps giving details of the research site should The importance of this step cannot be emphasized be included in the permanent file. (a) Loca- enough, especially if different people are resampling tions where western larch and subalpine larch the plots. Several scales of maps are needed to relo- are sympatric (Carlson and Theroux 1993). cate plots. Section .. includes a more detailed (b) Sketch of investigated stands of Pinus discussion of recording site parameters for relocation. sylvestris (Bergsten 1985).

section 1 planning tree seed research in the field 9 .. Preparing to collect data and samples . Selecting and Describing the Study Site Computer-generated administrative aids (labels, data sheets, random order lists) will simplify data and .. Selecting the study site sample collection. Organic tissue and soil samples An essential part of planning is selecting a suitable collected during the study must be properly coded site—the where of the experiment. Site selection and archived for analysis or future reference. Pre- should take into consideration practical aspects such printed labels simplify collection of samples in the as accessibility, the frequency of site visits, and how field and act as an additional check that coding seasonal changes may affect access and any perma- sequences are complete. Colours and symbols nently installed instrumentation. (e.g., stars, circles, triangles) used in addition to, As early as possible in the planning process, or instead of, numerical codes will help to reduce contact the local forest district or the forest manage- errors, which may result from performing repetitious ment office responsible for the proposed study area. tasks under arduous field conditions. Establishing a good relationship with those ulti- The sequence prescribed by the randomization mately responsible for the site can have unexpected scheme can be used to arrange labels, sample con- benefits and will also serve to promote your research tainers, and data sheets. If you have a large number amongst the forestry community. Local staff may be of items, it may be convenient to subdivide them able to suggest potential sites that meet your criteria into smaller groups (e.g., by plot number). and provide more detailed information if they know Permanent markers (stamped metal or plastic the objectives and the key factors you wish to study. are best) should be generated for all treatments, and From them you can gain considerable information securely attached to durable, highly visible posts or and knowledge about local forestry practices that other stationary devices in the field plots. For further will directly and indirectly influence your work or details, see Section .., Table .. research, now and in the future. Because they are located near the site, they may be able to maintain .. Collecting the samples and recording the data security or assist with site maintenance. In addition, Data can be recorded in the field using manual if local forestry staff know about your study, the records or hand-held dataloggers or other automatic chances of the trial being damaged by concurrent recording apparatus. Pre-labelled sheets can be used industrial or silvicultural activities is greatly reduced. for manual data entry, or datalogger files can be pre- The scientific criteria for selecting a site depend on programmed with plot, treatment, and sample codes. the goal of the study, but all critical site-related factors Refer to Spittlehouse () and Section . for addi- must be identified. For example, if the objective of tional information on using dataloggers in the field. the study is to determine the difference in seed pro- Transfer of data is now relatively easy using com- duction between north- and south-facing slopes, puters and computer interface devices, but all files then site selection will be dictated by the aspect must be regularly backed up to avoid loss of data. and grade of the slope. In other studies, slope and You should have spare power sources, in case primary aspect would not be primary factors. If the goal is to equipment fails. describe seed production in mixed stands as opposed to uniform stands, then the primary selection criteria .. Reporting would be the species composition of the stands. For A complete analysis of the research and a summary long-term research on seed production in a natural report should be prepared annually or at the end of stand, it would be important to locate each site away each field season. This can be considered the so what from openings or roads. In this case, a fixed area on of the experiment—what do the results mean in the each site might be delineated in the centre of the greater scheme of things. Strive to disseminate as stand, with the trees surrounding the plot acting as a quickly as possible the interim results or updates at buffer zone to reduce edge effects. technical meetings, in short articles, or in newsletters. Site illustrations are useful in documenting the Prompt reporting will help maintain support while key elements of the field site (Figure .), and should the research is in progress. form part of the permanent file (Section ..). They

10 field studies of seed biology may also be used to find the plots again for repeated The choice of plot size often depends on stand measurements (Section ..). density and heterogeneity. To compensate for differ- ing stand densities or species mixes in a study, the .. Deciding on temporary or permanent plots plot size could vary to maintain a constant number A decision must be made whether to use temporary of trees or species types within each plot. See Smith sites (entirely new units are randomly selected for et al. () for an example of this approach. observation each time), or permanent sites (the Another approach for studying tree density effects same units are observed repeatedly over time). The is to use rectangles of fixed dimensions (width and choice of temporary or permanent plots depends length) for all sample areas. Tree density can be esti- on the degree of correlation you expect between the mated by dividing the number of healthy trees within initial and final plot values. If a high positive correla- a sample rectangle by the area of the rectangle. This tion is desired, permanent plots will generally give approach is preferable because the fixed dimensions better precision. If large-scale changes are expected provide consistent estimates of site variation across in the nature of the site, temporary plots should be all study areas, and ensure the validity of tests for tree used (Freese ). Sometimes a combination of density effects. temporary and permanent sites can be used—perma- nent (intensive) sites for detailed aspects of the study .. Installing, marking, and relocating the plots and temporary (extensive) sites for broadening the Once the experimental site has been selected, the number of samples and site types. plots must be identified and the boundaries clearly marked. Markings must be highly visible and dura- .. Determining size and shape of the plots ble. The choice of marking method will depend on a The size and shape of the plot depend on a number number of site factors such as the distance from the of factors, including the goal of the research, the cost road, steep terrain, annual snowpack, rocky ground, or time required for sampling, the required precision, or height of vegetation. The durability required of and the uniformity or heterogeneity of the area markers will depend on the amount of exposure to (Freese ). There are obvious trade-offs between the elements, the possibility of crushing or toppling plot size and homogeneity of samples. You want as by large animals (bears, moose, cows, humans), and large a sample size as possible, with good treated the total length of time the plot will be sampled. A buffers, but the larger the plot size, the more likely summary of important points for installing and you are to introduce heterogeneity (in soils, nutrient marking plots is given in Table .. regime, moisture regime, slope, etc.) into your plot After installation, the site location should be re- selection, thereby increasing the within-plot error corded in detail in the permanent file. This is an sources. To assess homogeneity, a full site description important step, and will prove particularly valuable of each plot is recommended. Plots should only be if the plots are resampled by different people or over accepted for inclusion in the study if the variation in many years. site type would not compromise the long-term re- sults. In general, moving more than one full site series • To locate the general vicinity of the site: Mark site (or other environmental gradient) within a plot is locations on topographic maps, forest cover maps, probably sufficient reason to abandon it. Larger plots airphotos, orthophotos, etc. Write out directions in- (more than  ×  m) should have multiple soil pits cluding distances (km) to each turning point and to ensure homogeneity. road names, etc., from likely starting points (towns). The duration of the study will also govern plot Use GPS locations if you can afford and have access size. If there is any possibility of continuing the study to this technology. for  years or more, consider the plot size carefully. For example, if your original objective was to study germ- • To locate the site from the point of access (e.g., road): ination and initial development, but later you decide Draw a site map of the plot(s) and surrounding area to extend the length of the study, you may be unable with local landmarks (e.g., roads, water, rocks, slopes, to do so if the plot size initially chosen was too small. directions, etc.). If there is more than one plot, ensure

section 1 planning tree seed research in the field 11 they are mapped in relation to each other as accu- (Agriculture Canada Expert Committee on Soil Sur- rately as possible, using compass bearings and dis- vey ), humus classification (Green et al. ), tance measurements. and for sites in British Columbia and some other areas, the appropriate biogeoclimatic ecosystem • To locate the plot(s): Flagging tape and painted classification (refer to the B.C. Ministry of Forests stakes will help you spot the plots once you are at the regional field guides listed in Appendix C). Topo- site (See Table .). Plots marked with metal stakes, graphic grid references are also useful to locate the pins, or tags can be relocated with a metal detector. general area of the site. Keeping the objectives of the study clearly in focus • Take photographs of the plot to have a visual record will help identify other factors that should be docu- of changes that occur and assist in relocating the mented in the site description because they might plots. Mark the photo points on your site map. affect the outcome of the study. For example, the percent cover of major non-tree species that com- .. Describing the site monly invade to sites following disturbance should General site characteristics should be described for be listed if their presence could influence the results a field site even though they may not be identified of your experiment. Soil profile details could be as the primary factors under investigation. A site included if the experiment would benefit from this description should include the slope, aspect, eleva- information. If the site has been harvested, the de- tion, longitude and latitude, soil classification gree of soil disturbance should be quantified and

 . Installing and marking the research plots

Stakes

Weight: This factor is critical (unless few stakes are needed) if the site is inaccessible and materials have to be carried a long way. In rocky ground, thinner stakes are easier to install. On steep slopes, stakes usually get pushed over in the winter, especially where there is a lot of snow and/or vegetation. Use strong, slim stakes and pound them far into the ground to reduce this problem.

Visibility: Stakes should be taller than the tallest understorey vegetation, but short enough that you can reach the top to pound it in. Allow enough length to compensate for the amount that is pounded into the ground. Ensure the stakes are clearly visible by painting the tops bright, contrasting colours (white, fluorescent pink, orange, or blue; not yellow, green, or dull or dark colours).

Wood: Pros: relatively lightweight; broader surface more visible when painted; easy to attach labels; moderate cost. Cons: bulky; eventually rot; can split and break; may be harder to pound into the ground; greater surface area, more easily pushed over by snow.

Steel Bend the ends of rebar stakes to prevent injury to people and animals. Pros: compact, long-lasting reinforcing (but rust); relatively cheap; easy to pound into even rocky ground; can be relocated with a metal bar (rebar): detector. Cons: heavy; not very visible even when painted, difficult when rusty; harder to attach labels; <1 cm diameter can bend fairly easily; larger diameters (>1 cm) are too heavy (except for short stakes). Aluminum: Use either conduit or Y-beam. Pros: lightweight; visible when unpainted; strong (doesn’t bend easily), won’t corrode, can engrave plot information directly on so don’t have to attach separate labels; can be relocated with a metal detector. Cons: expensive (3–4 times cost of rebar).

12 field studies of seed biology  . Continued

Installation

Pound stakes until they are as steady as possible. Carpenters’ hammers (unless they have metal handles) break too easily. A short-handled 2 lb. (0.9 kg) sledge hammer is ideal because the handle is stronger and the head has a broader surface area. If topofil (hipchain) is used to measure distances, remove the thread after measurement because it can entrap birds and other small animals and kill them.

To form Use one rope to measure the length of two sides of the plot (marking the middle), and a second square plots: rope for the diagonal length. A tape measure can be used instead if it is long enough. Install the first stake. Measure the distance to the diagonally opposite stake and install. Measure the length of a side towards a third corner from each of the first two stakes. Where these meet, install a third stake. Repeat the last two steps to locate the fourth stake. Ropes with loops on the ends (one the diagonal length and the other the length of two sides with the middle marked) or two flexible fibreglass tapes can be used to make the measurements.

For circular Install a single stake in the middle of the plot. Use a rope the length of the plot radius to measure plots: from the centre stake to the plot boundaries and mark with flagging tape.

Labelling

Labels are essential if there is more than one plot in the installation. Identify the plot with a number code and other pertinent information. (See also Section 1.5.1.)

Washers: Stamp large washers (3.5 cm diameter with 1.5 cm hole) with plot numbers using a die set, then slip them over the rebar stakes. Pros: easy to use. Cons: eventually rust so much you cannot read the numbers; work their way into the ground and must be excavated; can slip off and be lost if the stake is pushed over in the winter.

Aluminum Can be wired, nailed, or stapled onto wooden stakes, or folded around rebar stakes and stapled. sheets: Pros: easy to engrave; won’t corrode; easy to see; can attach them to the tops of stakes (no exca- vating); easy to mold to stake. Cons: easily ripped by animals or vegetation rubbing in winter, etc.; sharp edges unless each edge is bent. Plastic or Can be purchased from engineering or survey equipment suppliers either pre-numbered or blank. metal tags: Some suppliers will engrave custom numbers on blanks. Pre-numbered plastic livestock ear tags are also available from agricultural suppliers. Plastic tags come in different colours but may break or fade over time. Metal tags are more durable but less visible. Use coated wire to attach tags to plot stakes, trees, etc.

Flagging tape: If resampling frequently (e.g., every 1 or 2 years or less), use plastic flagging tape on stakes. Use the most durable winter-weight flagging; although more expensive, it lasts much longer. Use fluorescent pink, orange, etc. (same as for paint) for the best visibility. A long tail of tape moving in the wind will catch the eye better than many short pieces and wrap-arounds. Biodegradable tape is not recommended as it is almost impossible to see. If sampling is infrequent, don’t bother with flagging; it is not very durable, and animals chew on it. Rely on painted stakes, photos, and good site maps instead. Felt pen on flagging tape is OK for temporary labelling purposes (about one year), but not for long term.

section 1 planning tree seed research in the field 13 documented using the methods prescribed in the entered into the SiteTools software (available from Forest Practices Code Soil Conservation Surveys Guide- Research Branch, B.C. Ministry of Forests) to obtain book () (Appendix C). an estimate of site index. When a stand is present on the site, then the de- When good top height trees are not present on the scription should include an estimate of tree species site, the site index can be approximated through a site composition based on basal area. This can be done series–site index correlation, by which a site index is using variable-radius sample plots. The basal area estimated from the site series present. This method is factor of the prism or relaskop and species of “in” generally less accurate than the tree-based estimates, trees are used to estimate species-specific basal area and should not be used if good top height trees can (British Columbia Ministry of Forests ). Rela- be found. Site series–site index correlations are not skops are used most commonly, but a set of prisms yet widely available, but may be found in the Minis- works as well and is less expensive (about $ instead try of Forests field guides for Nelson (Braumandl and of $ for a relaskop). Curran ), Prince Rupert (Banner et al. ), and Environmental conditions, such as relative humid- Vancouver (Green and Klinka ), and in the scien- ity, rainfall, hourly temperature averages, and daily tific literature (Green et al. ; Klinka and Carter maximum and minimum temperatures, can be ; Carter and Klinka ; Wang et al. ; monitored using on-site dataloggers. Note that Kayahara et al. ; Wang et al. ). First approxi- climatic data collected from standard weather sta- mation provincial correlations are available in draft tions may not be sufficient to accurately document form (Meidinger and Martin []). factors that affect flowering, pollen dispersal, cone opening, and seed maturity at the stand level. . Analyzing and Interpreting the Data Weather variables monitored several metres above the ground may not reflect the conditions within the The great tragedy of Science: crown, or on the north and south sides of a tree. In the slaying of a beautiful hypothesis by an ugly fact. some cases it may be useful to establish correlations (T.H. Huxley) to capture these relationships. For further discussion on these and related topics, refer to Section ., Data management, and Section ., Designing an environ- For many researchers, the most enjoyable (and chal- mental monitoring program. lenging) part of a study occurs after the data have been acquired and entered into the database. At the .. Site index data analysis and interpretation stage (Stage III) the The site index is commonly used in forestry to meas- relevance of research results must be recognized and ure site productivity. The site index is the average articulated. If the project has been well planned height of top height trees (unsuppressed dominant (including site selection, experimental design, and or codominant trees) measured at breast height age analyses), this stage is usually straightforward. How- . The more productive the site, the higher the site ever, unexpected things happen, and you may need to index. Site index can be obtained in at least two ways: manipulate and analyze the data in ways not initially from tree measurements or from the site series. To planned. For example, look for confounding factors obtain accurate tree-based estimates of site index, that may be influencing your results, or try grouping good top height trees should be present in the plot. your data differently and reanalyzing. A description of what constitutes good top height The value of reanalysis is best demonstrated using trees can be found in Forest Productivity Councils an example (D. Coopersmith, pers. comm., ). It of British Columbia () and Soderberg and Nigh also demonstrates how a detailed site description can (). These publications also detail the sampling later be used for other purposes. protocol. The total height of the tree and its breast Recently, an analysis was performed on perma- height age are required to estimate site index. This nent sample plots at the Aleza Lake Research Forest information and the name of the tree species can be in north-central British Columbia (lat °'N, long

14 field studies of seed biology °'W). The stand had been logged in the s .. Obtaining site approvals using diameter-limit selection (all trees larger than For sites in British Columbia, researchers are respon-  cm were taken, and smaller trees were left). The sible for obtaining the B.C. Ministry of Forests site was a very productive moisture-receiving site at district manager’s agreement for the location and the base of a long slope. It was also micromounded, purpose of the research. The district manager will probably from previous windfall events in the stand. want to ensure that the project can be accommodated Some of the spruce were as old as  years, so it within the objectives of the management plan for the was probably  or more years since this site had site. Researchers must adhere to the Forest Practices been burned. An initial examination of the tree Code of British Columbia; for silvicultural system data showed that basal area and volumes had not and natural regeneration trials, this may require increased since the last evaluation in . This amendment of silviculture plans and cutting permits. was surprising because some very large spruce and Plans normally must be filed at least  year in advance subalpine fir appeared to be growing very vigorously with the district office. For trials not affecting silvi- on the site. culture prescriptions, the district manager must be A second analysis was performed. This time the notified a minimum of  months before the proposed trees were separated into two classes: those growing research work. in the wetter hollows of the micromounds (charac- Before beginning any study on developed lands, terized by Equisetum); and those growing on the check with the local offices of major utilities to ensure drier mounds of the site. The results of this reanalysis that site activities will not disrupt water, electrical, or were dramatic: all trees in the hollows showed little gas services in the area. Field personnel should know or no growth (in fact, large numbers had died since the name and telephone number of the appropriate the last evaluation), showing that trees on these utility companies to contact in case of emergency. microsites had not contributed anything to basal area and volume in the stand, while those on the higher .. Registering field installations microsites were still growing vigorously and adding Some regional offices maintain a list of the objectives significant additional growth. By not differentiating and locations of all known research plots within the between these two microsite types, much of the story region. Researchers establishing research plots or of these plots was lost in the “noise.” installations on provincial Crown land must convey Researchers should consult a statistician before the site location and other pertinent information embarking on any of the more elaborate statistical to the regional research manager, as well as to the analysis methods to ensure the proper application of relevant district managers. the techniques. A protocol for plot registration and map notation has now been established for all permanent sample . Administration of the Research Site plots (psp) by the B.C. Ministry of Forests for the forest inventory mapping system (famap). Map Field research must always be undertaken with the notations are made on all mylars furnished to the knowledge and approval of the land owner and the forest districts. When a district is proposing a stand local land manager. In British Columbia, researchers treatment, such as thinning or fertilization, all district wishing to locate research sites on provincial Crown mylars are checked for the area of the proposed treat- land must follow the regulations and guidelines set by ment. This is how the forest district avoids treating the B.C. Ministry of Forests and other agencies. The well sites, archaeological sites, and research plots. steps outlined in this section are specific for research There is a standard procedure for getting informa- sites in British Columbia, but are similar to require- tion entered on the map mylars, such as a harvesting ments in other areas. Ensure that you contact the tenure or an experimental project (ep), usually by agencies with jurisdiction over the area you have providing a sketch map and documentation. Note, chosen and that you know and follow the appro- however, that the researcher is responsible for keep- priate regulations. ing track of sub-eps in the permanent research file.

section 1 planning tree seed research in the field 15 .. Security unlikely to be discovered by passers-by. If working For security purposes, the location of research sites alone is necessary, you should know and closely ad- must be on file with the applicable district, regional, here to policy guidelines of the forest management and licensee offices. Installations that will be repeat- agency, your employer, and your local authority. edly visited (i.e., more than once) should be registered In British Columbia, the safety of crews working on forest cover maps as either a map notation (coded in the field is governed by Workers’ Compensation on the forest cover map to notify users that an activ- Board (wcb) regulations and Industrial Health and ity is occurring there) or a map reserve (to reserve Safety and Occupational First Aid Regulations. Ensure the site from harvest within a specified time period). that you are aware of and comply with all the require- Map notations or reserves can be critical in saving a ments—only a few are highlighted here. Regulations site from disturbance or inadvertent damage. stipulate that a Level I first aid kit and someone with Experimental plots benefit greatly from having appropriate first aid training must be on-site. Specific signs posted on the site. This will keep out most of written procedures for transporting injured workers the public. At the minimum, the sign should state must be developed and be present at the field site “Research Site, Do Not Disturb. If you would like before operations begin. more information, please contact the nearest district office.” Locate valuable equipment such as meteoro- .. Using registered seeds logical stations so they are not visible from the road; In British Columbia, only registered seeds may this will lessen the possibility of equipment being be used for reforestation on Crown land (Forest vandalized or used for target practice. Practices Code, Silviculture Practices Regulation, Sec. ()(a)). This regulation also applies to forestry .. Safety research trials if the seeds will be planted on Crown Ensure that you know the radio and check-in proto- land. Refer to the Forest Practices Code Guidebook: cols for the district you are working in (see B.C. Seed and Vegetative Material or consult with district Ministry of Forests Research Branch Operating staff for current seed use guidelines. The Seedling Policies and Procedures). Planning and Registry (spar) system can identify Use radio or cellphone communication when available registered seed sources. Contact the district possible. Radios are essential on active logging roads. office for assistance with spar and other Code-related The forest district office or logging company can matters. (See Appendix C for resources.) supply you with radio frequencies, but they are also usually posted at the beginning of the road. The .. Making seed collections appropriate frequencies can be programmed into If the seedlings resulting from individual seed collec- radios by staff of the B.C. Ministry of Forests Techni- tions will not be planted on Crown land, the use of cal and Administrative Services Branch (for ministry registered seeds is not required. However, if you plan employees) or at most radio shops. In the field, call to collect your own seeds, you must obtain a cone your location frequently and monitor the location of collection permit from the district office in which logging trucks so you can pull over well before you the collections will be made. B.C. Ministry of Forests meet them. Logging trucks always have the right-of- researchers are not required to have a permit, but way. Always drive with your headlights on when on they are still encouraged to inform district staff of logging roads. their intention to collect cones. Report your destination and return time before any Anyone making seed collections is responsible for field trip, and check in during the day so that someone rigorously adhering to all precautions restricting the knows where you are and your next stop. B.C. Minis- use of climbing gear and collection equipment. In try of Forests policy advises against working alone in British Columbia, aerial operations are subject to the field and strongly discourages the practice. This wcb regulations, the helicopter company must be policy relates specifically to situations in which an certified, and the pilots appropriately qualified for employee is working in a remote area off paved roads, making aerial collections. Refer to wcb regulations, and may be unable to call for help if injured, or is Eremko et al. (), and Camenzind () for

16 field studies of seed biology additional information on the safety aspects of tree understanding draw these circles closer together. seed collection in British Columbia. Opportunities for sustainability increase when we manage so that these spheres can overlap. . Ecosystem Management Information is a primary resource, and as re- searchers, it is our major contribution; the success From time to time, it is beneficial for researchers to of adaptive ecosystem management depends on the stand back and view their research in the larger con- generation and transfer of our scientific knowledge text in which studies are conducted. By viewing (Bormann, Brookes, et al. ). Monitoring and studies of tree seed biology in the broader perspective research must be integrated with decision-making of ecosystem management needs, research results can processes to continually improve the scientific basis have a greater impact and enhanced value to society. of ecosystem management (Jensen and Everett ). Ecosystem management is a scientifically based Thus, it is critical that we allocate our efforts to bridge land and resource management system that integrates this interface between science and management. In a ecological capabilities with social values and eco- topical article, Larsen et al. () defined  principles nomic relations to help sustain ecosystem integrity of ecosystem management which provide useful guid- and use over the long term. In recent years, the term ance to ecosystem researchers to make their research adaptive management has been used to describe a projects relevant to management needs for informa- modified approach to managing ecosystems. One of tion. Not all research projects will be able to strictly the main distinctions of adaptive management is that adhere to these principles, but they provide a useful it emphasizes learning through conscious experimen- reminder of context for natural resource studies. tation, monitoring, and adjustment (U.S. Dep. Agric. For. Serv. b). . Management and research must deal with large The goal of adaptive management is to create and landscapes. The cumulative effects of processes that maintain sustainable ecosystems. To achieve the goal typically function at smaller scales, such as stand- of sustainability requires that we integrate both the level silvicultural treatments, can be observed only if human societal and economic needs and ecological we step back to take a wide-angle view of the forest. processes. This concept may be visualized by viewing Some important processes, such as patterns of forest the needs of society and the earth’s ecological capacity distribution or natural disturbance, can be observed as separate spheres (Figure .). Knowledge and only at the landscape scale.

 . Sustainability can only be achieved when the needs of society and the potential capacity of the earth we live in overlap. Learning draws these circles closer together and increases our opportunities and options for sustainability. (Adapted from Bormann, Cunningham, et al. 1994.)

section 1 planning tree seed research in the field 17 . We must be concerned with long time frames. . Researchers must share sites so that they can Just as the extent, structure, and condition of today’s integrate their findings and investigate change in forests have been determined by harvesting practices each ecosystem component over many different that took place a century ago, so the impact of the spatial and temporal scales. Agencies must make current management activities will persist at least a long-term commitments to maintain research sites century into the future. as well as to fund basic site measurements. The mar- ginal cost of additional projects is quite low as long . We must consider both where and when we create as a base level of measurements exists. a disturbance. Important spatial and temporal com- ponents are associated with any forest management . We must simultaneously focus our collabor- activity or any natural disturbance. If, for example, ative research efforts on real landscapes. We w ill our management activities will disturb large areas in increase our understanding of the interactions and a given landscape over the next century, it makes a trade-offs only when experts from many fields apply difference whether the affected areas are contiguous their collective wisdom to the same piece of land or dispersed, and whether the disturbance occurs in a over the same time frame. Purely theoretical ap- single year or is spread over the full century. proaches to ecosystem management research have great merit, but ultimately the evaluation must be . We have enough scientific knowledge to start in the field. managing ecosystems, but we will never fully under- stand all aspects of forest ecosystems. We know a . We must remember that people are part of the great deal about some parts of forest ecosystems and ecosystem. Human activity has left an indelible mark at least a little about most parts; a prudent approach on our forest resources, and ultimately, it is people is to begin by using the best science we have available who decide which forest practices are acceptable. now, while we continue with our research. Our role as scientists and practitioners must be to: (a) identify and discourage those activities that will . We must synthesize the results of research that likely cause short-term or long-term ecosystem address many different ecosystem attributes and degradation, (b) clarify the trade-offs among many processes. We must combine what we know about acceptable management alternatives, and (c) identify ecosystem components and ecosystem processes to and encourage the alternatives that will most likely arrive at a more complete understanding of how eco- produce the desired outcomes. systems work and how they respond to disturbance. Synthesis also serves to identify the major gaps in . Summary our knowledge. A long-term experiment whose sole sponsor has left, . The complexity associated with ecosystem died, or lost interest is a sad orphan, management is so great that we must employ and adoption is seldom quite successful. mathematical models. Tracking details, measuring (Dyke ) interactions and trade-offs, dealing with long time frames, dealing simultaneously with many species, and mapping the results—all require the use of Planning constitutes the major activity associated computer models. with field studies, and may even take longer than the study itself. Care is needed in defining the experi- . We must facilitate cooperation and collabora- mental protocol, data management, and reporting tion. The complexity of forest ecosystems requires routine. Good plans are especially critical when the the attention of teams of scientists and managers main investigators are not readily available at all representing a wide range of expertise. times during plot establishment and site selection.

18 field studies of seed biology Flexibility is also required, and possible modifications The role of a scientist in the ecosystem manage- should be considered even while the study is being ment model is to provide information for the conceptualized. The plan must try to anticipate some decision-making process. Such information helps level of uncertainty and be flexible enough to cover to identify the current status of an ecosystem as unexpected conditions in the field. Change is inevita- well as potential options for addressing the social, ble, and consideration of alternative approaches during physical, economic, and biological issues (Haynes et the design stage will help to focus the planning effort al. [technical editors] ). This information helps and secure long-term success of the project. clarify feasible limits, options within the limits, Field experiments require sustained commitment consequences of those options, and trade-offs by the scientific staff so that the study will reach its between options. It is the role of the decision-maker full potential. Sustained commitment by funding to choose among options; it is not the role of science. organizations is also essential to maintain stability. The challenge for resource managers is to balance Finally, the information needs urgently required biological science with social science and with the by natural resource managers necessitates that the philosophical views of how society values renewable results of field studies reach the end user as quickly and nonrenewable natural resources (Haynes et al. and as accurately as possible. [technical editors] ).

section 1 planning tree seed research in the field 19 SECTION 2 DESIGNING AN ENVIRONMENTAL MONITORING PROGRAM

Every raincloud, however fleeting, leaves its mark, not only on trees and flowers whose pulses are quickened, and on the replenished streams and lakes, but also on the rocks are its marks engraved … (John Muir “Gentle Wilderness, the Sierra Nevada”)

. Background and frost-free period. Wind speed and direction are available for a few locations. For some parts of British Environmental and site factors influence the produc- Columbia, annual temperature and precipitation tion, dispersal, survival, longevity, and germination summaries by subzone can be obtained from the of tree seeds. Researchers must have a general under- biogeoclimatic ecosystem classification database standing of the effects of various environmental (see Appendix C). factors to select the most suitable location, time Site climate and site weather conditions involve frame, experimental techniques, and types of sensors on-site measurements of air and soil temperature, for field studies. The nature of these factors and the precipitation, humidity, wind speed and direction, overall objectives of the study will also determine solar radiation, and soil moisture. Monitoring usually which environmental variables should be measured, requires an electronic datalogging system. Weather and how frequently. data may only be needed for a short time during an Environmental variables such as temperature and event of interest (e.g., pollen release); in this case, precipitation may be considered in the context of you may need hourly rather than daily summaries. long-term average conditions, as ranges and extremes However, to characterize the climate (averages and (climate), or as day-to-day conditions (weather). Fur- variation), – years of data collection are required. thermore, environmental variables can be viewed at These data should be referenced to the nearest three scales: macro (or regional) weather, site weather, long-term weather stations to determine how and tree weather. The complexity involved in obtain- different the period being measured may be from ing data increases as we go from macroclimate to tree the “normal.” weather. Tree weather describes conditions in cones or Macroclimate and synoptic weather conditions can flowers, or beside germinating seeds. Small, delicate be obtained from the nearest Environment Canada or sensors, such as thermocouples, are usually required other government-operated weather stations, and to make these measurements. Variables of interest in may be adjusted for the elevation of the site of inter- regard to tree weather are temperature, radiation est. Usually, climate data are summarized as monthly balance, and soil moisture (for germination). The or annual values and include average, maximum, data can be used to develop physically based models minimum, and extreme values for temperature, total or regression models of tree conditions as a function precipitation, and derived data, such as degree-days of site conditions.

section 2 designing an environmental monitoring program 21 . Designing an Environmental Monitoring Program

In designing an environmental monitoring program for a research site, you must first decide which vari- ables you need to measure. For field germination studies, near-surface (– cm depth) soil temperature and soil moisture are the important variables. How- ever, soil temperature and moisture will be affected by a variety of other environmental variables. Soil temperature, for example, depends on soil moisture, solar radiation, wind speed, air temperature, soil tex- ture, and surface colour. On the other hand, surface soil moisture depends on rainfall, solar radiation,  . Climate station with wind direction and wind evaporation, vegetation cover (transpiration), soil speed sensor, a rain gauge (lower right), and texture, and soil temperature. The humidity of the tower with solar radiation, air temperature, and air and solar radiation can critically affect the initial humidity sensors. A robust tower is required at establishment of germinants through its effects on this site to support the large precipitation soil evaporation and plant transpiration. Humidity gauge used for winter snowfall measurements. also affects seed production through its effects on pollination. Slope and aspect affect temperature and however, is that you may be able to demonstrate dif- moisture because they influence the solar radiation ferences between treatments only when they are large. and rainfall reaching the surface. Wind is of interest Whatever the means of recording data, the com- primarily for studies of seed dispersal (see Section ). plexity of environmental factors and their interactions The frequency of environmental measurements necessitates careful planning of all field measurements. will vary depending on the type of measurement. Four steps to developing an environmental monitor- Light and temperature can vary rapidly and thus ing program are illustrated here using an example of require frequent monitoring. Relatively stable site a study to determine conditions that initiate flowering. factors, such as soil type, soil pH, presence and type of duff layer, biogeoclimatic zone, elevation, slope, . Why do you need environmental data? and aspect, may need only to be measured once. To determine weather conditions that initiate Researchers sometimes rely on environmental flowering, and their variation from year to year. data from the nearest weather station to provide data such as rainfall and daily minimum and maximum . What data are needed? temperatures, but if the microclimate of the site Air and bud temperatures and solar radiation, from is significantly different from that of the weather bud initiation through flowering (over  months or station it is advantageous to set up a small weather more). The year-to-year variation could be obtained station at the site (Figure .). Dataloggers can be by monitoring for many years, or by calculating used to continuously record a variety of environmen- regression equations that are based on weather data. tal variables (Figure .). Spittlehouse () provides Site weather conditions could be related to the near- guidance on using dataloggers in the field and the est long-term weather station to provide the data accuracy that can be expected from such measure- that would be needed to drive the model. Ideally, a ments. While it is tempting to collect large amounts physically based model of bud/flower temperature of weather data on the assumption that somehow as a function of site weather conditions and bud they will be useful, a few days of manual measure- characteristics should be developed to allow porta- ments under a range of weather conditions may be bility to other sites with a minimum of calibration. just as effective as installing electronic dataloggers Both methods require – years of on-site data for on the site. The disadvantage of manual sampling, development and validation.

22 field studies of seed biology . Methods for Measuring Environmental Factors

.. Soil temperature Near-surface soil temperature (– cm depth) can be easily measured, but measurements must be adequately replicated. Soil temperature varies substantially, not only horizontally and vertically, but temporally as well. Individual locations can be averaged by using a series of thermocouples connected in parallel. Data- loggers are a convenient way to monitor the number of sensors required to assess the spatial and temporal variability. The diurnal trend of the near-surface temperature parallels the diurnal course of solar radiation; thus a reasonable approximation of the daily maximum temperature can be obtained by making the measurement about an hour after solar noon. An estimate of solar noon in your area is avail- able on the Internet at http:/www.crhnwscr.noaa.gov/ grr/sunlat.htm. The average near-surface soil temper-  . Electronic datalogger used to monitor and ature (during the summer) can be approximated by capture data from a series of environmental measuring at about  a.m. solar time ( hours before sensors. The datalogger is housed in a solar noon). A comparison of treatments with this waterproof box (shown open) and data are approach requires that measurements be made under routinely retrieved using a portable computer. the same weather conditions. This manual sampling This datalogger can monitor many sensors at method will only be useful for showing differences once; less expensive dataloggers are available ° that will monitor only one sensor. larger than C and should only be used to give an idea of trends. Shade can significantly reduce solar radiation, . What needs to be done to get the data? resulting in a corresponding decrease in the near- Is continual monitoring necessary, or is a short-term surface temperature. On the other hand, shade also manual measuring program adequate? Fine wire reduces night-time cooling. When solar radiation thermocouples are needed in the buds to measure is reduced by over %, the surface temperature temperature; they are sensitive and inflict minimal can be approximated by the air temperature at damage. The site may need frequent visits to ensure . m above the ground. For more information that bud temperature sensors have not been on forest soil temperature, see Stathers and disturbed. Spittlehouse ().

. Can the work be done physically and what .. Soil moisture does it cost? There is no easy way to obtain good soil moisture Is the site readily accessible, and are equipment and measurements, and the difficulty increases as one personnel available? A datalogger-based monitoring gets closer to the surface. Gravimetric sampling and system would cost $ to $. Installation requires time-domain reflectometry (tdr) measure soil water – days depending on the number and location of content (Hook et al. ), but further work is re- sensors, and monthly site visits are required to check quired to develop tdr probes and techniques. tdr equipment and collect data. At least  day per month requires substantial replication, and although it is should be allowed per site for data processing and usually done manually, it can be automated. Water analysis—an important consideration that is content can be converted to soil water potential (or often overlooked. tension) using a soil water retention curve obtained

section 2 designing an environmental monitoring program 23 from undisturbed soil samples analyzed by a com- Radiant flux density is the energy in the light emit- mercial soil physics laboratory. ted, transmitted, or received per unit area (W/m2). Gravimetric sampling is labour intensive and Irradiance is the radiant flux density incident on a destructive. Some – replicates at each depth of surface; emittance is the radiant flux density emitted interest are required. It is best to use a sharpened by a surface. metal tube to take a soil core of known volume, The units and the instruments used for light rather than a grab sample. The sample is sealed in measurement will depend upon the intent of the plastic bags, and returned to the laboratory for study. Some radiometers (for example the li-cor weighing and drying. Gravimetric samples are pre- model li- radiometer) can be fitted with a variety sented on either a weight of dry soil or a volumetric of sensors to measure irradiance. A quantum sensor basis. The latter is the common approach and can be is used to measure photosynthetic photon flux den- converted to soil water potential (or tension) using a sity (ppfd) or par. A pyranometer (or radiometric soil water retention curve. sensor) is used to measure solar radiation. Photomet- Gypsum or fibreglass soil moisture blocks can be ric measurements using illumination units (lux or used to measure soil moisture potential (or tension) footcandles) should not be used in plant studies. in the  to -. MPa ( bars) range; they can be read Plants do not respond to the light spectrum in the manually or with a datalogger. Moisture blocks pro- same way as the human eye, so such measurements vide relatively coarse resolution, and require testing have no value unless the characteristics of the light over several drying cycles before installation. They source are precisely known (Salisbury and Ross ). may have poor contact with substrates such as par is usually the radiation measurement made when coarse sandy soil or partially decomposed organic assessing physiological responses such as plant produc- material and cannot be used at soil depths shallower tivity (although other wavelengths may have specific than  cm. photomorphogenic effects such as the induction of Tensiometers measure soil water potential in the flowering or cold hardiness). par is commonly meas-  to -. MPa (. bars) range. They are usually read ured in units of µmol • m-2 • s-1. Special sensors are manually but can be automated. As with moisture required to measure ultraviolet radiation, and various blocks, they cannot be used at depths shallower than filters are available that modify sensor output to  cm. Soil water potential can also be obtained by match the biological response of tissue. Longwave equilibrating soil samples with the air or filter paper radiation sensors are not easy to use, so longwave in a sealed container, then measuring the humidity estimates are usually obtained by subtraction of solar of the air or filter paper. In situ measurements of soil radiation from total radiation measurements (see water potential using soil psychrometers or hygrom- Black et al. ). eters is extremely difficult, particularly in the top The controlling influence of vegetation on the  cm of the soil. Further information on measuring light regime will render a shaded surface more moist soil moisture can be found in Schmugge et al. (). and cool than a bare surface. The amount of direct cover over the study area, the distance from the edge .. Solar radiation (light) of openings, and the aspect of the edge will influence Three ranges of the radiation spectrum are usually the light regime in openings. These influences can be considered when assessing the light regime at the estimated from measurements of above-canopy light earth’s surface: ultraviolet radiation from  to and the amount of cover. When measuring irradiance  nm, photosynthetically active radiation (par) under a plant canopy with uneven light levels, rea- from  to  nm, and solar radiation from  to sonable averages can be obtained by moving a small  nm. Radiation above  nm is called longwave sensor repeatedly along a track (Figure .), by using or thermal radiation. Different sensors are required a long linear sensor, or by using many spot sensors. to measure each of these bands of energy. There is a It is generally best to spend some time generating good correlation between the energy in each band radiation interception curves with an intensive meas- both above and below the canopy (Yang et al. ; urement program over a short period. These curves Alados et al. ). are used with continuously monitored above-canopy

24 field studies of seed biology reaching the forest floor passes between needles and other gaps in the canopy. Consequently, the change in the shape of the spectrum is not as great as in hardwoods where more radiation passes through the thinner leaves. The biggest change is in the in- crease in the ratio of red (– nm) to far-red (– nm). It changes from .–. under clear skies above the canopy to .–. under hardwood forest canopies, and to .–. under coniferous canopies. Light quality—the incident light spectrum—af- fects the germination of many conifer seeds (Section     . Automatic tram system that moves back and . . ) and the production of female cones (Durzan et  forth over a 50 m span to determine the al. ). For measurement of irradiances under for- variation in short- and longwave radiation, est canopies, see Black et al. () and Yang et al. and surface and air temperature under a forest (). The measurement of the total light spectrum canopy. The system is controlled by the datalogger in the tube hanging on the end. (System designed by R. Adams, B.C. Ministry of Forests.) radiation to give below-canopy data through the year. The variability or patchiness of the canopy may indicate that there is a range of light environments that must be quantified separately. Radiation rapidly decreases as canopy cover increases. The interception curve is of the form

(-KC) I = Ioe ,

where: I = the radiation at the forest floor,

Io = the radiation above the canopy, C = percent canopy cover (range –), and K = an extinction coefficient (range .–.).

Shaded and sunny areas of small openings and clear- cut edges can be determined using the formulae for length and direction of tree shadows at different  . Influence of forest canopy on the intensity and times of the day and year (Sit a). spectral distribution of solar radiation reaching Forest canopies change the quality as well as the the forest floor. The upper panel shows the intensity of light reaching the forest floor (Vezina and incident radiation from a clear sky during the middle of the day. The lower panel (note the Boulter ; Atzet and Waring ; Ross et al. ; difference in scale) indicates that pine and Messier et al. ). Figure . illustrates how the maple canopies have greater absorption in the spectral distribution is changed due to plant foliage middle range (400–700 nm) than in the near differentially absorbing and reflecting the various infrared range (700–750 nm). (Based on data wavelengths. The relatively thick needles of coniferous in Federer and Tanner 1966 and Vezina and trees transmit very little radiation and most radiation Boulter 1966.)

section 2 designing an environmental monitoring program 25 at a site requires a portable spectroradiometer. Both .. Air temperature and humidity Atzet and Waring () and Floyd et al. () con- A hygrothermograph in an instrument shelter ducted spectroradiometric analyses to determine the (Stevenson Screen) located on the study site can light-filtering capacity of coniferous forests. However, record air temperature and relative humidity for up the changes in light quality can simply be measured to a month before the chart requires changing. Elec- with a portable red:far-red meter, since most of the tronic temperature and humidity sensors can be canopy effects are due to the canopy cover shifting placed inside the Stevenson Screen. Smaller shields the ratio of red to far-red light. can be built or are available commercially, but some commercial varieties overheat under low wind .. Wind speed and wind direction speeds. Spot measurements can be made using Wind, acting either directly or by wind-induced aspirated or sling psychrometers. vibration, plays a major role in the distribution of pollen and seeds. Seeds of some species are very .. Plant temperature responsive to updrafts or vertical air movements (see Obtaining temperature measurement in cones and Section ..). Wind speed is a vector quantity with buds requires extremely small sensors. Thermocouples attributes of direction and magnitude, although only (Figure .) are the best option, being more robust and the horizontal component is usually measured. Cup or easier to make than thermistor or platinum resistance propeller anemometers are generally used to monitor sensors. They are best monitored with a datalogger. wind speed. They can be connected to a datalogger or Surface temperature can be measured using an infra- have their own display. Many ane-mometers come red thermometer with a narrow field of view. with a wind direction indicator. Topography and veg- etation cover affect wind speed and direction and .. Canopy cover care must be taken in locating the anemometer and Canopy cover is the environmental factor most im- wind vane. The sensors should be located – m mediately affected by forest harvesting activities and above vegetation canopy and away from clearcut by silvicultural practices (see Section ). It signifi- edges. Robust anemometers usually have stall speeds cantly affects the microclimate of a site, influencing of . m s–1 or greater. Low stall speed anemometers the solar radiation, air and soil temperatures, wind are required if you are interested in conditions at the speed, humidity and rainfall experienced at the edge of a clearing, in a small opening, or below the ground (Hanley ). Canopy cover is also impor- canopy. Hot-wire anemometers (commercial or tant because the position of vegetation within the home-made) can be used to measure wind flow canopy is used as a criterion for the relative domi- around cones and flowers, but they are delicate and nance of individuals within a plant community easily damaged. They can be monitored manually or (Section ..). with a datalogger. A discussion of wind dynamics and instrumentation can be found in Pearcy et al. ().

.. Precipitation Rainfall can be reliably measured with tipping-bucket or storage gauge. The gauge should be located in an opening where a line projected from the top of the gauge to the top of the surrounding vegetation has an angle of no more than ° to minimize any shading effects. Snowfall cannot be measured with a standard  . A fine wire thermocouple is used to measure tipping-bucket gauge. Although gauges that melt the the temperature inside the leader of a young snow and can be monitored with a datalogger are spruce tree. The thermocouple is monitored available, they are expensive. A low-power, reliable with a datalogger, as are the accompanying sensor that measures the depth of snow on the environmental sensors. The same technique ground can be used in conjunction with a datalogger. can be used to measure cone temperature.

26 field studies of seed biology Canopy cover is often expressed as a percentage and Brown  for a sample layout). The length of value, usually by species, growth form, or canopy the canopy intercept of each species along the line is stratum; in a dense or multilayered community, total measured from the tape or with a ruler. If the cano- vegetative cover may exceed %. The method pies overlap in layered vegetation, it may be desirable chosen to measure canopy cover depends on the to measure each height layer separately. Transect lines available technology and the type of site; some should be – m in length. Many short lines are methods are suitable for low herbaceous vegetation generally preferred to a few long lines; – transects or clearcut areas, while others are designed for are usually required for an adequate sample. Several forested areas. Bunnell and Vales () present a cover values can be calculated: comparison of different methods of measuring forest canopy cover. percent cover for each transect by species = Many researchers obtain percentage cover of length intercepted by a species × ; different species and canopy layers with the visual- transect length estimation technique (Mueller-Dombois and Ellenberg ). Cover can be estimated to the nearest percent cover of a species by sampling unit = percentage point or to the nearest th or th percen- sum of all transect lengths intercepted × . tile. Cover estimates may be restricted to the plots total transect length sampled being studied for germination or other responses, or may be used for more general descriptive purposes While this method is generally precise and accurate (such as describing the study site, Section .). The (Cook and Stubbendieck ), it can also be time visual-estimation method is especially suitable for consuming. grasslands or clearcuts, because of the low profile of the vegetation. Plot size for cover estimation averages about . m2 (either circular or square), but may be smaller when working in exclusively herbaceous veg- etation, or larger when working with tall shrubs and trees. A good guideline for plot size is that plot diam- eter should be approximately equal to the height of the vegetation being described. Visual estimates are subject to personal bias, thus human error will introduce variability into the data (Bunnell and Vales ). This can be checked and corrected (calibrated) by other people working on the same project. It is also useful to have examples of how different spatial arrangements affect one’s per- ception of canopy cover. Figure ., for example, compares different spatial arrangements of % canopy cover. These kinds of comparisons are espe- cially useful when observers are not experienced in canopy estimation methods. Canopy cover can be measured more objectively using a line-intercept method, which is suitable for woody plants, shrubs, and trees (Chambers and Brown ; Habitat Monitoring Committee ). A line or tape measure is stretched tightly across the  . Different spatial arrangements comprising vegetation between two stakes. The best sampling 50% canopy cover. Some experience may procedure is the stratified-random sample using a be needed to estimate different proportions baseline and perpendicular transects (see Chambers of cover.

section 2 designing an environmental monitoring program 27 Objective measurements of canopy cover can canopy cover. It is measured by sampling the foliage also be obtained with a point-intercept method or by using light penetration techniques (Gholz et al. (Owensby ; Levy and Madden ) or point- ; Smith et al. ; Fassnacht et al. ; Chen ). quadrat method (Chambers and Brown ). These In the former method, all the foliage in the shrub and methods use a point- or pin-frame, often consisting herb layers is removed from – samples of known of  pins spaced  or  cm apart, with pins posi- area (usually  m2). The area of the leaves is then tioned vertically or at an inclined angle. The frame is measured using an image analyzer. All the leaves from positioned randomly within the sampling units or a tree (or from a representative branch in each whorl) along a transect and a single pin lowered towards the are sampled and leaf area measured with an image ground. The first strike of any part of the vegetation analyzer. Trees of different diameter at breast height canopy becomes a “hit.” Each “hit” is recorded by (dbh) are sampled to produce a dbh/leaf area rela- species or growth-form (Chambers and Brown ). tionship (or sapwood cross-sectional area/leaf area) The sample size required for statistical adequacy is which is then used with stand dbh distribution to usually – pins. Several cover values can be calculate tree lai. Leaf area can be determined using derived from this information: percentage canopy light sensors such as the ceptometer and the lai-. cover for each species or life-form (Chambers and Both measure the “effective” leaf area and must be Brown ); percent total canopy cover; and percen- corrected for leaf clumping to get the true leaf area tage vegetation cover by species. The user should be index (Smith et al. ; Fassnacht et al. ; Chen aware that the line is the sample unit, so it is better ). These sampling techniques can be used to to have fewer points per line and more lines, rather determine how leaf area changes with height and than vice versa (Bonham ). The frame should to calculate foliage area density. be held vertically; if the frame is at an angle, the The canopy can be photographed from ground number of intercepts may increase and overestimate level using a camera fitted with a hemispherical or the cover. fisheye lens with a ° field of view. Film exposure In woodland areas, other instruments such as the must be standardized (Chen et al. ).The resulting moosehorn and the spherical densiometer are fre- photographic negatives, prints, or slides can be digi- quently used to measure tree canopy cover (Lemmon tized, and then analyzed by a computer program to ; Bunnell and Vales ). The moosehorn is a accurately measure canopy cover above the point of point-intercept method where the canopy is viewed measurement. Available computer programs include through a screen and coincidences between vegeta- solarcalc (Chazdon and Field ), gli (Canham tion and dots on the screen are tallied. The spherical ), sunshine (Smith and Somers ), and densiometer has a curved mirror with a grid that hemiphot (ter Steege ). This photographic reflects the overstorey at a particular point, then method is suitable in herbaceous, scrub, forested, or provides an estimate of the relative amount of the mixed cover, but has the drawback that considerable grid covered by vegetation. At each location, four office time is required to obtain cover estimates. readings (facing north, east, south, and west) are These same programs also model solar radiation recorded and averaged. input for the point at which the photograph was A canopy analyzer uses measurements from a taken (see Section ..), but the cover estimates fisheye optical sensor placed above and below the require fewer assumptions. plant canopy; in this way canopy transmittance data can be used to calculate the leaf area index and the .. Soil variables mean leaf inclination angle (Chen et al. ; Welles Soil nutrient levels are important because they affect and Norman ). The canopy analyzer functions in seed production, germination, and seedling growth. a variety of sky conditions, with overcast being the Three principal methods are used to diagnose nutri- best; the instrument can be used in canopies ranging ent deficiencies: deficiency symptoms, soil chemical in size from short grasses to forests. analysis, and foliar analysis. (For other methods see The area of leaf per unit area of ground (leaf area Morrison .) Soil chemical analysis has some value index - lai) is another measure used to quantify for diagnosing site nutrient status on sites where

28 field studies of seed biology foliage sampling is impractical. There are major Oxygen is usually a limiting factor only in water- problems with conducting soil chemical analysis in logged soils, where water may fill pore spaces. Oxygen forest soils. Typically, the root zone is not homogene- is difficult to measure in the field, but under suitable ous, often containing dissimilar horizons that may conditions, an oxygen electrode may be used. This yield different analytical values. Nutrient standards technique uses glass electrodes that are delicate and for forestry soils are not available, and criteria cannot easily broken, and is not generally suitable for field be extrapolated from one kind of soil to another. It studies. It is primarily designed for laboratory can be problematic to integrate these disparate results studies, but can be set up and operated adjacent to a to determine the nutrient status of the composite study site. A key requirement is constant-temperature soil profile. water, obtained from a thermoregulated circulator The high variability of some soils may require a or a large-reservoir flow-through system. Some large number of samples. The most useful routine instruments can be configured to determine oxygen soil chemical analyses for forest soil fertility in indirectly by heating the sample in a stream of inert B.C. are likely to be: pH, organic carbon concentra- gas and converting all oxygen-containing gases to tion, and total nitrogen concentration (Watts carbon monoxide or carbon dioxide. Refer to Pearcy [editor] ). et al. () for further information on such methods.

section 2 designing an environmental monitoring program 29 SECTION 3 NATURAL SEED PRODUCTION

O sweet spontaneous earth how often have the doting fingers of prurient philosophers pinched and poked thee , has the naughty thumb of science prodded thy beauty . (e.e. cummings)

. Background of seeds also tends to be poor (Edwards ; Caron and Powell a, b). Depending upon the The path to the production of a viable seed begins species, conifer seed production varies in length and with the growth and development of reproductive complexity of the production cycle (Figure .), repro- buds, continues with pollination and fertilization, ductive success (due to different sexual mechanisms), and ends with embryo development and seed matu- and the timing of natural seedfall (Zobel ). ration. Throughout all these developmental stages The reproductive structures of trees are derived losses occur for a variety of reasons; losses may be from reproductive buds. The time of initiation of due to environmental factors, or may result from male and female reproductive buds can vary from various diseases and animal predators that attack year to year due to factors such as the relative abun- cones and seeds. Researchers investigating tree seed dance of seed trees (trees/ha) (Smith et al. ) and production must be able to assess which and to what tree age (Caron and Powell b). Natural seed pro- extent these factors limit natural seed supplies. duction is rare in trees younger than  years. Angiosperms characteristically produce seeds an- Generally, the volume of seeds produced increases as nually, but production can vary considerably from the tree ages. Bergsten (), however, found no bio- year to year (Table .). Most conifers do not produce logical differences between mature Scots pine seeds collectable crops every year (Table .), a phenomenon obtained from young stands and those collected from called periodicity. Mature cones are produced at in- old stands. Environmental conditions, such as tem- tervals ranging from  to  years, and sometimes as perature, drought (Eis a, ), and nutrient infrequently as every  years. Crop yields vary in availability (Heidmann ), affect reproductive bud different years, and in poor crop years, the quality production. Environmental stresses can reduce the

section 3 natural seed production 31 1 year 4 years 1 year 2 years 2–4 years 1 year 1 year ); Wyckoff and Zasada ); Wyckoff 2 (1 / samara); no endosperm 50–100 seeds / cone no endosperm angled 20 hard, seeds + endosperm Numerous no endosperm 1–2 / stone; + endosperm 1 / samara + endosperm 3–5 carpels 1–2 seeds / carpel + endosperm Numerous; no endosperm Numerous; no endosperm Paired winged seeds Paired 3–6(samaras); cm long cones (2Brown cm) winter; over remain winged oval, contain nutlets Orange-red berry-like (1 finely surface cm), granular with wings Nutlets broader than body ofClusters bright red (1“berries” cm); globular ovoid or drupes samara Paddle-shaped, (3–5 cm) in large on female trees clusters small reddish to Yellow (10–15 mm) fleshy pomes not capsules; 2-valved hairy hairy green Round, that split into capsules 3 parts; seeds covered with fluffy hairs white Seeds Interval Ripen Sept–Oct;Ripen disperse Oct–Jan Aug–Sept; Ripen disperse Aug–spring Fall Aug–Sept; Ripen disperse Aug–spring late Ripen summer/fall Late summer/fall Late fall June–July May–July Appear with orAppear leaves before (Apr–May) previous Male, fall; female, Feb–May Apr–May previous Male, fall; female, Apr–June in Always in spring; often fall before Appear leaves Apr–June Apr before Appear (Apr– leaves June) Greenish yellow (3 yellow Greenish mm in numerous across); hanging cylindrical cluster; in male and female partsdifferent of crown male reddish Drooping catkins (5–12 female cm); cones woody catkins are (2 cm) urn-shaped (7White, mm), clusters in large drooping and flower staminate Male (2–4 strobile female cm); up at maturity break Small (3 mm) white in surrounded tight clusters bracts showy 4–6 white by (2–7 cm) Small (3 mm) male and female (yellowish) in flowers (greenish) twigs on clusters bunched pink fragrant to White (2blossoms 5–12 in cm); clusters flat-topped catkins and female Male catkins (2–3Male cm) with 40–60 stamens/ catkins female flower, (8–20 cm) with 3 stigmas/ flowers Monoecious; imperfect flowers Monoecious Monoecious; flowers perfect Monoecious Monoecious; flowers perfect Dioecious Monoecious; flowers perfect Dioecious Dioecious ssp. ssp. [1998]; Zasada et al. [1998] Seed production characteristics of hardwoods native to British Columbia. Sources: Schopmeyer (1974); Pojar and MacKinnon (1994 native to British Columbia. Sources: Seed production characteristics of hardwoods balsamifera balsam poplar trichocarpa black cottonwood bigleaf maple alder red arbutus paper birch dogwood Pacific ash Oregon crab apple Pacific Common nameCommon type Tree (description) (month) (month) (description) seeds/fruit crops Species Flowers Flowers mature Fruit # Average between  .  Alnus rubra Alnus menziesii Arbutus Betula papyrifera Cornus nuttallii latifolia Fraxinus fusca Malus balsamifera Populus balsamifera Populus Acer macrophyllum Acer

32 field studies of seed biology 4–5 years 2–3 years Good crops withalternate poor 2–7 / capsule; 77– 2–7 / capsule; 500 seeds / catkin no endosperm 1 / drupe 1 / acorn; cotyledons only, no endosperm seeds / 2–3 nutlike drupe + endosperm 14–18 seeds / capsule 5–7 seeds / capsule; 144–311 seeds per catkin 8–12 / capsule 15–36 / capsule 12–20 / capsule Catkins of tiny, greenish Catkins of tiny, with covered capsules down cottony Dark drupes (1 red cm) (2–3Acorns cm) in scaly cups shallow, round Purplish-black, berrylike drupe capsule 2-valved Small, hairycontains seeds 24–48 capsule; 2-valved capsules/catkin capsule 2-valved capsule 2-valved capsule 2-valved capsule 2-valved Seeds Interval May–June July–Sept Aug–Dec July–Sept July May–June May–June June–Aug May–July Apr–May Apr–June Feb–May Apr–July May–June Apr–June May May–July Apr–May Apr–June Male catkins (2–3Male cm); female catkins (4–10 cm) pinkish 5–10 white to (10–15flowers cm) in flat- topped cluster inconspicuous Tiny in hanging male, flowers; single or catkins; female, in small clusters ofLoose 5–12 tiny, clusters yellowish-green flowers Staminate catkins soft, silky Dioecious Monoecious; flowers perfect Monoecious Monoecious Dioecious Dioecious Dioecious Dioecious Dioecious Dioecious (Continued) ssp. quaking aspen cherrybitter Garry oak cascara peach-leaf willow willow Bebb’s pussy willow sandbar willow lasiandra willow Pacific willow Scouler’s Common nameCommon type Tree (description) (month) (month) (description) seeds/fruit crops  .  Species Flowers Flowers mature Fruit # Average between Populus tremuloides Populus Prunus emarginata garryanaQuercus purshiana Rhamnus Salix amygdaloides Salix bebbiana Salix discolor Salix exigua Salix lucida Salix scouleriana

section 3 natural seed production 33  . Seed production characteristics of conifers native to British Columbia (Eremko et al. 1989). (Cones refer to female cones only.)

Cone Cone- Years Viable seeds Position of Ease of Species length bearing age between per hectolitre cones in cone Common name (cm) (years) crops of cones crown detachment

Abies amabilis 9–13 20 2–3 30 389 Top ¼ Difficult amabilis fir Abies grandis 5–12 50 2–3 50 776 Top ¼ Difficult grand fir Abies lasiocarpa 6–12 20 2–4 40 582 Top ¼ Difficult subalpine fir Chamaecyparis 0.5–1.5 Unknown 4 or more 93 965 Throughout Easy nootkatensis yellow-cedar Larix laricina 1.5 40 3–6 32 000 Non-shaded Moderate tamarack part of crown Larix occidentalis 2–3 25 1–10 119 312 Non-shaded Moderate western larch part of crown Picea glauca 3–6 40 6 347 163 Top ⅓ Moderate white spruce Picea mariana 2.5 10 4 or more 108 000 Top ¼ Difficult black spruce Picea sitchensis 5–10 25–40 3–4 194 270 Top ⅓ Moderate Sitka spruce Pinus albicaulis 3–8 20–30 3–5 515 Throughout Difficult whitebark pine Pinus contorta 3–6 15–20 2–4 coast: 176 660 Throughout Difficult unless shore pine interior: 70 546 frozen Pinus flexilis 7–20 20–30 2–4 6 454 Throughout Moderate limber pine Pinus monticola 10–25 20 3–7 7 687 Top ¼ Moderate western white pine Pinus ponderosa 7–9 12–16 4–6 31 522 Throughout Difficult ponderosa pine Pseudotsuga menziesii 5–10 20–25 2–10 coast: 39 577 Top ½ Easy Douglas-fir interior: 70 343 Thuja plicata 1–2 20–30 2–4 897 837 Throughout Easy western redcedar Tsuga heterophylla 2–3 25–30 3–4 366 698 Throughout Easy western hemlock Tsuga mertensiana 2–8 30 3–6 356 428 Top ⅓ Easy mountain hemlock

34 field studies of seed biology number of reproductive buds or, in other cases, can is sometimes dispersed by wind, but generally angio- stimulate prodigious production of cones. Plant sperm pollination is less affected by climatic variables, growth regulators (pgr) such as gibberellins have although extreme conditions (cold temperatures, been used to increase cone production in conifer seed heavy rain) may still affect pollination success. orchards (Ross and Bower ; Ross ), but pgr Fertilization efficiency may be reduced due to poor levels are difficult to alter, for logistical reasons, in female cone production, self-pollination (which often natural stands. results in embryo abortion), lack of pollen tube growth, Pollen, produced in male cones or anthers, is or freezing temperatures (Shearer and Carlson ). transported to female cones or flowers in the process Some conifers (e.g., Douglas-fir) can produce seeds of pollination. Successful pollination results in the (megagametophyte, but no embryo) without fertili- fertilization of ovules; ovules then develop into seeds. zation, but other conifers (e.g., pines) require the Reduced pollination efficiency may be due to low presence of pollen to form seeds (Owens and Molder pollen-cloud densities (few pollen-cone buds initi- b). Additional background on the sexual repro- ated), climatic conditions (e.g., rain, frost), or the duction of conifers may be found in Owens and presence or absence of pollen vectors. In conifers, Molder (a, b, c, d, ). which are wind pollinated, the absence of wind, or Once fertilized, seeds may fail to mature due to barriers to wind may inhibit pollination. Thus, the abortion (which may be caused by self-incompatibility, positioning of cones relative to tree height or relative insects, or disease), or because of climatic conditions to the windward and leeward sides of a tree can in- during embryo development, particularly cool, fluence the frequency of filled seeds (Smith et al. ). cloudy weather during the summer (Eis ; Zasada In angiosperms, animals (insects, birds, and mammals) et al. ). Some conifers do not shed their seeds usually are the primary vectors of pollination. Pollen when they mature in the fall, and instead may retain

Mature seeds

 . Typical development and maturation cycles of British Columbia conifer seeds (Leadem 1996, adapted from Eremko et al. 1989). Most angiosperms exhibit a reproductive cycle similar to that shown for Douglas-fir, redcedar, true firs, and hemlocks.

section 3 natural seed production 35 their seeds in the cones several years, a phenomenon of this size probably would be sufficiently mature referred to as serotiny (see Section ..). to produce seeds (McCaughey and Schmidt ). Further information on tree seed biology may be found in Leadem (). Once a site has been selected, data on individual Seed production studies are usually undertaken to trees may be collected. Examples of the data that determine: could be included are (Alexander et al. ): • the quantity and quality of seeds that may be • dbh to the nearest . cm (trees . cm dbh produced relative to some variable, or and larger); • the reasons for seed loss. • total height, to the nearest . m; • crown class; Other, more specific objectives may include: • species; • to predict the frequency of good seed crops • average length of live crown to the nearest . m (e.g., relative to climatic variables); (average of four sides); and • to relate seed production to stand, tree, or crown • average width of live crown to nearest . cm characteristics; (average of two measurements). • to examine the relationship between pollen abun- dance and filled seeds per cone; .. Determining sample size • to relate the number of seeds per cone to cone age The choice of sample size, such as the number of seeds or cone size; to sample per tree, can be made by applying statistical • to determine the relationship between the number efficiency calculations to a preliminary set of meas- of seeds in the cone half-face and the total number urements (Sokal and Rohlf ; Ager and Stettler of filled seeds per cone; ). See also Stauffer (, ) for sample size • to determine the date of cone and seed maturation; tables prepared specifically for forestry applications. or Sample sizes for measuring cone characteristics • to establish the relation between seed quality and will depend on the species and the sites from which collection date, or cone handling methods. the cones were collected. Carlson and Theroux () randomly selected  cones each from some sub- Several examples of seed production studies are alpine larch, hybrid larch, and western larch collections. described as case studies in Section .. Only five cones were selected from six other western larch collections because initial sampling error esti- .. Collecting stand and study plot information mates indicated that five cones would be adequate. Before conducting seed production studies in natural Sample sizes for seed measurements should also be stands, data should be collected on the tree and stand determined before the study. For a study of western characteristics known to influence natural seed pro- larch and subalpine larch, length, width, and thick- duction. Examples are: ness were measured on only  seeds randomly • density (number of seed trees per hectare); selected from each lot; initial sampling estimates in- • spatial arrangement of seed trees; dicated this would enable standard errors to within • age of seed trees; % of the mean (% confidence) • evidence of past production; (Carlson and Theroux ). • evidence of animal use (e.g., squirrel caches, cones Environmental changes may result in year-to- that have been broken or split); year variations in cone and seed measurements. • height and diameter at breast height (dbh); Ponderosa pine seeds collected in , , and  • assessments of the general health and vigour of the showed negligible differences in seed weight, length, crowns; and and width when comparisons were made within the • basal area values (in square metres per hectare). same year. However, differences were found in all Specific selection criteria may be included, for three measures when year-to-year variations were example the basal area of all Engelmann spruce removed by adjusting values to be relative to those trees with dbh of . cm and larger, because trees observed in  (Ager and Stettler ).

36 field studies of seed biology . Predicting Natural Seed Yields For estimates of Douglas-fir, grand fir, and western white pine cone crops, Eis (a, ) counted It is often desirable to be able to predict the occur- cones on one side of mature trees in July. Counting rence of natural seed production, to better was done using -power binoculars mounted on a understand what factors promote seed crops, to tripod at a permanent station that offered a good determine whether seed production will be great view of the crown. Cone counts were multiplied by enough to merit collection of the crop, or to provide conversion factors (obtained by comparing binocular advance notification for organizing pre-collection observations with physical cone counts on felled activities. In the sense used in this section, a distinc- trees). Weather variables were derived from various tion is made between prediction and correlation. expressions of temperature, precipitation, sunshine, Prediction is the objective of these studies (we are and wind velocity. Starting  months ( months for trying to predict natural seed production) and corre- western white pine) before the cones matured, cone lation is the means to do so (correlations with various estimates were correlated with all monthly meteoro- variables are used to predict the size of the crop). logical parameters. Where several meteorological variables were important in the same month, the data .. Correlation with weather variables were combined and analyzed by stepwise, forward, Many models for predicting the size of natural seed multiple regressions. crops have been developed, and those based on cli- Caron and Powell (b) correlated annual pro- matic variables indicate that the influence of weather duction of black spruce seed cones with warm weather conditions may be cumulative. In Douglas-fir and in early May and early July and with low June rainfall, grand fir, a cool, cloudy summer – months all in the year preceding maturation. Cone produc- before crop maturation appears to be a prerequisite tion data were recorded branch-by-branch during for abundant lateral bud initiation. These conditions later spring. Seed-cone estimates of previous crops must be followed by cold, sunny weather through the were obtained from a combination of () cones per- winter (– months before maturation); a wet sisting on the trees, () stubs and basal cone scales left April ( months in advance) to promote lateral bud on the bearing shoots where squirrels had removed differentiation; and a warm, dry, sunny June before cones, and () cones or stubs on nearby shoots of pollination ( months before maturation) (Eis a, comparable size and position within the crown when b). (See Figure . for a summary.) The impor- shoots of bearing type had been removed. tance of dry summers to floral initiation has also Mosseler () used accumulated growing been demonstrated in other species, such as spruce, degree-days (gdd) to predict when cones of black larch, and ponderosa pine (Eis and Craigdallie ). spruce and white spruce could be collected without

Seedfall

Aug Sept

0

 . Climatic conditions required for cone crop production in Douglas-fir (Eremko et al. 1989).

section 3 natural seed production 37 adversely affecting seed quality. Accumulated gdd is .. Correlation with aspect and slope a cumulative sum of the degrees of temperature above Aspect and slope can significantly affect cone produc- °C counted on each day that the daily mean tem- tion, especially in northern regions. For example, perature exceeds the °C threshold. In this study, black spruce trees growing on southerly aspects bore Mosseler based the gdd on the simple mean of the . and  times more seed cones and pollen cones, maximum and minimum daily temperatures recorded respectively, than trees growing on northerly aspects at the Atmospheric Environment Service (Environ- (Simpson and Powell ). Variations in slope and ment Canada) weather station nearest to each site. aspect can be difficult to depict, yet Simpson and Cones were harvested at intervals of  gdd begin- Powell effectively conveyed their results by using ning at about  gdd. Mosseler found that natural concentric circles to show the percentage of cones seed release in white spruce occurred between  produced in all compass directions (see Figure .). and  gdd. Cones from black spruce can be col- lected as early as  gdd and white spruce as early .. Correlation with crown size and crown class as  gdd without significant losses in seed yield In a closed canopy, the crowns of the trees forming or quality. Similar results were found for white spruce the canopy are touching and intermingled so that in Alaska (J. Zasada, pers. comm., ). light cannot directly reach the forest floor. However, Note that when attempting to correlate environ- dominant trees have crowns extending above the mental factors to seed production, it is important to general level of the canopy and thus receive full light place sensors as near as possible to where pollen and from above and partly from the side. The crowns of seed cones are produced to ensure you are monitor- codominant trees, which form the general level of the ing the conditions actually present in the canopy. canopy, receive full light from above, but compara- Also, because comparable events in the reproductive tively little light from the sides. The relatively more cycle are not always synchronous, male and female favourable light environment for tree crowns in the flowers, for example, may not experience the same upper canopy appears to enhance the cone produc- climatic conditions, so the environmental effects may tion of dominant and codominant trees. be different. In paper birch, male flowers are induced For example, dominant and codominant crown in May before bud burst and thus must depend on classes of Engelmann spruce produced three-quarters resources stored in overwintering materials. Female or more of the total seedfall in an experimental forest flowers develop in late June to early July, so they are in the Colorado Rocky Mountains (Alexander et al. able to draw on current metabolites for their growth ) (see Case Study , Section .). Also in black (Macdonald and Mothersill ). spruce, dominant trees produced almost three times

a) b) c)

 . Percentages of black spruce trees (concentric circles) 8–12 years old from seed, growing on slight (2–12%) slopes and on various aspects (Simpson and Powell 1981), which in 1980 bore: (a) more than five, (b) more than 15, and (c) more than 25 pollen cones.

38 field studies of seed biology as many cones as codominant or the intermediate and of cone-producing branches. He concluded that trees. Intermediate trees, on the other hand, produced total cone production was related to the square of the about twice as many seeds per cone as dominant trees trunk diameter just below the lowest living branch   2   (Payandeh and Haavisto ) (see Case Study , Sec- (DB in Figure . ). He concluded that cone produc- tion .). Note, however, in exceptionally good cone tion was not related to the total amount of dry matter years, trees in all crown classes produce cones, not in the tree, but rather to the amount of dry matter just the dominant and codominant trees (J. Zasada, accumulated in the neighbouring trunk and branches pers. comm., ). adjacent to cone production sites. Thus, cone pro- One possible reason for the periodicity observed duction per branch depends on the resource status of in conifers is the substantial drain that cone produc- a branch, to which at least part of the resources may tion evidently places on the tree’s resources. In be imported from other branches or the trunk for Douglas-fir, decreased needle, shoot, and xylem ring local cone production. In this way, the investment in growth was noted in good seed years in the trees that seed production in individual branches may not nec- regularly produced cones (Tappeiner ). No such essarily cost the whole plant its vegetative growth or reductions were seen in trees that did not produce future survival. cones. Similar effects have been seen in grand fir, western white pine (Eis et al. ), subalpine fir, .. Sampling methods using bud counts and mountain hemlock (Woodward et al. ). Eis (b, ) developed a sequential sampling meth- Seki () wanted to know the specific location of od to estimate white spruce and western white pine resources used to produce seeds in Abies mariesii, so cone crop potential in the fall preceding the seed year. he studied the allometric relationship between cone The method is based on the cumulative total count of production and the productivity of the entire crown female buds from one branch per tree collected from

 . Position of measurement for trunk diameters, the diameter of the base of a branch, and main axes for estimating of the number of cones (Seki 1994). Allometric relationships between various parts of a tree can be used for relating cone production to tree dimensions.

section 3 natural seed production 39 the third whorl from the top. Bud counts from three  . Cone crop rating based on the relative number terminal nodes on a branch of the fourth or fifth of cones on the trees (Eremko et al. 1989) stem node may also be used, but with slightly lower based on dominant and codominant trees only accuracy. Trees should be – years old, – m high, of dominant class, with well-developed Class Rating Definition crowns. The observer must be able to distinguish repro- ductive buds (both male and female) from vegetative 1 none No cones buds (identification based on general morphology, 2 very light Few cones on less than 25% location in the crown and along the branch, and col- of the trees our). When the cumulative bud count falls between 3 light Few cones on more than 25% given limits, cone crop potential can be classified with of the trees % probability, and no further samples are required. 4 light Many cones on less than 25% Male pollen buds can be used in birch and alder to of the trees indicate the next year’s seed production. Male buds 5 medium a Many cones on 25–50% of the trees are easily identifiable any time after September, and a can be counted from the ground to provide reason- 6 heavy Many cones on more than 50% of the trees able estimates of female catkin production the a following spring (J. Zasada, pers. comm., ). 7 very heavy Many cones on almost all of the trees A multistage variable probability sampling method, originally developed to estimate seed orchard effi- a Crops rated as class 5 or higher are generally considered ciencies, could be applied to assess seed production collectable. in natural stands. Bartram and Miller () first implemented a standard multistage approach in many seed orchards over several years. The effective- The rating of potential cone crops is highly subjec- ness of this approach was evaluated against several tive and dependent on the surveyor’s experience. The alternative methods using the efficiency data initially number of cones produced—and their distribution collected for the study. Refer to the original paper through the crown—varies considerably with tree for an example using this methodology in coastal species. Thousands of cones can constitute “many” Douglas-fir seed orchards in British Columbia. on a spruce tree; the same number could be classed as Mattson () also suggested a multistage approach “few” on a mature cedar. to evaluate red pine cone and seed production. In this A method based on seedfall data has been used in scheme, the first stage is based on weather factors and Oregon and California for rating cone crops of Abies, the second stage on insect predators. Pseudotsuga, Tsuga, and Chamaecyparis (Zobel ). Seeds were collected from traps approximately once a .. Scales for rating cone crops month over a -year period. The monthly trap sam- Crop rating is an operational assessment procedure ples from a site were usually combined, except where used by the B.C. Ministry of Forests to determine seed production was high enough to separately count whether developing cone crops are collectable individual traps. Basal area of each tree species over (Eremko et al. ). Suitable stands are located, and  cm dbh in each stand was measured using a wedge the relative size of developing cone crops is assessed prism, with each trap as a sample point. Seed produc- in late June or early July. A visual assess-ment is made tion effectiveness of a site was expressed as the annual of the relative number of cones in the cone-bearing seedfall per square metre of basal area of each species. portion of dominant and co-dominant tree crowns. In another study, seed production of Engelmann The number of cones on each cone-bearing tree and spruce was based on seeds captured in traps and percentage of trees bearing cones in the stand are also grouped into categories (Table .). assessed. Observations are grouped into classes Note that there may be some discrepancy between depending on the relative number of cones observed the cone crop rating and the number of seeds col- on the tree (Table .). lected in seed traps. Such discrepancies can occur

40 field studies of seed biology  . Rating of seed crops by number of filled seeds fertilization take place in the same growing season, per hectare (Alexander et al. 1982) but the total cycle usually lasts about  months. Under natural conditions, seed maturation is delayed Filled seeds per hectare Seed crop rating by a period of dormancy until the following year (Figure .); however, under favourable conditions < 25 000 Failure in seed orchards, pollination, fertilization, and 25 000–125 000 Poor seed maturation can occur within the same year 125 000–250 000 Fair (El-Kassaby et al. ). 250 000–625 000 Good In conifers, male and female strobili appear on the same tree (with the exception of yew). However, the 625 000–1 250 000 Heavy distribution of cones within the crown varies with > 1 250 000 Bumper the species (Table .), and even within the same species, male and female cones may occur in different parts of the crown. In hardwoods, it may be necessary in areas where there is heavy predation of cones by to identify male and female clones before monitoring, squirrels. Thus, generally it is best to count when since dioecious hardwoods, such as Salix, Populus, cones have attained maximum physical dimensions, and Fraxinus, bear male and female flowers on and before squirrels begin to harvest. different trees. Cone crop estimates also can be obtained by direct sampling of cone-bearing regions or fertilized Monitoring pollen flowers. For example, cone crops of eastern redcedar Pollen abundance in the air during female receptivity (Juniperus virginiana) were estimated by multiplying is believed to be closely related to seed production. the average number of cones per sample branch Estimates of pollen abundance can be used to formu- (–) in October by the cone-bearing canopy late a relationship to () total seed production for the foliage (Holthuijzen et al. ). stand, and () total filled seeds per cone. In lodgepole pine, the number of male meristems produced on .. Monitoring the seed crop individual trees was correlated to the frequency of The sequence and length of different components filled seeds on those trees (Smith et al. ). Weather of the reproductive cycle must be understood when data for previous and current year also can be planning and executing seed production studies be- incorporated to develop predictive models for seed cause different components are not the same in each yield. Possible relationships that can be studied are species. For example, you need to know the length of the amount of rain during flower initiation and the entire reproductive cycle of each species involved temperature minima during critical stages of cone in the study, since this will determine when monitor- maturation (Stoehr and Painter ). ing should begin. You must also know the timing of Many different kinds of pollen samplers have been other critical events, such as pollination, fertilization, used to assess pollen abundance. Each has its advan- and periods of bud dormancy, to ensure you are at tages and disadvantages, but often the choice depends the right place at the right time. on the financial resources available. The least expen- For Douglas-fir, redcedar, spruces, true firs, and sive approach is a microscope slide placed on a flat hemlocks, the development and maturation cycle surface or on the ground. The results from such a takes about  months (Figure .). For these species, method may have little relationship to the actual male and female strobili appear in the spring, and densities experienced at the female cone level, but seeds mature in the fall of the same year. In maple, it may be possible to correlate the data with crown alder, birch, Garry oak, and willows, seeds are also measurements. produced in the same year as the female flowers. Another inexpensive method used in Sweden and However, in pines, complete development takes Finland is to trap old pollen strobili that fall to the  months because fertilization is delayed for  year ground to obtain estimates of pollen cone production after pollination. In yellow-cedar, pollination and (Leikola et al. ). At the other end of the spectrum

section 3 natural seed production 41 are automated air samplers that pull a known volume long) with a fitted rubber stopper is used to protect of air through the sampler so that air movement is the catching surface from contamination before in- dynamic. This differs from strip chart recorders, stallation and after collection. The traps are collected which are “passive” and depend on natural wind and and replaced daily. air movement to adhere pollen to the “sticky” surface. Smith et al. () sampled airborne pollen for den- The abundance of pollen in conifer stands can sity estimates of lodgepole pine at canopy level near be estimated using a -day pollen monitor (Webber the centres of two squares that made up a central and Painter ). Several monitors (three is good, rectangle surrounded by a  m bank. Air samples depending upon the size of the plot) should be placed were taken at -minute intervals over  or  days in the experimental sites  week before expected pol- with Kramer-Collins air samplers (Kramer et al. ). len shed and left in the field until the pollination Pollen counts averaged over the  peak days were used period is completed. The monitor is mounted on a to calculate relative pollen densities in the  stands pole  m above the ground and always turns into the chosen for study. Pollen cone production was estimated wind. The monitor consists of a permanent chart by counting the number of terminal meristems pro- wrapped around a drum, which is rotated with a ducing male strobili. Using binoculars, terminal clock mechanism. The chart is made of mylar coated meristems were counted for either the entire tree, with petroleum jelly so that pollen will adhere; heat- or  or  meristems were counted on a portion ing the petroleum jelly slightly creates a smooth, even of larger trees, and that portion was estimated as a coat. Since the drum completes one turn each week, percentage of the total surface area to provide a pollen charts must be changed weekly. Pollen charts meristem total for the tree. Estimates ranged from are assessed in the laboratory, where pollen densities,  to  male meristems per tree. To check the timing of dispersal, and pollen identification can be consistency of the technique,  trees were counted determined on a daily basis. Proper evaluation of on consecutive days. Of the  counts that differed pollen charts depends on the experience and exper- on the  days, the second day had the smaller count tise of an analyst familiar with pollen density patterns.  times. The smaller count averaged % of the Sarvas (, ) used different types of pollen larger count for the  trees that had male meri- samplers to estimate pollen density in Scots pine stems. Using a similar sight estimation of female (Pinus sylvestris) and Norway spruce (Picea abies), cones on trees that were later cut down so that cones and was able to closely relate flowering to heat-sum could be counted, Elliott () and Smith () calculations. He placed all his pollen samplers on found that their under- or overestimation deviated towers at the height of the female flowers. Zasada et from the actual count by an average of %. al. () used this method for white spruce. Sarvas Pollen from different species can usually be identi- also used a small globe sampler to measure pollen fied microscopically (Figure .). In a study of black rain density. The globe shape was chosen because it spruce, the pollen catch was systematically examined more closely resembled the shape of a female cone, (-power magnification) on the four directional and thus better approximated the pollen density pat- faces of each trap (Caron and Powell a). Pollen terns expected on a “cone-shaped” surface. identification to the species level (or at least genus) Another pollen trap described by Caron and was accomplished by comparing pollen samples Powell (a) consists of a glass rod ( mm diameter collected directly from trees with micrographs and ×  cm long); one end is coated with a thin layer of species descriptions (Richard ; Adams and white petroleum jelly to serve as the catching surface, Morton ). In mixed stands comprised of species the other end is tightly fitted through a rubber stop- with similar-looking pollen, it is advisable to install per into a hole on a wooden base. The square wooden additional pollen monitors in pure stands of the base is grooved to slide into a holder, which can be species located near to the study site. Pollen density set so that the edges are aligned to the four cardinal records from mixed stands can then be compared directions. Each trap holder is protected from rain by to those from pure stands to determine the relative polyethylene film installed on a wire frame – cm pollen abundance and time of pollination of different above the base. A vial ( mm diameter ×  mm species (Stoehr and Painter ).

42 field studies of seed biology Monitoring seed cones descending branches within the lower two-thirds of This section focuses on monitoring female conifer the crown (Shearer and Schmidt ). In British Co- cones, although the same procedures could also be used lumbia considerable overlap of the seed and pollen for the flowers, fruits, and seeds of hardwood species. cones occurs within the lower half of the crown To sample for cone production (either male or (Owens and Molder a, b) (Table .). female cones) it is necessary to determine where In species such as whitebark pine, cone scars can the cones are produced. In western larch, most seed be monitored to estimate past cone crops (Morgan cones are produced on ascending branches or on and Bunting ) (Figure .). Morgan and Bunting recent terminal leaders within the upper third of the chose a  m transect that contained  mature, crown; most pollen cones are found on horizontal or cone-producing whitebark pine trees with crowns

a) b)

c) d)

e) f)

 . Scanning electron micrographs showing whole pollen and details of the exine. (a), Chamaecyparis nootkatensis (×1100); (b), Betula (×860); (c), Abies amabilis (×400); (d), Pinus contorta (×720); (e), Pseudotsuga menziesii (×360); (f), Tsuga heterophylla (×540). (Owens and Simpson 1986).

section 3 natural seed production 43 Similarly, in Douglas-fir, it is possible to trace pedicle remains to estimate previous cone production (Tappeiner ). In Abies, cone spindles remain on branches for several years after the cones have disin- tegrated; these also might be used to estimate previous production (J.C. Tappeiner, pers. comm., ). In a study of ponderosa pine, four branches in the upper half of the crown were randomly selected and permanently marked for the presence of male and female flowers (strobili), conelets, and mature cones (Heidmann ). Flowers were counted in July of each year for  years. Because of the great number of male flowers on some branches (as many as  clus- ters per branch), an average was determined for a sample of  clusters, then multiplied by the number  . The oval, raised cone scars of Pinus albicaulis of clusters to obtain the total flower count for that can be counted and aged by the nearby branch. All female flowers were counted. annual bud scars on twigs (Morgan and Fourteen western larch stands ranging in age from Bunting 1992).  to  years were monitored by Shearer and Carlson () in Idaho, Montana, Oregon, and Washington. Using binoculars, they estimated the readily visible from the ground from at least two angles. number of new seed cones in spring. Five trees with The data were obtained by visiting each transect in late the greatest seed cone counts were climbed and the July or early August (before appreciable harvesting by number of developing cones was estimated by count- red squirrels and Clark’s nutcrackers); binoculars ing the number of branches with seed cones, and new were used to count the mature cones visible from seed cones (living and dead) on six random branches the ground. Each tree was climbed to access cone- (two from each third of the crown). The number of bearing branches normally found in the uppermost potential seed cones was estimated by multiplying the portion of the canopy; cone-bearing branches are average number of cones per sample branch by the visibly shorter and stouter than other branches. Five number of cone-bearing branches. Seed cone survival branches were sampled that bore either cones or re- was estimated in August by counting the number of cent cone scars. The mean number of cones or cone cones that matured on the six branches selected in scars was calculated for each sampled stand. The the spring. Seed cone mortality was determined by transformed values of the nonzero data were stand- subtracting surviving cones from the total cones ardized by calculating z (the difference between counted at the first visit of the year. During the first the individual observation and the means of all visit, researchers marked  cones on the two trees observations, divided by the standard deviation of all bearing the most cones at each site. During subse- observations). The z-scores were calculated separately quent visits they documented cone development, as for scar and cone counts, then combined for classi- well as the time and cause of cone damage. Dead fication into poor, average, or excellent cone crops. cones were removed and the probable cause of death Morgan and Bunting recommend counting imma- was identified. ture cones as an index of cone production in the following year. Another method involves counting Cone and seed analysis immature cones on high-resolution aerial photo- Initially developed for southern pines, cone and seed graphs. The immature cones are readily visible from analysis is an excellent procedure for identifying above; they occur at the ends of upswept branches actual and potential seed production and the causes near the top of crowns and their deep purple colour of seed loss in conifers. Bramlett et al. () provide contrasts with the foliage. complete background and procedures.

44 field studies of seed biology Cone and seed analysis is based on calculating cone and seed efficiencies would require some four critical ratios (efficiencies), which are used modifications since, in species such as Salix and to identify the sources (stages) in which major Populus, catkins are comprised of capsules, each of losses occur: which has several to many seeds. In Alnus and Betula, catkins are more like conifer strobili. In Acer and number of cones harvested cone efficiency = Fraxinus, fruits are paired or single samaras, respec- number of conelets initiated, tively, each containing one seed per samara. Refer to Table . and to Section .. for further information number of filled seeds seed efficiency = on hardwood fruit characteristics. number of fertile sites, Whether the method is used for conifers or hardwoods, two cautions should be considered in extracted seeds per cone extraction efficiency = conducting cone and seed analysis and extending the total filled seeds per cone, and results to the species: . If at all possible, the cone analysis should be number of germinated seeds germination efficiency = repeated during another good seed crop year. total filled seeds. . If an unharvested stand can be found near the study plot, cones should also be collected from the The analysis requires the determination of: unharvested stand for comparison. • the potential number of seeds per cone; • the total number of seeds per cone; Filled seeds per cone • the number of extracted seeds per cone; Measurements of the number of filled seeds per cone • the number of filled seeds per cone; and are obtained primarily for predictive purposes. • the number of empty and insect-damaged seeds Numbers are used to plan the size of cone collections per cone. in a particular area, or to estimate the potential of a site for natural regeneration. Large samples are Note that the number of filled seeds must be generally not needed. For example, in ponderosa pine determined in addition to the total number of seeds. only  closed cones from each lot were required to This is essential to reflect the actual seed production obtain good corre-lations between filled seeds per potential of the species. cone and kilograms of seeds per hectolitre of cones Commercial services are available if you do not (Ready ). wish to perform your own cone and seed analyses Determining the total number of filled seeds per (refer to Portlock [compiler] ). cone is time consuming, as it requires complete dis- Cone and seed analysis has been applied in British section of the cone. Special tools are needed as most Columbia to analyze seed production of lodgepole cones are hard and woody. For these reasons many pine and Douglas-fir (McAuley a, b). For studies have attempted to relate the number of filled, the analysis, samples were randomly selected from sound seeds seen in the cone half-face to the total five sacks among those filled that day. Random sub- number of seeds in the cone (see Figure .). Schmid samples consisting of one cone per sack were placed et al. () tested several sampling designs to deter- in separate bags for subsequent analysis. Based on mine the accuracy and precision of each design in previous experience, a sample size of  Douglas-fir estimating the mean numbers of filled seeds. They cones per orchard ( cones for lodgepole pine) was found that half-face counts on  cones (two cones considered to yield a reasonably precise estimate of from each of  trees) from a ponderosa pine stand single, orchard-level means. Samples for cone and estimated the filled-seed percentage for whole cones seed analysis of western larch consisted of  cones within ± units of the mean. Olsen and Silen () collected from each of  individual trees per hectare multiplied the number of filled Douglas-fir seeds (Stoehr and Painter ). seen in the cone half-face by . for an estimate of the Hardwood trees could also be assessed using total seeds per cone. From each . L of undried coneand seed analysis methods. Determining the cones, they cut  cones in half longitudinally,

section 3 natural seed production 45 counted the number of full seeds on one cut surface, British Columbia tree species follows (summarized then dried and extracted the cones to determine the in Table .). In addition, various cone and seed at- total number of filled seeds. tributes (such as colour, weight, and length) can be Half-cone counts have been used extensively in used to indicate seed maturity. Assessment of seed British Columbia to determine whether developing maturity may be the objective of the study or may be cone crops are collectable. Recommended collection important for obtaining the best-quality seeds for standards based on filled seeds in the cone half-face another study. are given in Eremko et al. () for Abies amabilis, Because conifer and hardwood seeds can vary so Abies grandis, Abies lasiocarpa, Chamaecyparis greatly, the procedures for collecting, processing, and nootkatensis, Larix occidentalis, Picea glauca, Picea storing seeds of various species are discussed sepa- mariana, Picea sitchensis, Pinus contorta, Pinus rately in Section . according to the characteristics monticola, Pinus ponderosa, Pseudotsuga menziesii, of their fruits. Thuja plicata, and Tsuga heterophylla. .. Description of conifer and hardwood fruits . Determining Fruit and Seed Maturity and The dry multiple fruit of a conifer is called a cone or Quality strobilus (plural strobili). A female cone consists of a central axis supporting overlapping bracts, each of Understanding fruit and seed morphology is vital in which subtends a scale bearing naked seeds. Gymno- designing and implementing a seed production study. sperm, another term for conifer, means “naked fruit,” A brief description of the seed-bearing structures of referring to the fact that conifer seeds are borne

 . Longitudinal and transverse sectioning of cones: (a) longitudinal sectioning of an interior spruce cone; (b) sectioned Douglas-fir cone; (c) transverse sectioning of a lodgepole pine cone; (d) sectioned lodgepole pine cone (Eremko et al. 1989). The number of filled, sound seeds seen in the cone half-face can be used to estimate the total number of seeds in the cone.

46 field studies of seed biology naked on the ovuliferous scales of their cones. A male The seeds of hardwoods are enclosed in protective cone consists of a central axis supporting spirally fruits that vary considerably in size, colour, and arranged microsporophylls bearing pollen sacs that structure. The seeds of bigleaf maple and Oregon ash contain the pollen grains. Conifers in British Colum- are contained in dry winged fruits called samaras. In bia produce several to numerous seeds in a single maple, two winged seeds are joined to form a V, but cone. An exception is Pacific yew, whose fruit is a red, in ash, each fruit contains only a single winged seed. berry-like aril that contains a single “naked” seed. The fruits of red alder, birch, poplar, and willow are

 . Seed-bearing structures of trees occurring in British Columbia

Fruit type Definition Example achene Dry, indehiscent one-seeded fruit. Betula acorn One-seeded fruit of oaks; consists of a cup-like base and the nut. Quercus garryana aril Exterior covering or appendage that develops after fertilization as an Taxus brevifolia outgrowth from the point of attachment of the ovule. berry Pulpy fruit developed from a single pistil and containing one or more immersed Arbutus menziesii seeds, but no true stone. capsule Dry, many-seeded fruit composed of two or more fused carpels that split at Populus, Salix maturity to release their seeds. catkin Spike-like inflorescence, usually pendulous, of unisexual flowers (either staminate Alnus, Betula, Populus, Salix or pistillate). Also used to describe the fruit. Compare strobile. cone Dry multiple fruit of conifers. A female cone consists of a central axis supporting all B.C. conifers, except Taxus overlapping bracts, each of which subtends a scale bearing naked seeds. A male cone consists of a central axis supporting spirally arranged microsporophylls bearing pollen sacs that contain the pollen grains. Syn. strobilus. drupe Fleshy indehiscent fruit, usually one-seeded, containing a seed enclosed in a hard, Cornus, Prunus bony endocarp (pericarp). Syn. stone fruit. nut Dry, indehiscent, one-seeded fruit with a hard wall. Quercus garryana pome Many-seeded fruit of the apple family consisting of an enlarged fleshy receptacle Malus fusca surrounding the papery, fleshy pericarp. samara Dry, indehiscent winged fruit; may be one- or two-seeded. Acer (two-seeded), Fraxinus (one-seeded) strobile Spiky pistillate inflorescence or the resulting fruit; not a true strobilus. Alnus, Betula, Populus, Salix (pl. strobiles) Syn. female catkin. strobilus Male or female fruiting body of the gymnosperms. all conifers, except Taxus (pl. strobili)

Notes: carpel: simple pistil or single member of a compound pistil. imperfect flower: flower which contains either, but not both, functional male or female parts. indehiscent: refers to dry fruits that normally do not split open at maturity. nutlet: nut-like fruit or seed, as in Alnus or Betula. perfect flower: flower that contains both pistil and stamens. pericarp: wall of a ripened ovary that is homogeneous in some genera and in others is comprised of three distinct layers: exocarp, mesocarp, and endocarp. Syn. fruit wall. pistil (or pistillate): the female part of angiosperm flowers, containing the ovary, from which seeds develop. staminate: referring to male angiosperm flowers, containing the stamens, from which pollen is produced. stone: part of a drupe consisting of a seed enclosed in a hard, bony endocarp as in Prunus and Cornus.

section 3 natural seed production 47 catkins (or strobiles), which develop from the spike- like female flowers. The drooping catkins of poplar and willow comprise many capsules that split open at maturity to release many seeds per capsule. The catkins of birch break up at maturity to release the small winged nutlets. The female catkins of alder are woody cones; the cones contain oval nutlets that do not break up at maturity. The fruit of Garry oak is an acorn, consisting of a hard-coated nut in cup- like base. Several hardwoods have fleshy fruits surrounding  . Anatomy of a mature Douglas-fir seed. Ninety their seeds. The fleshy fruit of Pacific dogwood, bitter percent elongation of the embryo is the cherry, and cascara is a drupe, sometimes called a recommended standard for collection stone fruit. A drupe usually contains a single seed (Leadem 1984). enclosed in a hard, bony ovary wall (the stone). The arbutus fruit is a berry, a pulpy fruit developed from a single pistil (female part of a flower) and containing white cedar (Thuja occidentalis) seeds makes it diffi- one or more immersed seeds, but no true stone. The cult to determine if the seeds are filled. Briand et al. Pacific crab apple is a pome, a many-seeded fruit con- () therefore used swelling of the embryo area as sisting of an enlarged fleshy receptacle surrounding a a means of classifying developed from undeveloped papery ovary wall. seeds. Seed colour can also be used as a key to viabil- ity, and is discussed in more detail in Section ... .. Assessing embryo development Destructive methods of seed assessment include The most commonly used indicators of maturity in cutting seeds open to expose the embryo. This proce- conifer seeds are cone and seed colour, degree of cone dure allows for a greater variety of measurements, opening, condition of the megagametophyte, and such as embryo length, embryo cavity length, and length of the embryo (Edwards ; Shearer ; cotyledon length. Alternatively, you can germinate Eremko et al. ). Cones lose moisture as they the seeds and determine the anatomical characteris- mature, and cone colour usually changes from green tics of the embryos. Cotyledon numbers of ponderosa to brown. In the field, specific gravity of the cones pine were determined by germinating  seeds, and has been used to monitor maturation of Douglas-fir selecting  germinants for scoring (Ager and cones (Shearer ). The rate of maturation is in- Stettler ). fluenced by the number of degree-days (Mosseler See Sections .. and .. for quick tests and ), elevation (Shearer ), and latitude. other viability tests. Conifer seeds should not be collected until embryos fill at least % of the embryo cavity (Figure .). .. Assessing seed colour Although collection can begin when embryos fill % Colour is frequently used as an indicator of both cone of the cavity, collecting seeds when embryos are more and seed maturity. In some instances, the purpose of mature will result in better-quality seeds (Edwards the study may be to assess seed colour as an indicator ; Zasada ; Eremko et al. ). of seed maturity. In other instances, determining ma- Embryo development and size may be determined turity (using colour) may be simply a tool to ensure destructively or non-destructively. Non-destructive that the best quality seeds are collected. methods depend on an external visual assessment Seed colour is one of the more difficult indicators of the seeds. For example, with paper birch it is to quantify, as it relies on the subjective judgement possible to distinguish viable well-filled seeds from of the observer, and cone and fruit colour can vary non-germinable seeds by viewing the seeds with a among different individuals of the population. Sev- dissecting microscope equipped with substage illumi- eral instruments, such as Tristimulus colorimeters nation (Bevington ). The small size of eastern and video imaging systems (McGuire ), can be

48 field studies of seed biology used to quantify seed colour and reduce the subjec- Bract length was measured from the base to the tip of tivity of colour readings. the pointed apex; width was measured at the widest Seed colour variation in ponderosa pine was point of the bract. quantified by constructing a -seed gradient with Large differences in cone morphology may be seeds from the entire collection (one seed from each noted between stands and between years. Abnormal of  trees) (Ager and Stettler ). Trees were then cone morphology may also be observed, for example, scored by comparing the adaxial (exposed) surface “forked” cones, proliferated cones (with needles of five typical seeds from a given tree with the gradi- formed at the apex), and combinations of male and ent. Mottled seeds were evaluated on overall shade. female in the same cone. (See Zasada et al.  for Although there were large colour variations among examples in white spruce.) the populations studied, the seeds of a single tree were remarkably uniform when compared to seeds Seed dimensions from different trees. This uniformity is probably a Measurements of seed size (length, width, and result of the high heritability and maternal control thickness) will depend on the anatomical charac- of seed morphology in pines (Kraus ). teristics of the seed (Figure .). In ponderosa pine, Seed colour can be used as a key to the viability of Ager and Stettler () defined seed length as the willow and poplar. In willow the presence or absence distance between the micropylar and basal ends, and of the embryo can be determined by the dark green width as the maximum distance across the seed of the cotyledons showing through the transparent perpendicular to the long axis. Length and width data seed coat. were based on five seeds per tree.

.. Measuring cone and seed dimensions

a) Cone dimensions Cone length can be measured using vernier calipers (. mm precision) (Caron and Powell a). Depending on the type of data presentation or data analysis, it may be convenient to group measurements of cone length into classes. Bergsten () initially grouped cone length measurements of Scots pine in- to  classes (. mm each) from . to . mm, but subsequently combined them into six length classes. Temperature and humidity may affect some cone measurements. In western larch and subalpine larch, Carlson and Theroux () measured cone length b) and diameter on both wet and dry cones, because moisture differentially influences their shape. They hydrated dry cones by placing them in a chamber at % humidity for  hours, then measured cone diameter at the midpoint along the longitudinal axis of the cone. Cone measurements are sometimes used as stable taxonomic markers to distinguish between species of the same genera, and their hybrids. Carlson and Theroux () measured the length and width of five scales and five bracts of western larch and subalpine  . Tree seed anatomy (longitudinal sections): (a) larch, randomly selected from the middle one-third of red alder, an angiosperm; and (b) Douglas-fir, each cone (measured when dry) to the nearest . mm. a gymnosperm (Leadem 1996).

section 3 natural seed production 49 In western larch and subalpine larch, Carlson and If X-ray equipment is available, seeds can be Theroux () measured seed length, width, and placed on celluloid film and exposed to X-rays (see thickness to the nearest . mm. Width and thick- Section ..). Once developed, the films can be ness were measured at the widest part of the seed, placed on a microfiche viewer (of the type commonly then each seed was sliced longitudinally. The thick- used in libraries). Precise seed dimensions can be ness of the seed coat was measured to the nearest obtained by direct measurement of the projected . mm midway between the base and apex of images. Actual and projected dimensions can be the seed. Sampling was done on  seeds randomly compared to calculate an appropriate conversion selected from each lot, as initial sampling estimates factor. If the size and shape of seeds are suitable, an indicated that this sample size would enable standard overhead light projector and  mm camera film can errors within % of the mean with % confidence. be used in a similar manner. Briand et al. () used a dissecting microscope Anatomical measurements were made on white equipped with an ocular micrometer to measure the spruce seeds by cutting the seeds longitudinally and small seeds of eastern white cedar (Thuja occidentalis). measuring the embryo length, embryo cavity length, Seeds were positioned such that the micropylar end and cotyledon length with a micrometer mounted in was facing up and the concave face of the seed was the eyepiece of a binocular microscope (Zasada ). towards the viewer. The following measurements When multiple embryos were present, embryo meas- were determined to the nearest . mm: length and urements were made on the dominant embryo. width of the seed and the embryo area, length of the Samples consisted of  white spruce seeds taken from right wing, and right wing width measured at the the central portion of four cones from each tree. In midpoint (Figure .). many conifers, seeds at the apical and basal portions Extremely small seeds of Salix and Populus, which of the cone are poorly developed (Bramlett et al. ). can be especially difficult (and tedious) to measure, can be graded by sifting them through a set of soil .. Estimating seed weight and volume screens. Although this method is not as precise as Seed weight can be expressed as the fresh weight (fw) using a micrometer, it is effective and less expensive or dry weight (dw) of seeds. The expression used for (J. Zasada, pers. comm., ). seed weight will depend on the context in which it is used. International seed testing rules prefer the use of fresh seed weight (before drying in an oven), whereas ecologists more often use the dry weight of seeds. See Section .. where fresh weight and dry weight are more thoroughly discussed. Seed weights should be determined to at least two significant figures. The sample size required to estimate seed weight varies with the species and the variability of the crop. For example, two -seed rep- licates per tree were used by Ager and Stettler () to determine the weight of ponderosa pine seeds. International standards for sample sizes for weight measurements may be found in International Seed Testing Association () or the Association of Offi-   . Outline drawing of a typical seed of Thuja cial Seed Analysts ( ). Seed weights of tree species   occidentalis (Briand et al. 1992) showing occurring in British Columbia are listed in Table . . significant seed dimensions. LEA = length of For serotinous cones of species such as jack pine and embryo axis; W = width of entire seed; LRW = lodgepole pine, the volume of cones can be determined length right wing; WRW = width right wing; by immersing individual cones in a graduated cylin- WEA = width embryo axis. der containing water and a wetting agent (Rudolph et

50 field studies of seed biology al. ). A similar procedure has been used effec- studied. Conifers generally require some effort to tively for white spruce cones (Zasada et al. ). extract the seeds from the cone, and hardwood seeds are enclosed in a hard or fleshy fruit which must be . Collecting and Processing Seeds removed to obtain the seeds. Another factor affecting seed collection is the In many studies, seeds are an end product by which capacity of seeds for long-term storage. All conifer successful reproduction is assessed. Thus, efficient seeds and many hardwood seeds can retain viability methods of collecting, extracting, and storing seeds for long periods if seed moisture content (mc) is must be known. The method selected for collecting reduced to low levels (–%) and the seeds are and extracting seeds depends on the species being stored at subzero temperatures. Such seeds are called

 . Seed sizes of tree species occurring in British Columbia

Seeds per gram Seeds per gram Scientific name Average Range Scientific name Average Range

 Tsuga heterophylla 655 416–1119 Abies amabilis 25 18–36 Tsuga mertensiana 251 132–458 Abies grandis 50 26–63 Abies lasiocarpa 85 52–108 Chamaecyparis nootkatensis 240 145–396  Juniperus scopulorum 60 39–93 Acer macrophyllum 7 6–8 Larix laricina 701 463–926 Alnus rubra 1468 844–2396 Larix lyallii 313 231–359 Arbutus menziesii 570 434–705 Larix occidentalis 302 216–434 Betula papyrifera 3040 1344–9083 Picea engelmannii 300 152–709 Cornus nuttallii 10 9–13 Picea glauca 405 298–884 Fraxinus latifolia 18 13–21 Picea mariana 890 738–1124 Malus fusca 119 Picea sitchensis 465 341–881 Populus balsamifera 3766 3583–3949 Pinus albicaulis 6 5–7 ssp. balsamifera Pinus banksiana 290 156–551 Populus balsamifera 1652 1233–2070 ssp. trichocarpa Pinus contorta var. contorta 263 225–300 Populus tremuloides 8353 5984–10 707 Pinus contorta var. latifolia 263 225–300 Prunus emarginata 15 9–19 Pinus flexilis 10 7–14 Quercus garryana 0.19 0.17–0.22 Pinus monticola 60 31–70 Rhamnus purshiana 27 11–42 Pinus ponderosa 25 15–51 Salix amygdaloides 5720 Pseudotsuga menziesii 85 63–117 var. glauca Salix bebbiana 5500 Pseudotsuga menziesii 95 65–100 Salix discolor no data available var. menziesii Salix exigua 22 000 Taxus brevifolia 34 32–36 Salix lucida ssp. lasiandra 25 000 Thuja plicata 915 447–1307 Salix scouleriana 14 300

Sources: Stein et al. 1986; Wyckoff and Zasada [1998]; Zasada and Strong [1998]; Zasada et al. [1998].

section 3 natural seed production 51 orthodox in their storage behaviour. Some hardwood berries are usually collected in the fall by stripping seeds do not store well, remaining viable only for or picking by hand directly into bags or baskets, or several weeks up to – years. These seeds are called by shaking or flailing the fruits from the plant onto recalcitrant. They must be stored at relatively high a canvas spread on the ground (Johnsen and moisture content (–%) and above zero tempera- Alexander ). tures, and may require other special handling. With all collection methods, safety precautions must be rigorously maintained. Safety belts and .. Conifer seeds straps must be checked at least twice each day. Tools such as pruning poles and cone rakes should not be Collecting conifer seeds carried while the tree is being climbed. For aerial Conifer cones may be collected by climbing (Yeatman collections, the helicopter company must be certified, and Nieman ) or felling trees. Many aerial cone and the pilots appropriately qualified. Aerial collec- collecting techniques are also available (Camenzind tion operations in British Columbia are subject to ). Aerial methods are much more efficient, Workers’ Compensation Board regulations; make especially for species that produce cones in the upper sure that you have access to current regulations crown, but collection costs are much higher since the appropriate for the area, and confer with persons use of a helicopter is required. Advantages and disad- experienced in cone collection operations. vantages of various cone collection methods may be found in Camenzind (). The choice of method Extracting conifer seeds for specific cone collection projects depends both on Serotinous cones such as black spruce may require the crop and the techniques available. Factors to a period of high temperature to open the cone consider include species, crop size, quantity of cones scales, and sometimes may need multiple extrac- to be collected, site characteristics, the capabilities of tion cycles, as for example, the procedure used by each harvesting technique, safety, efficiency, and cost. Haavisto et al. (): For relatively small trees, and where conditions . soak cones in lukewarm water for  hours, permit, cones and fruits can be collected using a fruit . oven-dry cones at °C for – hours, and picker with a hydraulic lift. If cone-bearing regions . tumble cones in a revolving screened drum for are clearly visible, branches can be shot down with  minutes. a rifle. Occasionally, cones may be collected from squirrel caches, but it is not recommended because Using this procedure, an average of eight seeds per the seeds may be infected with moulds and other cone still remained after the th cycle (average seeds/ pathogens (Sutherland et al. ). cone = ). Collecting Pacific yew and Rocky Mountain juniper Note that the application of this procedure and the seeds requires strategies different from most other ones that follow will depend on the degree of serotiny British Columbia conifers. Both Pacific yew and of black spruce cones (see Section ..). Rocky Mountain juniper are dioecious and bear Mosseler () used only two seed-extraction their fleshy fruits only on female trees. cycles (sec) to remove most of the seeds from black The fruit of Pacific yew, which ripens in late sum- spruce cones. The first sec consisted of oven drying mer or autumn, consists of a red, fleshy, cup-like aril at °C for  hours. A second sec was conducted bearing a single hard seed. To prevent losses to birds, after a -hour water soaking treatment, which was yew fruits should be picked from the branches by followed by drying at °C for  hours. Few seeds hand as soon as they are ripe (Rudolf ). remained in the cones following this extraction pro- The scales of the female flowers of Rocky Moun- cedure, and no further attempt was made to retrieve tain juniper become fleshy and fuse to form small, the remaining seeds. Seeds were counted with an indehiscent strobili commonly called “berries.” electronic counter and were judged to be filled if they Immature berries are green; ripe berries are blue and sank in % ethanol. Verification of ethanol separation covered with a white, waxy bloom. The fruit coat of was made by cutting sample seeds from the filled seed Rocky Mountain juniper is thin and resinous. Juniper fraction, and crushing seeds from the empty fraction.

52 field studies of seed biology Although multiple extraction cycles were also improvised means in a well-vented laboratory oven used, the method used by Caron and Powell (a) with a circulating fan, over a hot-air register or radia- differs significantly from the previous two, in that the tor, or similar location. black spruce seeds are not heated during extraction. Cones should be shaken or slowly tumbled to Instead, after the cones were shaken individually in a extract the seeds from the opened cones. Small-lot covered jar to dislodge seeds, the remaining seeds collections of seeds can be efficiently extracted using were extracted with forceps. This shaking and seed- a multiple compartment tumbler-drier (Leadem and extraction step was repeated two or three times until Edwards ). Although some wings are loosened all seeds were extracted. Cone scales were separated during tumbling and preliminary cleaning, many into three general categories (basal, central, and conifer seeds must be dewinged. Wings of most pines apical) before being counted. Central scales, which and spruces separate readily from their seeds; the spread apart considerably on cone drying to permit wings are hygroscopic, so slight misting can facilitate easy release of seeds, were considered potentially seed their removal. For Douglas-fir, larch, and true firs, bearing (fertile). The extracted seeds were separated wings can be gently broken. Wings cannot be into filled and empty seeds by alcohol flotation removed from western redcedar or yellow-cedar (% ethanol) after dewinging. X-ray analysis (see without damaging the seeds. Section ..) indicated that .% of the seeds that Wings can be removed from small quantities of sank contained well-developed megagametophyte seeds by rubbing the seeds between the hands or tissue and a fully developed embryo, whereas .% against a screen or roughened surface. The same of those that floated were empty or had a rudimen- principle is employed for larger quantities by gently tary embryo. Empty and filled seeds were counted tumbling dry or wetted seeds in a rotating container and weighed to the nearest . mg. Cones (with seed such as a cement mixer. Loosened wings, small parti- wings) were dried in a forced-draft oven at °C for cles, and dust are removed from good seed in final  hours and weighed to the nearest . mg. cleaning. Small lots may be effectively cleaned using In jack pine, which is a predominantly serotinous a laboratory aspirator (Edwards ) or by flotation species, individual cones were dipped in boiling water in water. for up to  seconds to break the resinous bonds The seeds of Pacific yew may be extracted by mac- between cone scales (Rudolph et al. ). The cones erating the fleshy “berries” in water and floating off were dried in a circulating oven at °C until they the pulp and empty seed (Rudolf ). Alternatively, were fully open, after which the cone scales were the fruits can be soaked for – days in warm water, removed and the seeds were extracted by hand. then rubbed over screens and washed thoroughly to Seeds of most conifers (Douglas-fir, larch, western float off light seeds. The viability of yew seeds can be redcedar, western hemlock, etc.) are obtained by maintained for – years if, just after extraction, they drying cones to open them, shaking out the seeds, are dried at room temperatures for – weeks, and separating the seeds from cone scales and debris, then then stored in sealed containers at –°C. loosening the seed wings, and finally separating clean After twigs, leaves, and other debris have been re- full seeds from wings, dust, empty seeds, and other moved with a fanning mill (air separation combined small particles. It may be advantageous to run closed with screens), Rocky Mountain juniper seeds can be cones over sorting tables or screens to remove foliage extracted by running the fruit through a macerator and debris before the cones open. On freshly picked with water and floating away the pulp and empty seeds cones of many species (e.g., Abies), pitch is soft and (Johnsen and Alexander ). Dried fruits should be sticky. Chunks of pitch that become attached to soaked in water for several hours before macerating. extracted seeds may be extremely difficult to remove. Seeds should then be dried to less than % moisture Therefore, true firs should not be heated, but left content (mc) and stored between – and +°C. under cool, dry conditions on trays to disintegrate For additional information on conifer seed collec- naturally. Most other conifer species require only tion, processing, testing, and storage, refer to Stein et good ventilation and slight heating for several days to al. (), Edwards (), Eremko et al. (), and open the cones. Small lots of cones can be dried by Leadem et al. ().

section 3 natural seed production 53 .. Hardwood seeds Another good index of maturity is the presence of Hardwoods are more variable than conifers in the a firm, crisp, white, fully elongated seed within the time of flowering, seed maturation and dispersal, samara. The clusters can be picked by hand or with the type of seed-bearing structures (fruits), and the pruners and seed hooks. Fully dried samaras may be number of seeds per fruit (Tables ., ., .). Many shaken or whipped from branches of standing trees hardwoods are dioecious; in species such as ash, onto sheets spread on the ground. aspen, willow, and cottonwood, seeds are only pro- After collection, leaves and other debris can be duced on female trees. Since the fruits of species such removed by hand-stripping, screening, or using a as maple or ash contain only one seed, collection of fanning mill. Since the pubescence on the pericarp hardwood seeds may be more labour intensive. For can be very irritating to the nose and skin, a face convenience of discussion, the maturation, collection, mask and rubber gloves should be used when work- and processing of hardwood seeds is discussed by ing for extended periods with bigleaf maple seeds. fruit type. Maple seeds generally are not extracted from the samaras following collection. However, dewinging Samaras (Acer, Fraxinus) reduces weight and bulk for storage, since wings ac- Bigleaf maple seeds are double samaras, which turn count for about –% of samara weight (Zasada from green to reddish brown when ripe. The pericarp and Strong []). Empty samaras can be removed has a dry, wrinkled appearance when fully mature, readily on a gravity table. and the surface is covered with dense, reddish-brown Fraxinus samaras should be spread in shallow lay- pubescence. Within the pericarp is an embryo with ers for complete drying, especially when collected associated seed coats, but there is no endosperm. early (Bonner ). Dried clusters may be broken Seed collection may begin when the Acer samaras are apart by hand, by flailing sacks of clusters, or by fully ripened and the wing and pericarp have turned running fruits through a macerator dry. Stems and tan or brown (Zasada and Strong []). Acer seeds other debris can then be removed by fanning or with may be picked from standing trees or collected by air-screen cleaners. shaking or whipping the trees and collecting the samaras on sheets of canvas or plastic spread on the Catkins (Alnus, Betula, Populus, Salix) ground. Samaras may also be collected from trees Birch catkins should be collected while strobiles are recently felled in logging operations, and sometimes still green enough to hold together, or immediately gathered from the surface of water in pools or streams. after a rain to keep them from shattering (Brinkman Bigleaf maple seeds should be collected before a). In Populus and Salix, catkins should be the fall rains. Once the fall rains start, seed moisture collected as close to the time of seed dispersal as content () may increase from  to % (dry weight possible (Wyckoff and Zasada []; Zasada et al. basis) to as high as %. If bigleaf maple seeds re- []). Timing of collection can be based on catkin main attached to the tree, they may germinate colour (which changes from green to yellow or (Zasada ). Moisture also affects the longevity of yellow-brown) and the condition of the capsule. It bigleaf maple seeds, which apparently can exhibit is often best to wait until a few capsules start to split either orthodox or recalcitrant seed properties (Figure ., stage b) and then collect catkins from (Zasada et al. ; J. Zasada, unpublished data). The the plant, since this usually results in the most rapid significance of collecting before or after the start of opening and efficient seed extraction. Note that fall rains is that bigleaf maple seeds with low  be- insect-damaged capsules may appear to be dispersing have more like orthodox seeds, while seeds collected seeds, but are often still immature. Once capsules at high mc have characteristics similar to recalcitrant begin to open, the rate of seed dispersal is deter- seeds. The pubescent pericarp may play an important mined by weather conditions; under warm, dry, role in the moisture content of the samaras. windy conditions all seeds may be dispersed within Ash fruits occur in clusters of one-seeded samaras, a few days. and are collected in fall when their colour has faded If only limited numbers of seeds are needed, from green to yellow or brown (Bonner ). branches with attached, immature catkins of Populus

54 field studies of seed biology and Salix can be collected and ripened in a green- Populus catkins should be spread out in thin layers house or controlled environment (Wyckoff and in pans or on screens at room temperature (Wyckoff Zasada []; Zasada et al. []). Catkins must be and Zasada []). Seeds will be shed in – days, handled carefully after they have been removed from depending on the ripeness of the catkin. Seeds can be the tree. During transport catkins should be loosely extracted from the catkins with a shop-type vacuum packed in paper bags to allow for drying. Catkins cleaner with a clean cloth bag substituted for the dust placed in a warm dry spot will open in a few days, bag. Populus seeds can be freed from their cotton by and seeds can be collected as the capsules open. tumbling the seeds in a rotating drum or a stream of Since alder catkins do not disintegrate at maturity, relatively high-pressure air. For small quantities of they may be collected from standing or recently felled seeds, the uncleaned seeds can be placed between two trees as soon as the bracts (scales) start to separate on soil sieves and a high velocity air stream applied to the earliest-ripening strobiles. tumble the seeds in the container. Seeds should be After collection, catkins from Betula papyrifera extracted and placed in subfreezing storage (- to can be air dried on newspapers at room temperature -°C) as soon as possible, since seeds stored at –°C (–°C), and the achenes separated from catkin lose viability quickly. Storage with a desiccant appears bracts using a series of standard sieves, or with an air- to provide long-term benefit for Populus seeds driven seed blower (Bevington ). Seed samples (Wyckoff and Zasada []). can then be stored dry in sealed containers at -°C Salix catkins should not be left at ambient tem- until used. peratures, and seeds should be extracted and stored

a) e)

f)

b)

c)

g)

d)

 . Salix capsules at various stages of opening (a-e) and the dispersal unit at various stages (f,g) (Zasada et al. [1998]). (f) shows hairs while still in the capsule; (g) shows hairs fully deployed and separating from the seed. Seeds should be collected when capsules start to split (b).

section 3 natural seed production 55 at low temperatures as soon as possible (Simak ). acorns should be kept under moist, cold conditions. The seeds should be separated from the cotton to As a member of the white oak group, Garry oak ex- reduce bulk, and because storage with the cotton hibits recalcitrant storage behaviour (Section .), so may reduce viability (Simak ). To clean small- to the mc must not drop below –%. medium-sized lots, place catkins in a single layer in screen-covered boxes in a relatively warm, dry area Drupes (Cornus, Prunus, Rhamnus) (–°C, –% relative humidity), with good air Dogwood fruits are ovoid drupes which ripen in fall. circulation (Zasada et al. []). If capsules are be- To reduce losses to birds, fruit should be collected as ginning to open when collected, opening will be soon as ripe by stripping or shaking from the branches. completed in ‒ days. The seeds separate easily from Ladders may be useful for collecting fruit from taller the cotton if the catkins and the cotton containing trees (Brinkman a, b). the seeds are placed in a container, so the material Bitter cherry fruits should be collected in late can be blown in an air stream or tumbled in a cement summer or early fall when fully mature and dark red. mixer. Seeds can be separated from coarser and finer Fruits are collected by hand-stripping, or by spread- residue by passing them through a screen or sieve. At ing sheets of suitable material under trees to catch the this time seed mc should be close to the –% rec- natural fall or fruits shaken off the trees (Grisez ). ommended for storage (Simak ). To maintain Fruit may be carried in bags, but boxes or baskets maximum viability, seeds should be placed in sealed provide better protection against bruising and spoilage. containers and stored between - and -°C. Cascara fruits should be picked in late summer or Note that the seed quality of both Salix and fall. The fruits are relished by birds so they should be Populus can be graded to a certain degree by passing harvested about  weeks before they are fully ripe seeds through a nest of soil sieves. In general, the (Hubbard ). largest and best seeds will be found on larger sieve To extract seeds of fleshy fruits, most species can openings (J. Zasada, pers. comm., ). be macerated in a blender. Maceration can be facili- Alder strobiles will open after being exposed in tated by softening fruits for – days in running water drying racks in a well-ventilated room for several (or with daily water changes). The mixture is then weeks at ambient air temperature (Schopmeyer ). placed in water to separate the pulp and empty They can be opened in a shorter time by drying them seeds from the good seeds by flotation. Seeds are in a kiln at –°C. Most seeds will fall out of the thoroughly air dried and placed in sealed containers strobiles during the drying process; however, the for storage at –°C. remaining seeds may be extracted by shaking or Dogwood stones can be sown without extracting tumbling if necessary. them from the fruit, but seeds to be stored usually are cleaned to reduce bulk (Brinkman b). If fruits Nuts (Quercus garryana) cannot be cleaned soon after collection, they should Garry oak acorns are brown when they ripen in late be spread in shallow layers to prevent excessive heat- summer and early fall; they may be collected from ing, although slight fermentation may facilitate the ground, or flailed or shaken from branches onto removal of the pulp. The stones can be extracted by canvas or plastic sheets (Olson ). Garry oak macerating the fruit in water and allowing the pulp belongs to the white oak group, which is character- and empty stones to float away. Clean, air-dried ized by seeds with little or no dormancy, so acorns stones may be stored in sealed containers at –°C. should be collected soon after they have fallen to For bitter cherry it is usually desirable to clean retard early germination. seeds of all pulp and juice (Grisez ). Cleaning is The only processing required before storing or sow- done by macerators with water to float off or screen ing Garry oak acorns is removal of loose cups, twigs, out the pulp. Small quantities may be cleaned by and other debris (Olson ). However, the propor- soaking and rubbing over a screen. Fermentation tion of sound seeds can be increased by removing has been used to soften fruit, but germination may defective, hollow, and partially consumed acorns, be severely reduced if seeds are allowed to become either by flotation or by hand. To retain viability, too warm or to ferment too long.

56 field studies of seed biology Cascara fruits can be allowed to decay for a few collected either by picking the fruit from the tree or by days to soften the pericarp, but usually fruits are run gathering fallen fruit from the ground (Crossley ). through a macerator with water soon after collecting, Pacific crab apple seeds may be extracted by then the pulp is skimmed off (Hubbard ). putting the fruits through a macerator with water, floating off the pulp, and screening out the seeds. Berries (Arbutus menziesii) Seeds should be dried to less than % mc and stored The fruit of arbutus is a berry with a thin, rough, at –°C (Crossley ). granular skin, which is bright red or orange red when ripe. Berries can be collected from standing trees . Assessing Factors that Reduce Seed Yields from October to December (Roy ). Arbutus berries can be dried at room temperature Seed yields are sometimes lower than expected or or seeds can be separated from the pulp immediately predicted and we must identify when or why these after being collected. Fresh or dried fruit can be losses occur, either to verify the value of predictive soaked in water in a warm place to soften the pulp. equations or to prevent future losses. In this section, Fruits then can be macerated and the seeds separated we examine the effects of serotiny and predation on from the pulp by flotation. Seeds or uncleaned ber- seed yields. Seed crop losses may occur due to envi- ries should be thoroughly dried, then stored in ronmental factors, disease, or animal predation, and airtight containers at –°C (Roy ). can be analyzed using life tables (Figure .). Life tables quantify the magnitude and sources of loss Pomes (Malus fusca) and are helpful in interpreting seed crop failure. The pomes of Pacific crab apple are yellowish to reddish Life tables might also be applied to hardwood when they ripen in late fall. Ripe crab apples may be flower production and seed development.

Age interval Number Mortality Number Percent (months) cones alive factors dying mortality

0–1 1182 C. pinus pinus 37 3.13 Abortion 27 2.28 Missing 1 0.08 Breakage 2 0.17 Unknown 9 0.76 insects 76 6.42

1–2 1106 Abortion 13 1.10 Shoot borer 23 1.95 Missing 1 0.08 37 3.13

2–5 1069 Abortion 122 10.32 Missing 5 0.42 Squirrel 4 0.43 131 11.08

Conelets remaining 938 Total mortality 244 20.54

 . Partial life table for 1981 jack pine conelet crop, Oneida County, Wisconsin (adapted from Rauf et al. 1985). Life tables quantify the magnitude and sources of loss and are helpful in interpreting seed crop failure.

section 3 natural seed production 57 .. Assessing serotiny Classes of resin-bond rupture Some conifers do not shed their seeds when they . fully open cone scales free over –% of cone surface mature in the fall, and instead may retain their seeds . partly open scales free over –% of cone surface in the cones for several years. This is termed serotiny, . closed scales free over –% of cone surface but is sometimes called canopy banking (as opposed to seed banking in the soil). Serotinous species retain Classes of weathering as index of age their seeds in tightly closed cones until high tempera- – yr no evidence of weathering tures (such as those achieved in a forest fire) open the – yr weathered over –% of surface cones. The degree of serotiny appears to depend on – yr weathered over –% of surface such factors as the frequency of fire, the local climate, – yr weathered over –% of surface and hybridization between interior populations that + yr weathered over entire surface are predominantly serotinous and coastal popula- tions that are not. Serotiny has great silvicultural .. Assessing predation significance because large quantities of seeds are Seeds represent an excellent food source because potentially available for release after fires or of their stored reserves, thus mature seed crops are harvesting. attractive to insects, birds, squirrels, or other animals. In British Columbia, coastal lodgepole pine is This section primarily describes predation of imma- primarily non-serotinous, whereas interior lodgepole ture seeds (pre-dispersal); for a detailed discussion of pine usually bears serotinous cones (Eremko et al. ). the predation of mature seeds, see Section . In both varieties, cones remain on the trees for many During excellent seed years there are usually more years, but freshly ripened cones have the highest than enough seeds to support both animal predation number of viable seeds. Jack pine are serotinous over and natural regeneration. In moderate years, how- most of their range, although southern sources tend ever, predation can present a problem. Since predator to be non-serotinous. Black spruce cones are semi- populations usually lag a year or so behind abundant serotinous; the cones remain on the tree and the seed crops, a mast year is often followed by a poor seeds are viable for several years (Safford ), year with higher predator populations. Insect and and sometimes as long as  years (J. Zasada, pers. disease damage to seeds and cones may range from comm., ). moderate to severe, and sometimes can result in To estimate the quantity and quality of seeds the loss of an entire seed crop (Miller et al. ; available for regeneration, it may be necessary to Schmid et al. ). Depending on the type of insect assess the age of serotinous cones. Eremko et al. or disease, the attack may occur any time from bud () provide photographic examples of lodgepole initiation to final seed development. Damage may pine cones in different age classes, and recommend result in cone or seed abortion or in partial or com- that, for lodgepole pine, only cones in classes I and II plete destruction of cones or seeds (Mattson ). (i.e., less than  years old) be collected. The cones Effects are sometimes indirect, for example, insects or should be only partially weathered and completely disease may cause the premature opening of cones so closed. that seeds are shed before they are fully developed. Viability of seeds in serotinous cones of harvested Insect predation can alter cone crop phenology trees can decline rapidly, and older cones present (Rauf et al. ) and seed dispersal, and may cause in the slash may have to be discounted as a source conelet and cone mortality. Seed losses due to insect for natural regeneration of a site. Ackerman () predation can be determined by dissecting the cones found that  years after logging there was a sub- and examining cone length, width, and the number stantial decrease in the germination percentage of of sound, hollow, and insect-damaged seeds (Schmid seeds. To conduct his study Ackerman devised et al. ). The percentage of seeds damaged in each scales to classify the degree of serotiny and degree cone may vary, depending upon the insect species. of weathering of lodgepole pine cones present in In areas where squirrel predation can have a major logging slash: impact on natural seed production, it is advisable to

58 field studies of seed biology collect cones early, but only if seeds can be ripened suf- . Experimental Design ficiently under artificial conditions. Hurly et al. () found that intensive harvesting by red squirrels began “Cheshire Puss,” she began, “would you tell me, please, in early September, and most caching occurred in late which way I ought to go from here?” September and October. Early in this period most cones “That depends a good deal on where you want to go to,” cut were eaten rather than cached. Caches are easily said the cat. found; cones can be recovered from the caches within (Lewis Carroll “Alice in Wonderland”)  weeks following the peak of caching behaviour. More information on cone and seed insects is available in Hedlin (), Hedlin et al. (), Ruth A study design is a plan for obtaining the maximum (), and Ruth et al. (). amount of information from available resources (Sit Microbial diseases may also be considered seed ). A good design should begin with clear, well- predation, and substantial cone and seed losses due defined objectives. Three general objectives of natural to disease occur each year. For further information seed production studies are: on cone and seed diseases of North American • estimation (e.g., how many seeds per cone?) conifers, refer to Sutherland et al. () and • modelling (e.g., what is the relationship between Ruth et al. (). seed production and stand, tree, and crown characteristics?) .. Using X-ray analysis to determine causes • comparison (e.g., is seed morphology the same of loss in cliff and swamp areas?) Seed X-rays are a quick and effective way to analyze seed production, but they depend upon the use of .. Estimation studies expensive X-ray equipment. If such equipment is To estimate a parameter such as the average number available, X-radiography can provide non-destructive of seeds per cone or the total number of seeds in a measurements of the number of filled, immature, plot, proper sampling design must be considered to and empty seeds, as well as the numbers of seeds ensure that the estimates are unbiased. Many sam- which have been damaged or attacked by insects (Fig- pling designs can be used, such as simple random ure .). In research studies, comparison of seeds to sampling, cluster sampling, stratified sampling, and their X-ray images facilitates the efficient removal of multistage sampling. For detailed discussions on empty seeds. For detailed procedures, see Section these and other sampling schemes, refer to Cochran .. and Leadem (). (), Thompson (), and Buckland et al. ().

a) b) c) d)

 . X-rays of tree seeds (Leadem 1996). X-rays are used to determine whether seeds are fully developed, damaged, or have been attacked by insects. (a) mature seed: c = cotyledon, m = megagametophyte, r = radicle, s = seed coat; (b) immature seed; (c) insect larva; (d) damaged seed.

section 3 natural seed production 59 Regardless of which sampling design is chosen, Y = a ,  + e b-cX sampling should be done through a random mecha- nism. This will ensure that no systematic bias will be has three parameters (a, b, and c) and requires at least introduced to the data. Sometimes, due to convenience  data points. or convention, an investigator may subjectively select The data collected should also cover the full range samples that are considered typical for the popula- of interest. For example, suppose you want to model tion of interest. These samples, called representative the relationship between cone production and accu- and judgement samples, are discouraged because mulated growing degree-days (gdd). If you want they are subject to personal bias and their statistical to use the model to predict cone production for properties are unknown. Systematic samples are often – gdd, then the data you use in developing taken because of their ease of execution. However, the model must span the range – gdd. The they can be unreliable, especially when the sampling resulting model would only be suitable for predic- scheme coincides with an underlying pattern in the tions within this range; extrapolations beyond the sampling population. It is best not to consider sys- range would be unreliable. tematic samples for estimation. In general, more data points are needed for com- To capture the variability in the population of plex relationships than for simple relationships. interest, the sampling design must provide an ad- equate sample size. The sample size depends on the .. Comparative studies variability in the population, the accuracy desired, In contrast to sampling for estimation and modelling, and the cost. You may want to consider stratifying the comparative studies require an experimental design. collection of samples (e.g., collecting from different In a comparative experiment, treatments are randomly levels of the crown of a tree) to reduce variation and assigned to a number of experimental units (the better understand effects of position (vertical and smallest collection of the experimental material to horizontal). For estimation type studies, the sample which a treatment is applied). If you wish to compare size can be determined using confidence interval seed morphology in two different habitats (e.g., cliff methods. See Section .. for a discussion and an and swamp), five sites each can be selected randomly example of sample size determination methods from all cliffs and swamps within the population of using confidence intervals. interest. Within each site,  trees can be selected for cone measurement. In this example, a site is the ex- .. Modelling studies perimental unit; a tree or a cone is a subsample. To sample for modelling, the sampling guidelines Comparisons based on a single application of the discussed above should be followed. All variables treatments are unreliable because variations are ex- involved in the model must be sampled from the pected between experimental materials. Differences same sampling points. For example, if you want to between a cliff site and a swamp site could be due to relate the number of seeds per cone with the number differences in the habitat, or to natural variation from of exposed seeds in the cone half-face, then the total site to site, or both. The only way to distinguish seeds per cone and the seeds in the half-face must be the possible causes of variation is to replicate the determined from the same cone. treatments. To model a relationship, there must be enough Replication of a treatment is an independent data to capture the relationship between the vari- observation of the treatment. The number of replica- ables. A general rule is to have at least  data points tions is the number of experimental units to which a per parameter involved in the model. For example, a treatment is assigned. Replication should not be straight line model, confused with subsamples, which are multiple meas- urements of a single treatment. In the cliff/swamp Y = a + bX, example, each treatment is replicated five times. The  trees within each site are subsamples. Pseudo- has two parameters, Y-intercept (a) and slope (b), replication occurs when replication is claimed when and requires at least  data points. A logistic model, in fact there is none. Pseudoreplication usually leads

60 field studies of seed biology to underestimation of the variability in the data. See analysis depends on the design plan, while the design Bergerud () for additional discussion of pseudo- plan is strongly influenced by the analysis method replication. deemed to be the most suitable for the data. The The number of replications necessary for a study analysis method should conform with the design depends on the variability in the data, the size of of the study, the nature of the data, and the study difference you wish to detect, the significance level objectives. desired, and the desired statistical power. Power analysis is the computation of statistical power for .. Estimation studies an experimental design, and should be carried out If the objective of a study is sampling for estimation, before the experiment to determine the amount of then care must be taken to ensure that the formulae replication required. For more discussion on the use for computing mean, total, and standard error are of power analysis for sample size determination, see appropriate for the chosen sampling scheme. A com- Cohen () and Nemec (). mon mistake is to use formulae for simple random Random assignment of the treatments to the sampling design in more complex designs, which experimental material is also essential to sound results in underestimation of the standard error of experimentation. Randomization assures that no the estimate. That is, the estimate would appear more systematic bias is introduced to the experiment, reliable than it really is. Nemec () provides an and the natural variation is approximately the same example that illustrates the consequences of using within each treatment group. Sometimes random simple random sampling formulae for data collected assignment of the treatments to the experimental from cluster sampling. material is not possible. In the cliff/swamp example, it is not possible for the experimenter to assign a cliff .. Modelling studies or a swamp to a particular location; an area is a cliff When the objective is to sample for modelling, then or a swamp before the experiment is even conceived. regression and correlation are typical analysis meth- In this case, to satisfy the randomization criteria, cliff ods. Depending on the relationship between the and swamp sites must be randomly selected from variables of interest, linear or nonlinear regression all cliff and swamp sites within the population of may be required. If the goal is to determine the best interest. Subsamples for measurements must also be set of variables for predicting a relationship, then randomly chosen within each experimental unit. stepwise regression can be used to systematically The design of a comparative experiment depends eliminate any unnecessary variables. largely on the treatments to be compared, the experi- Regression assumes that the residuals (the differ- mental material available, and the type of data to be ence between the observed data and the predicted collected. Common experimental designs employed values) are normally distributed, with a mean equal in seed production studies include completely to zero and a constant standard deviation. The nor- randomized design, factorial designs, and random- mal distribution of the residuals can be checked ized block design. For discussions on these and other using a normal probability plot on the residuals. An experimental designs, see Sit (). apparently straight line indicates that the residuals It is vital that the sampling design and experimen- are approximately normally distributed. Regression tal design optimize all essential factors of the study. also assumes that the residuals are: a) independent of Researchers should discuss their designs and analysis the values in the explanatory variables (x-variables), plans with a statistician before implement-ing a study and b) have equal variance for all values of the ex- to ensure that all relevant factors have been considered. planatory variables. The independent and equal variance assumptions can be checked by plotting the . Data Analysis residuals against the predicted values derived from the regression model. A random scatter of the points The success of an experiment requires both a well- implies that both assumptions are satisfied. designed plan and an appropriate analysis method, Violation of the normality and equal variance because the two are closely related. The method of assumptions sometimes can be corrected by

section 3 natural seed production 61 transforming the data using square root, natural log, that is, it explains most of the variation in the data or exponential functions. Transformation should not with the minimum number of variables. Rawlings be done routinely without first checking the resi- () may be consulted for further information on duals. Keep in mind that the regression assumptions regression analysis. See also Sit and Poulin-Costello are for the residuals, not for the data. It is possible () for additional discussions on nonlinear to have non-normal data, but normal residuals. A regression. common mistake is to examine the data and apply If the objective is to assess the strength of the transformation when the data are not normal or relationship between two variables, then correlation have unequal variance. analysis can be used. There are two types of correlation: Regression is a robust procedure against slight Pearson product-moment correlation coefficient, departures from normality and equal variance when r(p), and Spearman’s rank order correlation coeffi- the data set is large. That is, with a large data set, you cient, r(sp). The Pearson correlation assesses the can still use regression on the data (without transfor- linear relationship between two variables (see mation) for slightly non-normal residuals. However, Figures .a and b), and is based on the observed like most statistical procedures, regression is not ro- data. Spearman’s correlation assesses the monotone bust against independence. That is, regression results relationship between two variables, that is, whether are invalid if the residuals are dependent (e.g., when the two variables have a strictly increasing (linearly large residuals tend to associate with large x-values.) or nonlinearly) or strictly decreasing relationship Provision for randomization during data collection (see Figures .c and d). Spearman’s correlation is will ensure that the data, and thus the residuals, are based on the rank order of the data, with tied scores independent. assigned the average of the scores that would have To assess the goodness of fit of a model to the been assigned had no ties occurred. data, the coefficient of determination, r2 value, can be A correlation coefficient must have a value between calculated. The coefficient of determination represents + and -. A positive value implies that the two vari- the proportion of variation in the data explained by ables increase together; a negative value implies that the model. The higher the r2 value, the more varia- one variable increases as the other variable decreases. tion is accounted for by the model. The r2 value is A Pearson correlation coefficient near zero implies directly related to the number of explanatory vari- there is no linear relationship between the two ables in the model: the more explanatory variables variables, but the two variables may be related in a there are, the higher the r2 value. When comparing nonlinear way (see Figures .c, d, and e). A Spear- several regression models, it is more suitable to use man’s correlation coefficient near zero implies 2 the adjusted coefficient of determination (r adj) which that the two variables do not increase or decrease is r2 modified by the number of explanatory variables together, but they may be related in a curvilinear 2   in the model. A model with large r adj is favourable, manner (Figure . e).

a) b) c) d) e)

 . Correlation coefficients for hypothetical relationships. r(p): Pearson product-moment correlation coefficient; r(sp): Spearman’s rank order correlation coefficient.

62 field studies of seed biology .. Comparative studies For discrete data such as seed crop rating or the If the objective is to compare the effects of several number of full and empty seeds on a tree, contin- treatments on seed production, then analytical meth- gency table tests (chi-square test, Sections . and ods should be used. The method chosen depends on ..) or log-linear models could be used. If data are the nature of the data and the design of the study. collected on the same units over time and the objec- For continuous data such as seed weight, seed length, tive is to assess trend, then repeated measures analysis or seed width, methods such as the t-test and analysis methods should be considered. See Nemec () for of variance (anova) F-test could be used for anal- detailed discussions of repeated measures analysis. ysis. Both the t-test and the anova F-test assume normally distributed residuals. This assumption can . Seed Production Case Studies be checked by plotting the residuals in a normal probability plot (see Section ..). If the residuals Six seed production case studies, taken from the are far from normal, then nonparametric procedures literature, are summarized below. To highlight the such as the Wilcoxon tests could be used. Refer to design and analysis aspects, the studies are presented Sections ., .., ., and .. for discussions of in point form. The cautions given at the end of each anova analysis. See Sit () for a detailed discus- case emphasize the items that require special atten- sion of anova. tion to ensure that study objectives are met.

CASE STUDY 1: Estimating potential Engelmann spruce seed production on the Fraser Experimental Forest, Colorado (Alexander et al. 1986) Objectives . Regression was used to relate the number of filled • To predict the frequency of good seed crops. seeds per trap and the total seedfall per trap. • To relate seed production to stand, tree, and or . Stepwise regression was used to select the best set crown characters (sampling for modelling). of stand, tree, and/or crown measures for Study Design predicting seed production.  . The sampling was carried out over a long period . Transformation of the data may be considered to of time (annually for  years). correct for heterogeneity of variance before . Thirteen permanent sample plots with  seed regression; possible transformations are square traps were randomly located in each plot. root and natural log. . Seed trap contents were collected each fall and Cautions again the following spring. • Sampling units should be randomly selected from . Only filled seeds were counted; the response all possible units. variable was the number of filled seeds per trap. • Seed traps should be randomly located within Data Analysis each sampling unit (sample plot). . The sample mean (by plot) was the best estimate of • Seeds from several traps should not be bulked. average number of filled seeds per trap. • If data are collected over a long period of time, . anova was used to test for location and year check whether the model residuals are effects. independent; and consider using time-series . Seed counts were transformed (x + ⅜ ) to models or repeated measures (incorporating lag stabilize variance. variables in the regression).

section 3 natural seed production 63 CASE STUDY 2: Comparative seed morphology of Thuja occidentalis (eastern white cedar) from upland and lowland sites (Briand et al. 1992) Objectives adjust the α-probability level for simultaneous • To test whether a relationship exists between seed comparisons to reduce the risk of Type I error. morphology and the habitat where seeds are . Satterthwaite’s approximation (Zar ) was used produced (cliff and swamp). to compute % confidence intervals based on the • To explain the greater root plasticity among t-distribution. upland seedlings. . Pearson product-moment correlation coefficients Study Design were computed between all characters. . This is a one-factor completely randomized design Cautions with subsamples. • Since cliff and swamp could not be assigned to an . Three cliff sites and three swamps were randomly area, sites were randomly selected from all cliffs selected. and swamps within the population of interest. . Ten trees were sampled at each site; five cones were • Be careful when identifying the experimental unit. collected from each tree for measurements. In this example, an individual site is the experi- . Responses included total number of seeds, mental unit, not tree, or cone. number of fully developed seeds per cone, seed • The trees in the two habitats should be as similar as fresh weight, seed length and width, embryo area possible to avoid confounding the habitat and tree length and width, and right wing length and characteristic factors (e.g., cliff sites had older trees). width. • If the two habitats had trees of different ages, age Data Analysis could be used as a covariate in the analysis of co- . anova was used to compare the responses of the variance, provided that the covariate (age) is not two habitats (at significance level .). affected by the factor of interest (in this case, . Sequential Bonferroni technique was used to habitat).

CASE STUDY 3: Cone size and seed yield in young Picea mariana trees (Caron and Powell 1989a)

Objectives Data Analysis • To investigate the variation in cone size, seed yield . Correlation was used to assess the relationships of per cone, and seed weight from cones collected in the nine response measures.  plantations in three consecutive years. . Regression was used to relate the number of filled • To determine the correlation between cone size, seeds per cone to the number of pollen cones seed yield per cone, and seed weight. per tree; logarithmic transformation was used on • To examine the relationship between pollen the response variables to correct for unequal abundance and filled seeds per cone. variance. Study Design Cautions . Five plantations (, , , , and  years from • Correlation can be used to assess the relationship seed in ), located in northwestern New between two variables, but Pearson correlation can Brunswick, were used in the study. assess only linear relationships. It is possible that . Two study areas were selected within each planta- two variables are nonlinearly related and the tion; trees were randomly selected for measurement. correlation coefficient is near zero. . Responses included cone length, cone weight, total • A variable that shows a high correlation to seed scales per cone, potential filled seeds per cone, yield per cone may not be the best predictor total seeds per cone, total filled seeds per cone, for seed yield; another variable that is nonlinearly seed efficiency, weight of  filled seeds, and related with seed yield may be a better predictor. weight of  empty seeds. • Always plot the data in a scatter plot.

64 field studies of seed biology CASE STUDY 4: Prediction equations for black spruce seed production and dispersal in northern Ontario (Payandeh and Haavisto 1982) Objective was fitted to the data to relate seed viability (Y) • To use nonlinear regression to relate the number with seed cone age (X). of black spruce seeds per cone with cone age and . An exponential decay-exponential model, crown class. B e-B3X2 B e-B6/X2 Study Design Y = B X 2 + B X 5 , . Data on seed production (number of seeds per cone and cone age in years) were collected from was developed to relate estimated seedfall per black spruce in three crown classes: dominant, hectare (Y) with strip width (X) and distance codominant, and intermediate. from stand edge (X). . Two sets of data on seed dispersal across the Cautions stripcuts were also available for modelling. • Enough points are needed to cover the entire Data Analysis X-range. -B X . A simple exponential decay function, Y = Be 2 , • Know the form of the equation, and the deriv- was fitted to the data to relate the total number of atives with respect to the unknown parameters, seeds per cone (Y) with cone age (X) for each and an estimate of the parameters (starting point crown class. for iteration). . An inverse sigmoidal function, • Models should be compared based on adjusted r2. • Do not extrapolate results from the fitted models -B X B Y = B - B(1 - e 2 ) 3, beyond the range of the original data.

CASE STUDY 5: Estimating sound seeds per cone in white spruce (Fogal and Alemdag 1989)

Objectives Data Analysis • To determine whether the number of filled seeds . The mean and coefficient of variation were cal- per cone half-face is a valid indicator of the total culated for each of the four variables for each number of seeds per cone. location and crop year. • To determine the relationship between the num- . anova was use to compare locations and years ber of filled seeds per cone and cone length and based on the number of seeds per cone. diameter. . Scattergrams were prepared for each location and • To determine whether the relationship is the same year to visually assess possible relationships across time and location. between number of sound seeds per cone and Study Design number of sound seeds per section, cone length, . Cone data were collected from three white spruce and diameter.  plantations in  and . . Multiple regression was used to relate the number . Seed counts were made on  cones from each of of sound seeds per cone with the following  trees at each location. independent variables: number of sound seeds per . Measurements included cone length and maxi- section, cone length, and cone diameter. Eight mum diameter, number of sound seeds per section models were fitted to the data. on one cone half-face, and number of sound seeds Cautions per cone. • Use adjusted r2 to compare models, not r2.

section 3 natural seed production 65 CASE STUDY 6: Consistency of cone production in individual red pine (Pinus resinosa) (Stiell 1988)

Objectives Data Analysis • To compare production by stand and by . Linear regression with square root transformation individual trees at two dates ( and ). was used to relate crop size to tree size at both • To relate cone production to stem diameter and dates, and to relate  crop size to  crop size. subsequent diameter growth. . Potential cone production was approximated Study Design using the sum of both mature and aborted cones.     . Data were collected from an -year-old red pine . Relationship between crop size to – plantation that was established as a spacing trial. basal area growth was also analyzed. . Permanent sample plots were established for each Caution of six spacings. • Use adjusted r2 to compare models. . Cone counts were made on mechanically selected, numbered trees in  and .

66 field studies of seed biology