Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modelling windthrow risk in coastal variable retention using tree, neighbourhood, and stand attributes Scott, Robyn Elizabeth 2005

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

Item Metadata

Download

Media
831-ubc_2005-0111.pdf [ 6.68MB ]
Metadata
JSON: 831-1.0074983.json
JSON-LD: 831-1.0074983-ld.json
RDF/XML (Pretty): 831-1.0074983-rdf.xml
RDF/JSON: 831-1.0074983-rdf.json
Turtle: 831-1.0074983-turtle.txt
N-Triples: 831-1.0074983-rdf-ntriples.txt
Original Record: 831-1.0074983-source.json
Full Text
831-1.0074983-fulltext.txt
Citation
831-1.0074983.ris

Full Text

MODELLING WINDTHROW RISK IN COASTAL VARIABLE RETENTION USING TREE, NEIGHBOURHOOD, AND STAND ATTRIBUTES. by Robyn Elizabeth Scott A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTERS OF SCIENCE IN FORESTRY in THE FACULTY OF GRADUATE STUDIES (FORESTRY) We accept this thesis as conforming to the required standard. THE UNIVERSITY OF BRITISH COLUMBIA May 2005 © Robyn E. Scott, 2005 ABSTRACT The adoption of the retention system in much of coastal British Columbia, Canada, has brought with it concern over the windfirmness of the retained trees. The present state of knowledge concerning windthrow risk factors is inadequate for the purposes of risk prediction in partial cuts and in structurally complex forests, such as those that exist in Clayoquot Sound. This study examined the relationship between the occurrence of windthrow after partial-cut harvesting and various stand, neighbourhood, and tree attributes. Measurements of 1215 trees were obtained from 234 sample plots in retention system cutblocks in Clayoquot Sound on the west coast of Vancouver Island, and of 1891 trees from 115 plots in the Silviculture Treatments for Ecosystem Management in the Sayward (STEMS) study site near Campbell River. At the Clayoquot site, 16.5% of trees were windthrown, while at the STEMS site, 5.3% of trees were windthrown. Logistic regression models were fit for both areas using tree, neighbourhood, and stand variables to predict the probability of individual trees being windthrown. The best-fit models for the Clayoquot and STEMS datasets correctly predicted windthrow status of 72% and 94% of the sampled trees, respectively. The proportion of damaged trees at the Clayoquot site increased with increasing tree height-diameter ratio, crown density (sparse, moderate, full), and an index of fetch distance equal to distance multiplied by removal level, and decreased with increasing percent live crown and post-harvest density. At the STEMS site, windthrow decreased with increasing tree diameter. It is recommended that forest managers plan to retain at least 20% of original stand density in areas where windthrow is a concern, and focus retention efforts on trees with low height-diameter ratios, sparse crowns, and high percent live crown. ii TABLE OF CONTENTS ABSTRACT .' i ii TABLE OF CONTENTS iii LIST OF TABLES v LIST OF FIGURES vi ACKNOWLEDGEMENTS . viii CHAPTER 1 - INTRODUCTION 1 WINDTHROW AND VARIABLE RETENTION 1 USING MODELS TO PREDICT WINDTHROW RISK 3 CHARACTERIZING VARIABILITY 4 THE IMPORTANCE OF SCALE 5 RESEARCH OBJECTIVES 6 APPROACH 7 THESIS OUTLINE 7 CHAPTER 2 - LITERATURE REVIEW 9 WIND BEHAVIOUR 9 MECHANICS OF WINDTHROW 10 EFFECTS OF WIND ON TREES 11 FACTORS INFLUENCING WINDTHROW RISK ...13 Climate 13 Topography 13 Soils 14 Stand attributes 15 Stand age, height, and density 16 Species composition 16 Tree attributes 17 Tree height and diameter 17 Height-diameter ratio 18 Crown size and density 18 Species 19 Harvesting 20 Opening size and fetch 21 Retention pattern and level 22 SUMMARY, RESEARCH QUESTIONS AND HYPOTHESES 24 CHAPTER 3 - MATERIALS AND METHODS 26 DESCRIPTION OF STUDY AREAS 26 Clayoquot 26 STEMS : : ' 29 FIELD DATA COLLECTION 30 Clayoquot 30 STEMS 31 GIS AND REMOTE SENSING INFORMATION SOURCES AND DATASET CREATION..... 32 Information sources and data conversion c..'. 32 iii Aerial photos and image warping 32 CREATION OF DERIVED VARIABLES 33 Retention type and level 33 Fetch 36 Spacing factor : 39 Topographic exposure 39 D A T A ANALYSES A N D MODELLING 40 Simple correlation and contingency tables 40 Logistic regression 40 Model fit and comparison of models 41 C H A P T E R 4 - R E S U L T S 43 CHARACTERISTICS OF WINDTHROWN TREES :.43 RANGES AND MEANS OF PREDICTOR VARIABLES 43 CORRELATIONS BETWEEN SELECTED VARIABLES 45 CONTINGENCY TABLES ; 47 MODELS 54 BEST-FIT MODELS 58 EFFECT OF ADDING STAND-LEVEL WINDTHROW RISK . . . 60 C H A P T E R 5 - D I S C U S S I O N . . . . . . — . . . . . 62 DIFFERENCES BETWEEN STANDS 62 MODELS. . . . , , •.....-.„• , ....... : 64 K E Y TREE-LEVEL VARIABLES... . .v 65 K E Y NEIGHBOURHOOD-LEVEL VARIABLES „U 67 SOIL CHARACTERISTICS 69 LIMITATIONS OF METHOD 70 C H A P T E R 6 - C O N C L U S I O N S 71 C H A P T E R 7 - R E C O M M E N D A T I O N S 74 R E F E R E N C E S 76 A P P E N D I C E S . . . . . . . . 83 APPENDIX I: DESCRIPTION OF VARIABLES 84 APPENDIX II: VARIABLE SUMMARIES O V E R A L L PLOTS 87 APPENDIX III: DUMMY VARIABLE CODING FOR CATEGORICAL VARIABLES 89 APPENDIX IV: INTERCEPT AND PARAMETER ESTIMATES FOR INITIAL MODELS 90 APPENDIX V: STAND-LEVEL WINDTHROW RISK MODEL CALCULATIONS 93 APPENDIX Vl: DETAILS OF IMAGES USED TO MAP RETENTION POLYGONS.. 95 APPENDIX VII: EXAMPLE OF FETCH CALCULATIONS 96 i v LIST OF TABLES Table 1. Example of the spatial scales and data sources for variables at the tree, neighbourhood and stand level 5 Table 2. Mean windthrow values reported along clear-cut edges and in different retention types by various windthrow studies in coastal BC 21 Table 3. Variables used to describe retention level 33 Table 4. Retention type classification, and number and mean size of retention polygons 36 Table 5. Description of fetch variables ; 38 Table 6. Summaries over all trees for key variables at the Clayoquot study area (n=1215) .44 Table 7. Summaries over all trees for key variables at the STEMS study area (n=l 891) 45 Table 8. Correlations between selected variables at the Clayoquot study area 46 Table 9. Correlations between selected variables at the STEMS study area. 47 Table 10: Clayoquot models using tree, neighbourhood, and stand variables (Dataset 1 n=608, Dataset 2 n=607) 55 Table 11: STEMS models using tree, neighbourhood and stand variables (Dataset 1 n=942, Dataset 2 n=949) 57 Table 12. Variables, coefficients, and odds-ratios for the best-fit logistic regression models at the Clayoquot and STEMS study site 59 v LIST OF FIGURES Figure 1. Area harvested using retention silviculture systems (bar) as a proportion of total area harvested (0) in the Vancouver Forest Region over the past four years 2 Figure 2. Map of Vancouver Island showing location of two study areas; a) detailed map of Snowden Forest and STEMS block locations; b) detailed map of Clayoquot block locations 27 Figure 3. Total monthly precipitation and mean daily temperatures from the a) Torino Airport weather station, and b) Campbell River Airport weather station 28 Figure 4. Maximum hourly wind speed (km/hr) recorded in each month during the time period 1971-2000 for Torino and Campbell River weather stations 29 Figure 5. Example of retention polygon mapping from georeferenced photos in Arc View; a) external block boundary only, b) mapped internal retention polygons 35 Figure 6. Location of points used to determine retention levels for the purpose of calculating fetch 38 Figure 7. Illustration of distance-limited TOPEX for a north (N) to south (S) transect located in a valley bottom 39 Figure 8. Proportion of windthrown trees at each study area that fell towards each of the eight cardinal directions; a) Clayoquot (n=198), and b) STEMS (n=97) 43 Figure 9. Proportion of trees wnidthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area (n=1215) by: a) stand structure, b) lead species by volume 48 Figure 10. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the stand level 49 Figure 11. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the neighbourhood level 51 Figure 12. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the tree level 52 Figure 13. Proportion of trees windthrown (bar) and number of trees sampled (0) for selected variables at the STEMS study site 54 v i Figure 14. Observed windthrow versus predicted probability of windthrow for best-fit Clayoquot and STEMS models 60 Figure 15. Proportion of trees windthrown (bar) and number of trees sampled (0) versus predicted risk of windthrow after clearcutting using stand-level models developed by Mitchell (2003) and Lanquaye (2003) at a) Clayoquot, n=1215, and b) STEMS, n=l 891.61 v i i ACKNOWLEDGEMENTS A thesis is a major undertaking, and this one would never have been completed without the help and support of several people. Foremost among these is my supervisor, Steve Mitchell. Steve, thank you for being so enthusiastic about this entire process, and for providing me with a clear sense of direction and many opportunities for learning. I'd like to thank my committee members, Bruce Larson and Valerie Lemay, for their advice and contributions throughout. Mike Vorhies was an excellent field assistant, and helped to make the field sampling in Clayoquot Sound go smoothly. Naa Lanquaye and Yolanta Kulis provided much-needed help with the many aspects of GIS and Arc View that I did not fully understand. Louise DeMontigny kindly provided the datasets from the STEMS study, as well as patient explanations of the contents. I'd like to thank my family, John, Laurie, Theresa, Jennifer, Stephanie, and especially Dave, for all of their support and the many encouraging phone calls, emails, and hugs when I needed them. Finally, my sincere appreciation goes to International Forest Products Ltd and the Natural Sciences and Engineering Research Council, who provided financial support for this project. R.S. viii CHAPTER 1 - INTRODUCTION WINDTHROW AND VARIABLE RETENTION Tree mortality due to mechanical damage by wind is often called windthrow; it includes trees uprooted or broken by wind. Windthrow is a dominant form of natural stand disturbance in wet coastal forests, and a challenge for forest managers. Windthrow results in economic losses to forest companies due to the higher costs of salvage logging. If left unsalvaged, windthrown trees increase the potential for insect outbreaks (Mitchell 1995). Severe windthrow can disrupt planning and may limit the types of treatment or retention prescribed for certain sites (Stathers et al. 1994). Windthrow has been identified as a significant threat to the successful use of partial-cutting and other complex forms of stand management (Coates 2001, Rueleftf/. 2003). Changing social values have resulted in a new philosophy of ecosystem management and sustainable resource use in the coastal forests of British Columbia (BC), Canada, and the adoption of new forest practices. Unprecedented public protests over logging practices and land use decisions in Clayoquot Sound in the 1980's and early 1990's led the BC government to convene the Scientific Panel for Sustainable Forest Practices. The panel was charged with reviewing forest practices standards and recommending changes to "make forest practices in Clayoquot not only the best in the province, but the best in the world" (Clayoquot Sound Scientific Panel 1995, p.l). The panel recommended a 'variable-retention' harvesting strategy, which is essentially partial-cutting with varying amounts and patterns of retained trees. The resulting stand is designed to maintain higher visual quality, structural heterogeneity, and ecosystem function; thus, variable retention is intended to meet ecological as well as economic objectives. 1 In 1996, International Forest Products Ltd. began harvesting in Clayoquot Sound using a variable retention approach, and in 1998, MacMillan Bloedel announced that the forest company would phase out clearcutting in its coastal operations in favour of retention systems. In 1999, the retention silviculture system was formally defined in legislation (Forest Practices Code of BC Act 1999, Mitchell and Beese 2002). Since then, several other coastal forest companies have adopted this approach, and the area on the coast harvested using retention systems has grown steadily (Figure 1). In this thesis, the phrase 'retention system' is used to describe the method of harvesting and is used interchangeably with 'partial-cutting', although the latter is a broader term. The phrase 'variable retention' is used to describe the stand resulting from the use of retention systems (Mitchell and Beese 2002). o> 0-3° c '55 3 tn | | 0.20 I » <0 c sz o I fo.io o *-a o * 0.00 4 • • * IBB I P E X l B i s SI mm :'?• »<i -wgm -s v: . . • i s i i |S * 'i iiMiiil iHii 111 i 1 1 1 1 40000 CO 30000 -g % I 20000 « 10000 « (0 1999/2000 2000/2001 2001/2002 2002/2003 Fiscal Year Figure 1. Area harvested using retention silviculture systems as a proportion of total area harvested (bar), and total area harvested (0) in the Vancouver Forest Region over the past four years. Data compiled from statistical tables in BC Ministry Of Forests Annual Reports, 1999/2000 through 2002/2003. 2 USING MODELS TO PREDICT WINDTHROW RISK A key approach to windthrow management has been to assess the risk of windthrow through observational approaches, and more recently, using mechanistic and empirical models. Observational approaches identify factors that predispose a stand to windthrow - the more factors observed, the higher the windthrow risk. Stathers et al. (1994) and Mitchell (1998) described a qualitative observational framework for evaluating windthrow hazard in BC forests. Mechanistic modelling uses the results of tree wmching and wind tunnel studies to predict critical windspeeds at which trees will fail, and combines these predictions with topographic windspeed models to determine where windthrow will occur (e.g., Gardiner and Quine 2000, Ruel et al. 2000, Talkkari et al. 2000). These models have been used to predict windthrow in uniform plantation forests in Great Britain and Finland, and in balsam fir (Abies balsamea (L.) Mill.) stands in Quebec. The variability in tree size and health, the variability in stand structure and composition, and the lack of windspeed data for remote areas limits the use of mechanistic models in BC. Empirical models have provided a quantitative approach that is better suited to complex stands (Mitchell et al. 2001). In empirical models, regression equations are developed that relate windthrow severity to site, stand, and/or tree attributes. Empirical windthrow models have been developed for strip-cut black spruce (Picea mariana Mill.) in Ontario (Elling and Verry 1978), and for stand edges exposed by clearcut harvesting in coastal BC (Mitchell et al. 2001, Mitchell 2003). Although pilot windthrow studies have been undertaken in areas harvested using retention systems (e.g., Scott and Beasley 2001, Rollerson et al. 2002), no attempt has been made thus far to develop empirical models for individual trees in variable retention stands. 3 Although useful for prediction and planning, it is important to remember that models are incomplete and simplified representations of highly complex processes. To be credible, empirical windthrow risk models should rely on variables that are functionally related to the processes involved, that capture existing variability and complexity, and that are relatively easy to measure. The knowledge gained from empirical models should be used to inform mechanistic model development, and be incorporated into tools for managing windthrow. CHARACTERIZING VARIABILITY A major challenge in monitoring the outcome of retention system harvesting is describing the structural variability that is maintained. The Clayoquot Sound Scientific Panel (1995) recommended that variability be characterized by specifying: 1) the structural elements to be retained, 2) the spatial distribution of those elements, and 3) the amount to be retained. Commonly retained structural elements include live and dead trees of various sizes, and coarse woody debris. Spatial patterns of retention are generally classified as dispersed or aggregated (Mitchell and Beese 2002). Dispersed retention consists of individual or small groups of trees that are evenly distributed across the harvested area. In aggregated retention (also known as group retention), larger groups or patches of trees are retained. Various combinations of the two types may also be used. The level of retention may vary, and prescribed retention levels are specified in various ways. Retention level is commonly expressed as a percentage of the initial stems per hectare or basal area of the stand, or as a percentage of the total area of the cutblock. The widespread implementation of partial-cutting on Vancouver Island allows comparisons to be made between outcomes in different stand types. In this study, outcomes in the heterogeneous stands on the west coast are contrasted with outcomes in more homogeneous second growth stands found on eastern Vancouver Island. The cutblocks sampled in Clayoquot 4 Sound were operationally harvested and represent a range of retention levels and patterns. Single cutblocks often contained multiple retention levels and patterns. The Silviculture Treatments for Ecosystem Management in the Sayward (STEMS) site on eastern Vancouver Island is a designed silvicultural systems experiment and each opening represented a single retention level and pattern. T H E IMPORTANCE OF S C A L E Choosing the right spatial scale for characterizing variability and measuring windthrow and other attributes is an additional challenge (Table 1). Previous empirical modelling in BC Table 1. Example of the spatial scales and data sources for variables at the tree, neighbourhood and stand level. Level Spatial scale Source Variables Tree Several meters field data Species Height Crownclass Percent live crown Crown density Neighbourhood 10-100'sof field data, meters air photos Retention type Retention level Fetch Soil moisture Rooting depth Stand 100'sof meters forest cover data, digital elevation Species composition Stand height Stand structure Topographic exposure Windspeed models, wind models 5 has used stand-level attributes. Forest cover and ecosystem data are readily available for most forested areas in BC and can be compiled along with topographic and wind data using Geographic Information Systems (GIS). The addition of windthrow data from aerial photograph interpretation allows the creation of large datasets for empirical modelling (e.g., Mitchell et al. 2001). While these models allow the production of coarse-scale windthrow risk maps for operational planning, forest managers also need to know the risk of windthrow associated with different retention patterns, levels of retention, and individual trees within a stand. The effects of certain factors may only be obvious at a more localized stand level (sometimes called 'neighbourhood'), or by examining individual tree attributes. The size and other characteristics of neighbourhoods that are meaningful for windthrow have yet to be defined (Coates 2001). For the purposes of this study, most tree and neighbourhood attributes were measured in the field, while stand attributes were derived from data compiled in a GIS. Some attributes can be defined at two levels. For example, height can be a tree-level measurement, and a stand-level mean. It is important to determine which attributes, at what level, are most useful for modelling windthrow risk. This includes an assessment of whether the effort required to obtain more intensive data, such as individual tree measurements, corresponds to an increase in predictive ability and improved understanding of the processes resulting in windthrow. RESEARCH OBJECTIVES This thesis reports on the incidence and risk of windthrow after partial-cutting in two study areas on Vancouver Island. This study builds on and extends work by Mitchell et al. (2001) and by Lanquaye (2003), in which cutblock edge windthrow was mapped from aerial photographs and compiled along with other GIS layers to produce stand-level empirical windthrow risk models. The objectives of this study were as follows: 6 1. to develop methods of characterizing retention level and retention pattern after retention system harvesting that are meaningful in terms of windthrow risk; 2. to test hypotheses about windthrow risk factors in variable retention stands; and 3. to build predictive empirical windthrow risk models using tree level attributes, and evaluate the benefit of adding neighbourhood and stand-level attributes. APPROACH Field and GIS-based data-sets were combined and used to model windthrow risk empirically. Ground-based sampling was used to obtain measurements of trees retained after retention system harvesting. Post-harvest canopy retention levels were mapped from low-elevation air-photos and digitized in Arc View GIS 3.2 (ESRI1996). A variety of neighbourhood variables were calculated for each sample plot. These data were combined with existing layers of site and stand characteristics in a GIS, to produce complete records for each tree. This dataset was exported for analysis with SAS statistical software (SAS Institute 2001), and logistic regression models were fit using different combinations of tree, neighbourhood and stand variables. Stand-level windthrow risk was calculated using models from other studies, and added to the tree-level models to evaluate to what degree risk of windthrow at the stand level contributed to tree-level windthrow risk. THESIS OUTLINE This thesis is divided into seven chapters. Chapter 1 has been an introduction to the rationale behind the research and the purpose of the study, and has briefly discussed the research objectives and approach. Chapter 2 is a literature review in which the current knowledge regarding key windthrow risk factors is summarized, key research questions are identified, and hypotheses are presented. Chapter 3 describes the study areas and field sampling methodologies, and defines the GIS and statistical methods and terms used. In Chapter 4, the results of the study are presented, and these are discussed in Chapter 5. Chapter 6 examines the implications of the study for forest management and presents conclusions, and in Chapter 7, recommendations for further research and development of techniques are offered. 8 CHAPTER 2 - LITERATURE REVIEW WIND BEHAVIOUR Wind is the movement of air resulting from pressure differences. Windspeed is the speed at which the wind passes a location, generally measured in metres per second, or kilometres per hour. Windspeeds vary from year to year, as well as from location to location (Quine 1994). Mean annual windspeed is the windspeed averaged over the entire year. Windspeeds near the ground are generally slower than windspeeds aloft, because of friction (Environment Canada 1992). Surface roughness also leads to increased turbulence (Bull and Reynolds 1968). The return period is the average length of time between winds of a given speed (Murphy and Jackson 1997). Return periods are calculated from long-term wind records. Peak yearly winds may be distinguished from extreme winds on this basis. In coastal BC, peak yearly winds are normally winter storm winds. Peak yearly winds on the west coast of Vancouver Island are often strong enough to cause damage in trees recently exposed by harvesting. Extreme winds occur less often, and may cause considerable damage to intact stands as well as trees that have been exposed by harvesting. Wind measurements in British Columbia are mainly made at airports or lighthouses (Murphy and Jackson 1997), and generally do not reflect forested conditions. Similar limitations in other countries have led to the development of several surrogate measures of windspeed. In Great Britain, a network of tatter flags established in fields and open moorland was used to estimate windspeeds across the country (Quine and White 1994). Physical models of windspeeds have also been developed, most of which are based on topography (Lekes and Dandul 2000, Hannah et al. 1995). BC Hydro has numerically modeled mean annual 9 windspeeds using a modified version of the Mesoscale Atmospheric Simulation System (BC Hydro 2001) and used this to produce provincial wind resource maps with a resolution of three and five kilometres. Windthrow occurs when trees are snapped or uprooted by wind. O n the coast, few recently exposed trees can withstand mean windspeeds of 100 km/hr (Stathers et al. 1994), although sustained winds at lower speeds may result in windthrow, especially i f combined with rain or snow events. Beese (2001) reported that a 25-year storm with windspeeds o f up to 32 km/hr, measured at a height of 3 m, caused significant damage in a coastal montane shelterwood and along patch-cut edges. Over-canopy windspeeds during this storm were likely much higher. A distinction has often been made between catastrophic and endemic windthrow for management purposes. Catastrophic windthrow is caused by infrequent and extreme wind events, and results in localized but extensive areas of severe damage. Endemic windthrow is less severe and affects fewer trees over a wider area; it results from routine peak winds on susceptible sites (Stathers et al. 1994). Endemic and catastrophic windthrow are best viewed as endpoints on a continuum of windthrow severity (e.g., Canham and Loucks 1984, Foster and Boose 1992, Everham and Brokaw 1996). The modelling undertaken in this thesis is most applicable to the low-severity end of the windthrow spectrum. M E C H A N I C S O F W I N D T H R O W A n y attempt to model windthrow must be based on an understanding of the mechanics of windthrow, and how various factors affect the forces involved. These forces are not complicated, but the measurements and calculations required to quantify them are. Forces that combine to break or uproot trees include horizontal drag from wind pressure on the crown and stem, and vertical self-loading from the displaced mass of the crown and stem, which increases 10 as the tree sways or leans (Peltola and Kellomaki 1993, Stathers et al. 1994). The horizontal forces are transmitted down the stem of the tree, creating a turning moment. This turning moment is a function of the lever arm length which, in the case of wind-loading, is equal to the height at which the force is applied. In the case of self-loading, the lever arm length is the distance that the centre of gravity of the stem and crown is displaced from the root pivot point. In general, the taller or more slender the tree, the greater the total turning moment for a given amount of wind force, and the more likely the tree is to be windthrown. Resisting factors include the strength and elasticity of the stem, and the strength of the root-soil interface. If the resistive forces of the stem are exceeded, the tree will break. If the resistive forces of the roots or soil are exceeded, the tree will be uprooted. The critical turning moment required to uproot and break trees is linearly related to stem mass and diameter cubed (Peltola et al. 2000, Meunier et al. 2002). EFFECTS OF WIND ON TREES Wind affects forests at both the individual tree and stand level. Trees acclimate to long-term wind exposure with developmental changes in all parts of the free. Wind breaks branches and removes foliage, resulting in 'flagging' under high wind conditions (Stokes et al. 1995). Fewer, smaller leaves may be produced (Telewski 1995). Height growth is reduced and radial growth at the base of the trunk and branches increases (Rees and Grace 1980, Telewski 1995, Mitchell 2000). Conversely, under competition for resources in closed-canopy stands, height growth takes precedence over diameter increment (e.g. Waring and Schlesinger 1985, Mitchell 2003). As a result, open-grown trees are typically more tapered, with low height-to-diameter ratios (HDR), while trees grown in dense stands are often tall and thin, with high HDR. In areas with restricted rooting, root morphology adapts to resist movement due to wind. The spread and mass of roots increases, and in some species, buttresses are formed (Nicoll and 11 Ray 1996). In conifers, large windward roots contribute the most to tree stability. Exposure to wind stimulates diameter growth in these anchoring roots (Stokes et al. 1995). Wind may also influence stand structure and dynamics. In the Coastal Western Hemlock very wet maritime subzone (CWHvm) (Green and Klinka 1994), the 'phase' level of the biogeoclimatic classification relates to forest patterns resulting from natural disturbance (Clayoquot Sound Scientific Panel 1995). The two extremes of disturbance are reflected in the Cedar-Hemlock (CH) and Hemlock-Amabilis fir (HA) phases. CH forests have an open structure and are dominated by western redcedar (Thuja plicata Donn) or yellow-cedar (Chamaecyparis nootkatensis (D. Don) Spach.), often with small crowns or spiked tops. Wind disturbance in this phase tends to affect individual trees, while the stand as a whole tends to be wmdfirm (Clayoquot Sound Scientific Panel 1995). In contrast, HA forests are dominated by more closely spaced western hemlock (Tsuga heterophylla (Raf.) Sarg.) and amabilis fir (Abies amabilis Forbes) stands, with windthrow episodically affecting patches of forest that range in size from small groups of trees to hundreds of hectares (Clayoquot Sound Scientific Panel 1995, Mitchell etai. 2004). Lertzman et al. (1996) and Pearson (2000) investigated disturbance regimes in two relatively sheltered watersheds in Clayoquot Sound, and found that wind generally affects individual trees or small groups of trees. A recent study of windthrow as a natural disturbance on south coastal Vancouver Island, found that stand replacing wind disturbance affected up to 4% of the landscape near the coast, with patch sizes ranging from 0.3 to 59 ha (Mitchell et al. 2004). In southeast Alaska, stand-replacing wmdthrow occurs mainly on exposed south-east facing slopes of the coastal islands, and affects areas of up to 1000 ha (Nowacki and Kramer 1998, Kramer et al. 2001). 12 FACTORS INFLUENCING WINDTHROW RISK Windthrow risk is affected by climate, topography, soil characteristics, stand attributes, tree attributes, and management regimes. Analysis of the individual contribution of these factors is complicated by the fact that soil conditions often reflect topography, and tree and stand properties likewise vary with soil and topographic conditions. In addition, trees and stands adapt to local wind regimes. Climate Climate consists of long-term weather patterns including temperature, precipitation and winds. Climate is affected by latitude, altitude, continentality, and by semi-permanent regional high and low pressure systems (Whiteman 2000). The province of British Columbia experiences a wide range of climatic conditions due to its varied topography. The moderating influence of the Pacific Ocean is restricted by the Coast Range Mountains. Regional circulations influencing BC include the Pacific high in summer and the Aleutian low in winter (Whiteman 2000). On Vancouver Island, gale-force and stronger winds result from storms originating over the Pacific Ocean. Wind speed varies seasonally, with the highest winds and strongest gusts occurring in the winter months (Environment Canada 2004). Prevailing winds are from the southeast in winter and the northwest in summer. On the east side of the Island, strong winds can also result from outflows of cold arctic air (Environment Canada 1992). Topography Certain topographic positions may be vulnerable to wind damage because they are more exposed to prevailing or damaging winds, or because they cause winds to accelerate or change direction. Slopes which face prevailing winds may be directly exposed, while turbulence and lee flow can affect otherwise sheltered aspects under certain conditions (Gratkowski 1956, 13 Stathers et al. 1994). Vulnerable topographic locations that have been identified in the literature include ridge tops, saddles, slopes, and occasionally valley bottoms (Alexander 1964, Hutte 1968). Narrow valleys can funnel winds, increasing windspeeds. An index of topographic exposure (TOPEX) correlates well with tatter flag studies and numerically modeled wind speeds (Quine and White 1998), and has been predictive of windthrow in several studies (Ruel et al. 2000, Mitchell et al. 2001). Soils The influence of soil properties on windfirmness is complex. Soil characteristics directly influence the strength of the root-soil interface. Rooting depth is generally considered to be the most important factor, although soil texture, moisture content, and nutrient availability, or site index, all affect wmdfirrnness (Stathers et al. 1994). Studies in Colorado, the Queen Charlotte Islands, and on northern Vancouver Island have found that windthrow increases as rooting depth decreases (Alexander 1964, Rollerson 1981, Holmes 1985). In an exception to this trend, Rollerson and McGourlick (2001) found that windthrow increased as rooting depth increased in on northern Vancouver Island and the Queen Charlotte Islands. They hypothesized that this was due to greater stand height on the deeper soils. The relationship between soil moisture and windthrow is variable. Alexander (1964), and Lohmander and Helles (1987) found more windthrow on poorly-drained soils than dry soils. In contrast, Mitchell et al. (2001) found that trees on moist soils were more susceptible than those on wet or dry soils, while Rollerson and McGourlick (2001) found more windthrow on well-drained soils than on poorly-drained soils. High water tables often affect rooting depth in coastal forests. In areas of low topography and poor drainage, soils may be saturated for much of the year. In areas where windthrow is common, pit-and-mound topography may 14 develop, leading to highly variable rooting depth over small horizontal distances. Rooting may also be affected by coarse woody debris, which occurs at high volumes in most coastal forests. Trees growing on coarse woody debris are thought to be less stable than trees growing in mineral soil (Stathers et al. 1994). Stands on richer sites are often more susceptible to windthrow (Elling and Verry 1978, Fleming and Crossfield 1983, Harris 1989, Mitchell et al. 2001). The increase in windthrow with site index may be confounded by stand height, which is generally greater on richer sites. Nutrient availability may vary widely at wet sites, affecting stand density and confounding the effects of moisture. It is difficult to separate the effects of soil from the effects of stand attributes, such as stand height and density, which are influenced by soil characteristics. We do not yet know whether nutrient rich wet soils result in more wmdthrow than nutrient poor wet soils, whether drainage is more or less important than site richness, or how much the differences in density or height due to nutrient or moisture limitations are contributing to windthrow risk. Stand attributes A stand is "a spatially continuous group of trees and associated vegetation having similar structure and growing under similar soil and climatic conditions" (Oliver and Larson 1990, p.l). Windthrow research has been undertaken in many stand types, although even-aged plantations dominate the literature. Stand age, height, density, and species composition have been identified as important stand factors in several studies. Comparisons of windthrow in different stand types should be accompanied by descriptions of stand structure, soils, and topographic exposure. 15 Stand age, height and density Studies in even-aged stands have found that windthrow increases with increasing stand age (Lohmander and Helles 1987, Foster 1988, Jalkanen and Mattila 2000). This trend may be related to the greater height of older plantations, as well as the increased incidence of disease in older stands. Most stands in unmanaged landscapes are both much older and more complex in structure than plantation forests. Such stands may contain more trees that have survived previous storms and are adapted to high winds (Everham and Brokaw 1996). On Vancouver Island, Mitchell et al. (2001) found more windthrow in immature than mature stands. The younger stands were dense and even-aged, while the older stands were multi-storied and multi-aged. Windthrow tends to increase with stand height (Rollerson 1981, Fleming and Crossfield 1983, Lohmander and Helles 1987, Foster 1988, Mitchell et al. 2001). Stand height is positively correlated with site quality as well as with age, making it difficult to determine which of these variables is most important Studies relating windthrow to stand density have had variable results. Windthrow along clear-cut edges has been found to decrease as density increases (Fleming and Crossfield 1983, Holmes 1985), but the opposite has also been reported (Mitchell et al. 2001). This disparity may be due to differences in how density and damage severity are determined. In dense stands, inter-tree damping and mutual shelter may increase wmdfirrnness as long as the edge trees remain standing (Smith et al. 1987, Harris 1989). Species composition Some combinations of species may only be found on specific sites or at specific times during stand development making it difficult to separate the effects of species composition from the effects of stand structure or site. Coniferous stands have been found to be more 16 susceptible than hardwood stands (Foster 1988, Jalkanen and Mattila 2000). Western hemlock dominated stands have proven to be less windfirm than western redcedar dominated stands in several coastal studies (Moore 1977, Mitchell etal. 2001, Rollerson and McGourlick 2001). However, the two stand types are found on different sites and have distinct disturbance histories and stand structures, as described previously. Tree attributes Individual tree attributes that have been related to windthrow risk include measures of tree size, height-diameter ratio, crown size or crown density, and species. Much of the information that has been gathered at the individual tree level on the mechanics of windthrow comes from winching studies of plantation grown trees. In general, healthy trees with good form are used. Therefore, the results may not be directly applicable to trees of varied health, size, or form, nor to species that exist in coastal BC stands. In field studies, individual tree data from plots in different stands are often pooled, which confuses within- and between-stand effects. Insights can be gained from carefully designed studies that have specifically examined the risk of windthrow for various sizes and species of trees within stands damaged by hurricanes or intense thunderstorms (e.g., Canham et al. 2001, Peterson 2004). Tree height and diameter Several studies have found that windthrow is greater in taller trees (Rollerson 1981, Holmes 1985, Veblen et al. 2001) and trees of larger diameter (Rollerson 1981, Holmes 1985, Canham et al. 2001, Peterson 2004). Everham and Brokaw (1996) have proposed a uni-modal relationship between stem size and resistance to catastrophic winds, in which windthrow is most severe in trees of intermediate size; the smallest trees are sheltered from the wind, while the largest trees are adapted to it. Absolute stem size may thus be less important than the size of the tree relative to its neighbours. 17 Crown class may be a useful variable. Jull and Sagar (2001) found that dominant or strongly co-dominant BC interior Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) trees were more windfirm than other crown classes following partial-cutting. In contrast, Beese (2001) found that a greater percentage of dominants and co-dominants were windthrown in montane coastal BC partial-cuts. Rowan et al. (2001) found no relationship between within-stand height and windthrow. Foster (1988) found that dominant trees were more vulnerable in young stands, while in older stands, trees from all crown classes were windthrown. Dominant and co-dominant trees, by virtue of their position in the canopy, are exposed to over-canopy winds, which may make them more vulnerable to windthrow, but this exposure should also result in more long-term acclimation to wind forces. Height-diameter ratio The height-to-diameter ratio (HDR) of trees has been used as an index of susceptibility to both snapping and uprooting (Wang et al. 1998). Trees with lower HDR are generally less susceptible to windthrow (Alexander 1964, Cremer et al. 1982, Coates 1997, Jull and Sagar 2001). In one instance where windthrown trees had lower mean HDR than standing trees, most of the windthrow occurred on subxeric sites with shallow soils (Huggard et al 1999). In other studies, no relationship was found (e.g., Dunham and Cameron 2000). Valinger and Fridman (1997) found that HDR and an upper diameter measure could be used to predict the risk of damage from wind or snow in Swedish forests. HDR was inversely correlated with the maximum bending moment required to uproot and break trees in Finnish free-pulling experiments (Peltola et al. 2000). Crown size and density Windthrow is associated with crown size and density, but results are variable. Crown size is generally measured as the ratio of crown length to total tree height (percent live crown), 18 or as a cross-sectional sail area calculated from both length and width measurements. Valinger and Fridman (1999) found that windthrow risk decreased with increasing percent live crown (%LC), while Rollerson (1981) found no relationship between %LC and windthrow. The stand conditions under which the crown developed may be important. Open grown trees tend to have crowns extending further along the stem (high %LC), while overstory trees grown in dense stands have high, small crowns because of competition for light. Shade tolerant understory trees may also develop high percent live crown. If crown size is defined as sail area, then crown density can be thought of as a measure of the permeability of the sail (the crown) to wind. Crown density will depend on both the amount and compactness of the foliage in the crown. For a given crown size and species, drag will increase as crown density increases (Peltola and Kellomaki 1993, Rudnicki et al. 2004). Crown density differs between species; western hemlock has the densest canopy of any coastal tree species (Pojar and McKinnon 1994). Crown density has seldom been measured in windthrow field studies. Species Other individual tree attributes are often associated with species. Several studies in temperate forests have shown deciduous species to be more wmdfirm than conifers; this is often attributed to deciduous species being leafless in winter, when storm winds are more likely (e.g. Veblen et al. 2001). Species susceptibility may differ between stands. Peterson (2004) found that some species were more vulnerable than others within southern-boreal stands in Minnesota, but patterns of vulnerability were not consistent between stands. Several studies have reported similar species susceptibility in coastal stands. In 40-60 year old riparian strips in the Pacific Northwest, Grizzel and Wolff (1998) found that Big-leaf maple (Acer macrophyllum Pursh.) was the most wmdfirm species, followed by western 19 redcedar, Douglas-fir and red alder (Alnus rubra Bong.); amabilis fir and western hemlock were the least wmdfirm species. Along the edges of clear-cut coastal Douglas-fir stands with an understory of western redcedar, western hemlock and amabilis fir, Gratkowski (1956) found that western redcedar was the most windfirm species, followed by Douglas-fir, western hemlock and amabilis fir. Similarly, Beese (2001) found that western redcedar was more windfirm than western hemlock or amabilis fir on well-drained soils. In coastal stands, large, old western redcedar often have low height-diameter ratios and spike tops, or small, sparse crowns. They also tend to be less affected by decay than other species (Minore 1990). However, western redcedar may prove to be more windfirm than other species under only certain stand conditions, rather than across all stand types. Harvesting Forest management has resulted in changes to the pattern and amount of windthrow occurring on the landscape. Proximity to harvesting seems to increase windthrow (Holmes 1985, Harris 1989, Ruel 1995, Sinton etal. 2000), although few coastal studies compare windthrow in harvested and un-harvested stands. The apparent increase in windthrow along clearcut edges is attributed to increased windspeeds and turbulence in harvested openings (Stathers et al. 1994, Lee 2000), and the increased wind load on edge trees. The rate at which trees are windthrown tends to decrease with time after harvesting (Fleming and Crossfield 1983), with exceptions occurring in years with particularly strong winds (e.g., Beese 2001). Windthrow reported along the edges of clear-cuts in coastal BC ranges from 27-38% of the stems within 25-30m of the edge (Table 2). Differences in the area over which windthrow is calculated, as well as the minimum tree size used, make comparisons between studies difficult. 20 Opening size and fetch Opening size, shape, and orientation have been found to affect both windspeed and turbulence. Studies of the effect of opening size on wind loading have had variable results. Stacey et al. (1994) found that wind loading on windward edge trees increased rapidly with the size of the gap created. However, Flesch and Wilson (1998) concluded that a quiet zone of reduced windspeed and turbulence exists near leeward forest edges, and extends approximately three tree heights from the edge. Table 2. Mean windthrow values reported along clear-cut edges and in different retention types by various windthrow studies in coastal BC. Windthrow is reported as a percentage of stems per hectare, unless otherwise indicated. % windthrow Location Reference Clear-cut edges 27a Tsitika Watershed Holmes 1985 31b N. Vancouver Island Rowan etai. 2001 38a Queen Charlotte Islands Rollerson 1981 Variable Retention cutblock edges 7 a Clayoquot Sound Scott and Beasley 2001 6 a West Island Timberlands Rollerson et al. 2002 Dispersed Retention 7, 34° Clayoquot Sound Scott and Beasley 2001 10,29d Campbell River Beese 2001 16 West Island Timberlands Rollerson et al. 2002 Group Retention (<0.5ha in area) 18e West Island Timberlands Rollerson et al. 2002 25e Clayoquot Sound Scott and Beasley 2001 Uncut Natural Standsf minimal Campbell River Beese 2001 3 Clayoquot Sound Beasley et al. 2002 16 Tsitika Watershed Holmes 1985 a - windthrow as a % of steins within 25m of the edge; b - windthrow as a % of stems within 30m of the edge; c - windthrow as a % of retained stems, retention level of 70% and 40% of initial stand density; d - windthrow as a % of retained stems, retention levels of200 and 25 stems per hectare; e - windthrow as a % of stems within retained group; f- windthrow calculated as a % of total stems within plots located in unharvested areas. 21 In areas with strong directional winds, more windthrow often occurs on windward-facing clear-cut edges (Alexander 1964, Rollerson 1981, Holmes 1985). Windthrow severity also tends to increase with fetch (Rollerson et al. 2002). Fetch is a measure of exposure related to opening size, and is generally defined as the uninterrupted distance the wind travels across an opening before hitting an edge. In many cases, fetch distance is measured only in the dominant wind direction. Rollerson et al. (2002) determined an average fetch distance based on the primary and secondary wind directions for each block, and categorized the fetch surface by retention type. Exposure variables that do not require any assumption of prevailing wind direction have been developed by Mitchell et al. (2001) and Lanquaye (2003). These variables sum the distance across the opening for each of the eight cardinal directions, or sum the number of directions in which a minimum fetch distance exists. The amount of protection afforded by adjacent stands has also been related to windthrow risk. Lohmander and Helles (1987) found that windthrow risk decreased as the windward stand increased in height and volume, and as the distance the windward stand extended increased. Fetch and shelter variables have yet to be developed for complex variable retention stands. Retention pattern and level A variety of retention patterns and levels have been employed in partial-cuts. Although windthrow varies with the type of retention employed (Table 2), trends are not consistent. In the interior of BC, areas harvested using small patch-cuts and group removal were less prone to windthrow than areas harvested with a dispersed retention pattern (Jull and Sagar 2001). When aggregate retention is employed, windthrow tends to decrease as group size increases (Burton 2001, Rollerson et al. 2002). Recent experimental trials found that partial-cutting with 22 dispersed harvest patterns and small opening sizes did not significantly increase windthrow levels (Coates 1997, Huggard et al. 1999). Windthrow risk seems to increase as retention levels decrease. In high-altitude interior BC forests, Jull and Sagar (2001) found that retaining a greater proportion of stand volume reduced windthrow. In coastal BC stands, higher levels of dispersed retention seem to result in less windthrow (Beese 2001, Scott and Beasley 2001), but small sample sizes require that these results be interpreted with caution. Results from thinning studies may provide some insight. Several studies have examined the effects of uniform thinning on windthrow risk (Cremer et al. 1982, Lohmander and Helles 1987, MacCurragh 1991, Neilsen 1995). Thinning tends to increase windthrow, with greater damage occvarring at higher levels of removal (Lohmander and Helles 1987). However, the timing of harvesting is important. There is a consensus that less windthrow occurs within naturally open stands and planted stands that have been thinned to a low density early in the rotation (Alexander 1964, Cremer et al. 1982, Harris 1989). This allows the young trees to develop under more exposed wind conditions, resulting in a future stand of windfirm trees. Windthrow following minning decreases over time, as the remaining trees acclimate to greater wind exposure. Wind tunnel measurements of the forces acting on thinned model forests demonstrate that shelter increases as tree density increases, and that final stand density is a better indicator of risk than the pattern of thinning (Gardiner et al. 1997, Novak et al. 2000). In the context of retention systems, the initial stand density, the change in stand density as a result of harvesting, and the time since harvest may be more descriptive than the post-harvest density of the stand by itself. 23 SUMMARY, RESEARCH QUESTIONS AND HYPOTHESES The complex stands that result from the use of retention systems offer new challenges for windthrow management, since much of our knowledge of windthrow is based on studies done in even-aged conifer plantations and along clearcut edges. Other researchers have stressed the importance of assessing risk factors and developing management strategies separately for different forest types (Jull and Sagar 2001), and noted that the increased use of partial cutting requires us to examine the factors that make individual trees within stands more wmdfirm (Ruel 2000). To date, few researchers have examined the windthrow risk of individual trees within coastal partial cuts, and quantitative models of tree-level windthrow risk in partial-cuts have not been developed anywhere. Therefore, data are needed that examine individual tree characteristics and identify the design factors, stand attributes and tree characteristics that are most important for windthrow. This study addressed the following questions: 1. Which patterns and levels of variable retention are most wmdfirm? Does this change for different stand types? 2. Which tree, neighbourhood and stand attributes can be used to predict windthrow risk after partial-cutting? 3. Which attribute level (tree, neighbourhood or stand) will dominate in predictive models? 4. Does higher stand-level risk for clearcut edge windthrow correspond to higher tree-level risk in variable retention? Several hypotheses were formulated based on the results of previous research. It is expected that damage will: 1. be greater in uniform than multi-storied stands, 24 2. increase with topographic exposure, 3. increase with stand height, 4. increase as time since harvest increases, 5. increase as the level of retention decreases, 6. be greater in dispersed than in aggregated retention, 7. increase as opening size (fetch) increases, 8. be more likely in slender trees with high height-diameter ratios, and 9. be more likely in trees with dense crowns. It is also expected that: 10. western redcedar will generally be more windfirm than other species. 25 CHAPTER 3 - MATERIALS AND METHODS DESCRIPTION OF STUDY AREAS Data were obtained for two separate study areas on Vancouver Island in BC, Clayoquot Sound and Snowden Forest. The Clayoquot study area is defined by Tree Farm License 54 and Forest License A19235, which constitute a large portion of International Forest Products Limited's West Coast Division. These two licenses cover much of Clayoquot Sound as well as some of the surrounding area to the north, for a total land area of approximately 65,000ha. The nearest communities are Tofino and Ucluelet (Figure 2). The STEMS project is a large-scale multi-disciplinary experiment led by the BC Ministry of Forests, and compares the effects of seven different silvicultural systems (de Montigny 2004). The first replicate in the study was established in the Snowden Demonstration forest, just north and west of Campbell River (Figure 2). Clayoquot The topography in the Clayoquot study area is variable and spans two physiographic regions. The flat, undulating Estevan Coastal Plain gives way to the steep ridges and valleys of the Vancouver Island Mountains (Clayoquot Sound Scientific Panel 1995). The forests of Clayoquot Sound fall mainly within the Coastal Western Hemlock biogeoclimatic zone, in the very wet hypermaritime and very wet maritime subzones (Green and Klinka 1994). Common tree species include western hemlock, western redcedar, and amabilis fir. Douglas-fir, Sitka spruce (Picea sitchensis (Bong.) Carr), yellow-cedar, and red alder also occur. 26 Figure 2. Map of Vancouver Island showing location of two study areas; a) detailed map of Snowden Forest and STEMS block locations; b) detailed map of Clayoquot block locations. 27 The climate in Clayoquot Sound is mild. Summer and winter temperatures are moderated by the area's proximity to the ocean. Most precipitation falls from October to March (Figure 3a). Precipitation in Tofino averages 3200mm annually but varies widely with topography and can be much higher (Environment Canada 2004). The highest winds and a) Tofino A El Precipitation -Mean'daily temperature b) Campbell River A E E c o 1 a. 'i C L 500 400 300 ^ 200 100 0 SE Precipitation —•— Mean daily temperature Figure 3. Total monthly precipitation and mean daily temperatures from the a) Tofino Airport weather station, and b) Campbell River Airport weather station. Data are from Environment Canada (2004). 28 strongest gusts occur in the winter months, and generally come from the southeast and west. The maximum hourly wind speed recorded in each month at Tofino over the past 20 years ranges from 50-1 OOkm/hr (Figure 4).; STEMS The STEMS study area falls within the Coastal Western Hemlock very-dry maritime biogeoclimatic subzone (CWHxm) (Green and Klinka 1994). Most of the stands in the area are 40-80 year old second growth resulting from either fire or harvesting. The stands are dominated by Douglas-fir, with lesser components of western redcedar and western hemlock. The climate is drier than on the west coast of Vancouver Island; Campbell River receives only 1400mm of precipitation annually (Figure 3b). The maximum hourly wind speed recorded in each month at Campbell River over the past 20 years ranges from 46-65km/hr (Figure 4). 120 100 80 60 40 20 0 la Tofino • Campbell River Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Figure 4. Maximum hourly wind speed (km/hr) recorded in each month during the time period 1971-2000 for Tofino and Campbell River weather stations. 29 FIELD DATA COLLECTION Clayoquot Retention systems have been used in Clayoquot Sound since 1996. All accessible variable retention cutblocks harvested between 1996 and the winter of2002 were sampled in the summer of2003. Cutblocks ranged in size from 1.3 to 92.0 hectares, while prescribed retention levels varied from 5-90%. A systematic sampling design was used, with plots established on a 200m grid within each block. Additional plots were located within mapped retention aggregates. Three sizes of circular fixed-area plots (0.01,0.02, and 0.04ha) were used. Before sampling, stand density (sph) from the silviculture prescription was multiplied by the prescribed retention level for each block to estimate post-harvest density. Where estimated post-harvest density was less than or equal to 100 sph, a 0.04 ha plot was used. For estimated densities of 101-400 sph, a 0.02 ha plot was used, and for estimated densities greater than 400 sph, a 0.01 ha plot was used. Where stand density was missing from the silviculture prescription, a mid-size (0.02 ha) plot was established. A total of 240 plots were sampled in 55 cutblocks. Individual tree measurements were collected for all standing and windthrown trees within the plots that were greater than 15 cm in diameter at breast height. Measurements included height (m), diameter outside bark at breast height (dbh; 1.3 m above ground; cm), diameter at stump height (0.3 m above ground; cm), and height to live crown (m). Height to live crown was measured to the crown base. Species and crown class (based on canopy position) were noted for each tree. The preceding measurements were taken according to Resources Inventory Committee standards (Ministry of Sustainable Resources Management 2002). The substrate on which each tree was growing was also noted (SUBST; organic, mineral, or wood), and crown density (CRNDEN) was classed as sparse, moderate, or full. 30 Diameter at stump height and species were noted for all cut trees within each plot, and used to reconstruct pre-harvest density (sph) and basal area (m2/ha). Tree status (alive or dead) was noted for all standing and windthrown trees, and was estimated for cut trees based on stump condition. Windthrown trees were classified as uprooted, broken, or leaning, and the direction of fall (degrees) was recorded. Site series and permeable soil depth were collected at each plot, and GPS coordinates were recorded for each plot using a Garmin Etrex GPS. Stand structure in each plot was classified as either uniform (single-storied or two-storied stand) or multi-storied (closed forest with all crown classes well represented). STEMS Field data from the STEMS study was provided by the BC Ministry of Forests. Windthrown trees were sampled in March 2003, using concentric circular fixed area plots (0.01, 0.05 and 0.10 ha). Plots were established in each block using a systematic line-plot approach. The distance between lines was constant within each block, and ranged from 60-70m, in order to achieve a pre-determined number of plots in each block. Plots were spaced 65m apart along lines. A total of 142 plots were established in the seven treatment blocks. Treatments included an uncut control, commercial minning, dispersed retention, aggregate retention, modified patchcuts, clearcut with reserves, and group selection. For this study only measurements from trees greater than 15 cm in diameter were used, and trees from the control treatment were excluded, as no windthrow occurred in that treatment. This resulted in a sample of 1891 trees in 115 plots. Measurements included classical topex determined for the plot in the field, height (m), height to live crown (m), dbh (cm) and crown class. Windthrown trees were classified as uprooted, leaning or broken. In addition, direction of fall (degrees) and average rooting depth (cm) were recorded for windthrown trees. De Montigny (2004) describes the sampling at the STEMS study site in greater detail. 31 GIS AND REMOTE SENSING INFORMATION SOURCES AND DATASET CREATION Information sources and data conversion GIS coverages for the Clayoquot study area were obtained from both the Vancouver Interfor office and the Ucluelet field office. Shapefiles were obtained showing stream locations, cutblock boundaries, the coastline, and road locations. Forest cover and logging history layers were obtained as either Arclnfo interchange coverages (.eOO extension) or as Arc View shapefiles. Cruise compilation reports were also obtained for each of the sampled cutblocks at Clayoquot. GIS coverages for the STEMS study area were obtained from the Ministry of Forests. Simulated mean annual windspeed data for all of BC were obtained from BC Hydro. Terrain Resource Information Management (TRIM) elevation maps were obtained from the Forest Information Resources Management Systems laboratory at the Faculty of Forestry, University of British Columbia. A Digital Elevation Model (DEM) was created from TRIM elevation points at 100 x 100 m resolution^ and was used to derive elevation (m)l, slope (°), aspect (°), and topographic exposure (m) for each plot. Stand-level data including stand height (m), stand volume (m ), and lead species in the stand (by volume) were obtained from forest cover layers at the STEMS study site, and from cruise compilations at the Clayoquot study site, as forest cover information was incomplete. Aerial photos and image warping Colour digital low-elevation aerial photos and IKONOS satellite imagery were obtained for several blocks in the Clayoquot study area from Interfor. Clayoquot blocks for which no photos were available were flown on October 14th, 2003 by Bazett Aerial Imaging. The new images were taken at heights ranging from 3000-4700m, and the resulting colour photos were scanned at high resolution and converted to digital images (.TIF extension). Pre- and post-harvest digital images of the STEMS site were obtained from the Ministry of Forests Research 32 Branch in Victoria. More details on the aerial photos used in this study can be found in Appendix VI. All digital images were imported into Arc View GIS 3.2. If not already georeferenced, the Image Warp extension was used to georeference the photos by 'rubber-sheeting'; a less expensive and time-intensive method than true orthorectification (Rullie and Shah 1998). Georeferencing ensures that mapped features are correctly aligned with base maps. A minimum of six ground control points (road intersections, creek crossings or obvious cutblock features) were chosen for each image. Images were rectified to a root mean square error (RMS) of 10m or less. Where RMS error was too high due to a lack of identifiable ground control points, photo-to-photo rectification was used (1996 orthophotos). Scattered windthrown trees were not easily discernable on even the lowest elevation aerial photos. CREATION OF DERIVED VARIABLES Retention type and level The level and type of retention were expected to be important variables in the windthrow risk models. In this study, three different methods were used to quantify retention level (Table 3). The percentage of stems and the percentage of basal area retained were Table 3. Variables used to describe retention level. Retention Variables Description STEMSRET % of stems retained within each plot, calculated as (standing + windthrown stems) / (standing + windthrown + cut stems) x 100 BARET % of basal area retained within each plot, calculated from diameter values of standing trees and stumps CANOPYRET % of canopy cover retained, mapped from low-elevation air photos in ArcView 33 calculated for each plot at Clayoquot, and for each cutblock at STEMS, based on the field data. The percentage of stems and basal area retained at STEMS could not be calculated at the plot level as the number of harvested trees in each plot was not recorded. For areas surrounding the plot locations in Clayoquot and STEMS, retention levels based on canopy cover were mapped in Arcview by estimating post-harvest canopy closure from the georeferenced low-elevation aerial photos (Figure 5). Pixel resolutions of less than lm for the Clayoquot photos, and less than 2m for the STEMS photos allowed very small polygons to be mapped (mmimum size=40m2). The percentage of canopy retained in each polygon was estimated to the nearest 10%. This method may underestimate retention in areas where scattered windthrow has occurred, but was moderately correlated with the other plot-level retention measures from within the same retention polygons (e.g., for STEMSRET, r=0.51, p< 0.001, n=234). In other field studies of windthrow in variable retention stands, separate sampling protocols have been developed for dispersed and aggregated retention (e.g., Rollerson et al. 2002). In Clayoquot Sound, classifying retention into these two types is difficult. Interfor implemented a complex version of retention, retaining more trees as the distance from yarding corridors increased. Many internal retention features were not mapped prior to harvest, and the choice of which trees to remove was often left to the faller. Mapped retention aggregates were often surrounded by dispersed retention, and were generally quite small, ranging from approximately 0.05-1.0ha. 34 b) Figure 5. Example of retention polygon mapping from georeferenced photos in Arc View; a) external block boundary only, b) mapped internal retention polygons. Shading of polygons indicates retention level. 35 Compared to the Clayoquot blocks, STEMS blocks had internally consistent structure. The dispersed retention treatment retained 45 evenly distributed stems per hectare. Patch-cuts ranged from 1.4 - 1.8ha in size, and group selection cuts were less than 0.5ha. For this study, retention polygons were grouped into five categories, based on the type of retention within the polygon and the type of retention adjacent to it (Table 4). Table 4. Retention type classification, and number and mean size of retention polygons. Clayoquot STEMS Retention Description Number Mean Number Mean type of size of size polygons (ha) polygons (ha) Harvested Includes roads, yarding corridors and areas with scattered single retained trees 123 2.66 17 1.78 Disppatch Patches of dispersed retention that are surrounded by harvesting 155 0.21 0 -Dispbdry Areas of dispersed retention that are adjacent3 to a block boundary or other large reserve area 180 1.32 3 14.33 Aggpatch Patches of 100% canopy cover with defined boundaries that are surrounded by harvesting or low levels of dispersed retention (<10%) 135 0.45 12 0.27 Aggbdry Areas of 100% canopy cover that are adjacent to a block boundary or other large reserve area 188 1.29 3 21.65 A t least 5 % o f total perimeter is immediately adjacent to another retention polygon. Fetch Fetch is a measure of exposure, and is often defined as the uninterrupted distance the wind travels across an opening before hitting an edge. Automated fetch-calculating programs developed in earlier research for cutblock edges and variable retention patches (e.g., Mitchell et al. 2001) could not be easily modified to run in the extremely complex retention polygons that were mapped in this study. In addition, uninterrupted openings of any great length were rare. 36 Fetch variables were therefore calculated from points located every 30m from each plot in eight cardinal directions to a maximum distance of 300m (Figure 6). Percent canopy retained was extracted for each of these points from the mapped retention polygons in Arcview. Conventional fetch distance was equal to the sum of the distance of continuous open area (< 5% retention) surrounding each plot in eight cardinal directions (FETCH). In addition, two new indices of fetch were developed, in which the fetch distance included all non-continuous open areas (SIMPLEFETCH), or was adjusted by a multiplier based on the level of removal (VRFETCH; Table 5). A detailed example of the fetch calculations can be found in Appendix vn. A distance of 300m was used for calculating fetch because it is equivalent to approximately ten tree heights, a distance beyond which little change in the windspeed profile would be expected in a conventional opening. The same distance was used to calculate the two other fetch variables so that the utility of the three variables could be compared. DIREX is another exposure variable that is easily derived in Arc View, and that has been predictive of windthrow risk in other studies (Lanquaye 2003). DIREX can be thought of as a simplified fetch measurement, which sums the number of cardinal directions in which there is a given length of opening. For this study, DIREX was calculated at three minimum distances: 30, 60 and 90m, equivalent to one, two and three tree lengths (Table 6). These distances were chosen because fetch studies by Novak et al. (2000) and Gardiner et al. (1997) indicate that wind loading on windward facing edge trees declines rapidly as gap width drops below three tree lengths. 37 Figure 6. Location of points used to determine retention levels for the purpose of calculating fetch. For clarity, the points associated with only one plot are shown. Lines delineate retention polygons, while shading indicates different retention levels. Point labels indicate the level of canopy retention associated with each point. Table 5. Description of fetch variables. Fetch variables Description FETCH Sum of continuous distance of <5% canopy cover surrounding each plot in eight cardinal directions, to a maximum of 300m in any direction SIMPLEFETCH Sum of distance of <5% canopy cover surrounding each plot in eight cardinal directions, to a maximum of300m in any direction VRFETCH Distance multiplied by removal level every 30m for 300m in eight cardinal directions surrounding the plot DIREX30 Number of cardinal directions with at least 30m of fetch (<5% canopy cover) DIREX60 Number of cardinal directions with at least 60m of fetch DIREX90 Number of cardinal directions with at least 90m of fetch 38 Spacing factor Gardiner et al. (1997) found that the increase in wind load (maximum bending moment) after thinning was linearly related to the spacing factor within the stand. Post-harvest density (sph) in each plot was used to determine the distance between trees, assuming square spacing. The spacing factor (S/H) in each plot was then calculated as the distance between trees (m) divided by average tree height (m) in the plot. Topographic exposure Topographic exposure (TOPEX) is calculated as the sum of the angle to the horizon in the eight cardinal directions. Classical topex includes only positive angles, while distance-limited topex includes both negative and positive angles, and is the maximum angle-to-ground within a specified distance (e.g., 1km, 2km, 3km). Highly exposed areas would have very negative angles to the horizon, which would lead to a value close to 0 in classical topex, and a highly negative value for distance-limited topex. In more sheltered locations such as valley bottoms, the angle to horizon in most directions will be positive, leading to a high positive TOPEX score (Figure 7). 1 km 1 km Figure 7. Illustration of distance-limited TOPEX for a north (N) to south (S) transect located in a valley bottom; 9S and ON are vertical angles (adapted from Ruel et al. 2002). 39 DATA ANALYSES AND MODELLING Simple correlation and contingency tables Pearson's simple correlations were used to explore the strength of linear relationships between variables. Where two variables were highly correlated (r > 0.7), only one was used in subsequent modelling, to allow model coefficients to be interpreted. Unless otherwise noted, all reported correlations are significant at an alpha of 0.05. The relationship between windthrow status and predictor variables was explored through contingency tables and Chi-square tests. Continuous variables were divided into classes for this purpose. Logistic regression Logistic regression is useful for analyzing data with binary response variables (e.g., windthrown, not windthrown). In logistic regression, a linear model is used to predict the probability of an event, which is constrained to lie between 0 and 1 by a logit transformation (Bergerud 1996). One of the advantages of using logistic regression for the purpose of modelling windmrow risk is that models can be fit for individual trees rather than for plot-level loss rates. Furthermore, windthrow is a relatively rare event and plot-level loss rates are non-normally distributed, violating the assumption for conventional regression analysis. Fifty percent of the observations from each study area were randomly assigned to one of two datasets (Datasets 1 and 2). For each study area, both datasets were used to build models, and to validate the models built using the other dataset. Finally, the datasets were recombined and a model was fit for the complete dataset with the variables common to both initial models. Models were fit in SAS using PROC LOGISTIC with stepwise selection (significance levels for entry into the model and removal from it were 0.10 and 0.15, respectively). The response variable for all modelling was windthrow status (WTSTAT) which was set to 1 if the tree was windthrown and 0 otherwise. 40 The importance of each predictor variable within the model can be evaluated using odds ratios. The odds ratio is the increase in odds of being in one response category when the value of the predictor increases by one unit, and is equal to the exponentiated coefficient. The farmer the odds ratio is from 1, the more influential the variable (Tabachnick and Fidell 2001). Model fit and comparison of models The Hosmer-Lemeshow goodness-of-fit test (Hosmer and Lemeshow 1989; H-L test) provides a measure of internal fit if used with the model-building data. It can also be used to test the robustness of the model, e.g., whether it can accurately predict outcomes for datasets not used in model-fitting (test datasets). In this test, the data are first sorted into ten groups based on their predicted probability of windthrow. The trees with the lowest predicted probabilities go into the first group, the next lowest go into the second group, etc. The actual proportion of windthrown trees in each group is then compared to the average predicted probability of windthrow for the group. A good model is expected to produce a non-significant Chi-square value, indicating no difference between the actual proportions and expected probabilities for both damaged and undamaged cases (Tabachnick and Fidell 2001). A model will be more useful if the 10 groups are distributed over a wide range of mean probabilities. There are several other ways to evaluate the predictive ability of each model. Concordance is calculated by comparing the predicted probabilities of all windthrown and non-windthrown trees. Each pair of observations is said to be concordant if the predicted probability of windthrow is higher for the tree that was, in fact, windthrown. The c-value is analogous to the coefficient of determination (R2) in a normal regression, and can be interpreted as the probability of correctly classifying a random pair of cases drawn from each outcome category (Tabachnick and Fidell 2001). The c-statistic varies from 0.5 (predictive ability no better than random chance) to 1 (the model always classified the tree with higher probability as 41 wmdthrown). Finally, classification tables can be used to determine the number correct predictions within each response class at a chosen cut-point. The cut-point is the probability level at which an event is assumed to occur (e.g., at a cut-point of 0.5 one would assume that a tree is windthrown if p=>0.5). 42 CHAPTER 4 - RESULTS CHARACTERISTICS OF WINDTHROWN TREES Of the 1215 trees sampled at Clayoquot, 200 were windthrown (16.5%). Most of the windthrown trees were uprooted (82%), while 9% were broken, and 9% were leaning. The most frequent directions of fall at Clayoquot were towards the northwest, west, and north (Figure 8a). At STEMS, 101 of 1891 trees were windthrown (5.3%). The most common type of windthrow was uprooting (62%), followed by leaning (34%) and breakage (3%). The most frequent directions of fall at STEMS were towards the west and northwest (Figure 8b). a) b) S 0 3 O 9 § = 0.2 o g % £ 0.1 m N N E E S E S S W W N W Direction of fall N N E E S E S S W W N W Direction of fall Figure 8. Proportion of windthrown trees at each study area that fell towards each of the eight cardinal directions; a) Clayoquot (n=198), and b) STEMS (n=97). RANGES AND MEANS OF PREDICTOR VARIABLES Summary statistics for key predictor variables in each study area are given in Tables 6 and 7. Models generated with these datasets will be most applicable to areas with attributes that fall within the ranges shown here. 43 Table 6. Summaries over all trees for key variables at the Clayoquot study area (n=1215). See Appendix I for variable descriptions and Appendix II for summaries over all plots. Variable Label Mean Std Dev Min Max Tree Variables Tree height (m) HT 20 9 4 56 Height to live crown (m) HTLC 10 6 1 39 Diameter at breast height (cm) DBH 41 31 15 330 Percent live crown %LC 46 17 2 91 Height-diameter ratio HDR 57 18 10 133 Neighbourhood variables Fetch (m) FETCH 197 207 0 1170 VRFetch (m) VRFETCH 765 298 105 1644 Direx30 DIREX30 3 2 0 8 Percent stems retained STEMSRET 73 23 4 100 Percent basal area retained BARET 63 31 1 100 Percent canopy retained CANOPYRET 68 35 0 100 Time since harvest ended (years) TIMEHARV 2 1 0 6 Permeable soil depth (cm) SOILDEPTH 31 8 9 50 Post-harvest density (sph) POSTDENS 458 258 25 1400 Post-harvest basal area (m2/ha) POSTBA 80 83 1 870 Spacing factor S/H 0.26 0.11 0.10 0.76 Stand variables Merchantable stand height (m) S T D H T 24 6 13 43 Stand volume (mVha) STD_VOL 571 198 230 1392 Topex 1km (°) TOPEX1K 34 35 -26 136 Mean wind speed (m/s) WIND 5 1 4 7 Elevation (m) ELEV 188 164 22 648 Slope (°) SLOPE 11 10 0 43 Aspect (°) ASPECT 197 102 3 358 Distance from coast (km) DISTCOAST 3 1 1 7 44 Table 7. Summaries over all trees for key variables at the STEMS study area (n=1891). See Appendix I for variable descriptions and Appendix II for summaries over all plots. Variable Label Mean Std Dev Min Max Tree variables Tree height (m) HT 29 5 7 45 Height to live crown (m) HTLC 17 5 0 28 Diameter at breast height (cm) DBH 36 10 15 84 Percent live crown %LC 85 16 27 197 Height-diameter ratio HDR 44 14 11 99 Neighbourhood variables Fetch (m) FETCH 92 250 0 1470 VRFetch (m) VRFETCH 529 339 90 1686 Direx30 DIREX30 1 2 0 8 Percent canopy retained CANOPYRET 90 26 0 100 Post-harvest density (sph) POSTDENS 494 184 10 1020 Post-harvest basal area (m2/ha) POSTBA 40 13 1 63 Stand variables Stand height (m) S T D H T 23 2 21 27 Site index SI 33 3 29 . 36 Topexlk (°) TOPEX1K 13 10 -6 44 Mean wind speed (m/s) . WIND 4 0 4 4 Elevation (m) ELEV 219 14 166 ... 247 Slope (°) SLOPE 6 3 1 11 Aspect (°) ASPECT 155 117 7 356 Distance to the coast (km) DISTCOAST • 7 1 6 8 CORRELATIONS BETWEEN SELECTED VARIABLES Several variables at Clayoquot were moderately to highly correlated (Table 8). At the tree level, height and diameter were positively correlated, while height-diameter ratio and diameter were negatively correlated. Height to live crown and percent live crown were moderately and negatively correlated. The correlation between height-diameter ratio and 45 percent live crown was negligible (r=-0.04, p=0.17). Several neighbourhood and stand variables were positively correlated, including TOPEXlk and elevation, TOPEXlk and slope, and elevation and slope. VRFETCH was positively correlated with the two other fetch variables, as well as with DIREX30. The proportion of basal area and the proportion of stems retained were highly positively correlated. Other measures of retention were correlated to a lesser degree. Spacing factor was negatively correlated with percent stems retained. Table 8. Correlations between selected variables at the Clayoquot study area. Variable 1 Variable 2 Correlation coefficient n HT DBH 0.63 1215 HDR DBH -0.55 1215 HTLC %LC -0.45 1215 TOPEX1K ELEV 0.64 234 TOPEX1K SLOPE 0.60 234 ELEV MEANWTND 0.61 234 ELEV SLOPE 0.83 234 VRFETCH DIREX30 0.41 234 VRFETCH SIMPLEFETCH 0.84 234 VRFETCH FETCH 0.53 234 STEMSRET POSTDENS 0.67 234 STEMSRET CANOPYRET 0.51 234 STEMSRET BARET 0.83 234 STEMSRET S/H -0.72 234 Many variables at STEMS were moderately to highly correlated (Table 9). Height and diameter were positively correlated, while height-diameter ratio was negatively correlated with both percent live crown and with diameter. Height to live crown and percent live crown were negatively correlated. Several neighbourhood and stand variables were correlated. Elevation was negatively correlated with both TOPEX1K and slope. All of the fetch variables were strongly positively correlated with each other. Percent canopy retained was strongly positively 46 correlated with post-harvest density, and both CANOPYRET and POSTDENS were negatively correlated with VRFETCH. Table 9. Correlations between selected variables at the STEMS study area. Variable 1 Variable 2 Correlation coefficient n HT DBH 0.78 1891 HDR %LC -0.43 1891 HDR DBH -0.72' 1891 HTLC %LC -0.84 1891 TOPEX1K ELEV -0.51 115 TOPEX1K SLOPE 0.43 115 ELEV SLOPE -0.66 115 VRFETCH DIREX30 0.87 115 VRFETCH SIMPLEFETCH 0.99 115 VRFETCH FETCH 0.88 115 CANOPYRET POSTDENS 0.84 115 CANOPYRET VRFETCH -0.72 115 POSTDENS VRFETCH -0.65 115 POSTDENS POSTBA 0.88 115 CONTINGENCY TABLES There was no significant difference in the proportion of trees windthrown in uniform versus multi-storied stands at Clayoquot (Figure 9a, j?dt=i=Q2l, p=0.64). There was also no difference in windthrow levels between stands dominated by western redcedar and stands dominated by other species (Figure 9b, x2dM=l-21, p=0.75). Trends between windthrow and other variables within different stand types were similar, thus the data were pooled and trends are presented in Figures 10,11 and 12. 47 a) b) Multi-storied Uniform Stand structure B A C W H W Y C Lead species (by volume) Figure 9. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area (n=1215) by: a) stand structure, b) lead species by volume. Windthrow increased as distance from the coast increased, and was highest on south facing aspects and from 200-400 m elevation (Figure 10). Windthrow was high in both the most exposed and the most sheltered topographic locations. Windthrow increased with mean annual windspeed, and was greatest in stands with merchantable heights of20-40 m. a) b) 0.3 § 0.2 o | 0 1 Wt • • -A, Hi - • f BI • i i i 1 i i 600 400 200 0 0.5 1.5 2.5 3.5 4.5 5.5 6.5 Distance from coast (km) 50 150 250 350 450 550 650 Elevation (m) 48 c) 0.3 | 0.2 •e o a. 2 0.1 Q-e) N E s Aspect w d) 600 400 -o a E 200 z f) 5 15 25 35 45 Slope (degrees) 0.3 § 0.2 •c o a. S 0.1 a. • • —-1 iff 4.25 4.75 5.25 5.75 6.25 6.75 Mean windspeed (knVhr) 600 400 X> E 200 z g) 0.2 c o o 0.1 a i 7*" • / tiki * 800 600 400 f 3 Z 200 0 15 25 35 45 Merchantable stand height (m) Figure 10. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the stand level; n=1215; a) distance from coast, b) aspect, c) elevation, d) slope, e) topographic exposure, f) mean annual windspeed, and g) stand height. A chi-square test indicates that windthrow varies significantly between classes for each of these variables (p < 0.05). 49 Several neighbourhood variables were related to windthrow (Figure 11). Windthrow varied with time since harvest, but did not increase as expected. Windthrow decreased as soil depth increased, but did not vary with soil moisture or nutrient regime. The proportion of windthrown trees increased with decreasing post-harvest density and percent stems retained. Windthrow increased with an index of fetch (VRFETCH), and with the number of directions to which each plot was exposed for at least 30m (DIREX30). Windthrow varied by retention type, and was highest in the two types of dispersed retention. Damage also increased as the spacing factor increased. a) b) <1 1 2 3 4 5 6 Time since harvest (years) 0.5 0.4 | 0.3 o £ 0.2 °- o . H 600 400 200 0 5 15 25 35 45 Permeable soil depth (cm) c) d) 0.3 g 0.2 o Q. 8 0.1 tf • • 500 1400 300 | 200 | 100 0 150 450 750 1050 1350 1650 VRFETCH 0.3 | 0.2 M O o. S 0.1 Q. 0 1 2 3 4 5 6 7 8 DIREX30 400 300 fm 200 | 3 z 100 0 50 e) f) e o € o a e a. 0.6 0.5 0.4 0.3 0.2 0.1 0 10 Percent stems retained 1— 30 SO 70 90 600 500 400 ._ a 300 £ 200 Z 100 150 450 750 1050 1350 Post-harvest density g) h) 0.3 I 0.2 •c o g 0.1 a. Retention type 600 400 % E 200 z 0 0.5 0.4 c ° 03 t: •I 02 0.1 0.1 0.3 0.5 0.7 Spacing factor 800 600 2 400 E 3 Z 200 0 Figure 11. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the neighbourhood level; n=1215; a) time since harvest, b) permeable soil depth c) VRFETCH, d) DIREX30, e) percent stems retained, f) post-harvest density (sph), g) retention type, and h) spacing factor. A chi-square test indicates that windthrow varies significantly between classes for each of these variables (p < 0.05). Trends in the proportion of trees damaged relative to class values of tree variables are shown in Figure 12. The proportion of damaged trees increased as height-diameter ratio increased, and as percent live crown decreased. There was greater damage in trees that were growing on organic substrates than in trees growing on wood or mineral substrates. Windthrow increased as crown density increased. There was no difference in windthrow by species or crown class. 51 a) 0.5 0.4 • c o 0.3 o Q. 2 0.2 Q.0.1 0.0 600 400 fc. .O 200 10 30 50 70 90 >100 Height-diameter ratio c) d) Mineral Wood Organic Substrate 10 30 50 70 90 Percent live crown 0.3 I °-2 -I O CL S 0.1 CL 0.0 • • Sparse Moderate Full Crown density 600 400 200 0 AO E 3 Figure 12. Proportion of trees windthrown (bar) and number of trees sampled in each class (0) at the Clayoquot study area for selected variables at the tree level; n=1215; a) height-diameter ratio, b) percent live crown, c) substrate, and d) crown density. A chi-square test indicates that windthrow varies significantly between classes for each of these variables (p < 0.05). At STEMS, all plots were located in the same stand type, a uniform second-growth stand dominated by Douglas-fir. There was a significant difference in windthrow by treatment, with 40% damaged trees in the dispersed treatment, while fewer than 20% of trees were damaged in all other treatments (Figure 13). Windthrow increased with slope, and was greater on north and west-facing aspects. Windthrow increased as TOPEXlk, VRFETCH and DIREX30 increased. There is a trend of increasing damage as post-harvest density (and post-harvest basal area) decreases. The proportion of windthrown trees increased with decreasing 52 diameter, and with decreasing percent live crown. The proportion of windthrow also varied with crown class, and with species. More windthrow occurred in trees of intermediate crown class, and more western red cedar were windthrown than Douglas-fir or hemlock. a) b) o a I 0.6 0.4 0.2 0 • • -• -11 Treatment 800 600 „ 400 f 3 t f & £ £ '• & P & Jr 2 3 4 5 6 DIREX30 c) d) 1 3 5 7 9 11 Slope (degrees) e) f) 5 15 25 >30 TOPEXlk 0.3 § 0.2 t o Q. S 0.1 a. • • • — B H „ • ...1 6 6.5 7 7.5 8 1200 900 a 600 f Z 300 0 DISTCOAST (km) 53 g) h) 150 250 450 750 1050 1250 1750 Post-harvest density (sph) 0.6 0.5 | 0.4 0 0.3 a. 1 0.2 0.1 0 • h • M 1 |™| — r - i Iii 1 9 150 450 750 1050 1350 1650 VRfetch i) j) 0.1 | g. 0.05 \ CL • • • M B 'msa 20 30 40 50 1200 1000 800 >_ 600 £ 3 400 Z 200 0 DBH (cm) 10 30 50 70 90 Percent live crown Figure 13. Proportion of trees wmdthrown (bar) and number of trees sampled (0) for selected variables at the STEMS study site; n=1891; a) treatment, b) DIREX, c) slope, d) aspect, e) TOPEXlk, f) distance to coast, g) post-harvest density, h) VRFETCH, i) DBH, and j) percent live crown. A chi-square test indicates that windthrow varies significantly between classes for each of these variables (p < 0.05). MODELS Logistic regression models were fit separately for Clayoquot and STEMS. Models were fit for each of the datasets from the Clayoquot study site using combinations of variables from tree, neighbourhood and stand levels (Table 10). Class variables were coded as dummy 54 Table 10: Clayoquot models using tree, neighbourhood, and stand variables (Dataset 1 n=608, Dataset 2 n=607). Critical x2 value=15.507. Model Dataset . Variables8 % Concordant c-value H-L test x2 statistic -fitting dataset H-L test x2 statistic -test dataset Tree variables only 1 CRNDEN, SUBST2, HDR, %LC 68.0 0.683 9.06 p=0.34 10.62 p=0.22 2 CRNDEN, SUBST2 , HDR, %LC 74.9 0.751 11.62 p=0.17 18.50 p=0.02 Neighbourhood variables only 1 2 VRFETCH, TIMEHARV, POSTDENS VRFETCH, DIREX30, POSTDENS, PREBA 66.5 68.1 0.669 0.685 2.63 p=0.96 4.97 p=0.76 16.79 p=0.03 23.87 p<0.01 Stand variables 1 TOPEX1K, DISTCOAST 57.8 0.587 8.50 p=0.39 9.86 p=0.27 only 2 ASPECT, WIND 60.8 0.620 13.91 p=0.18 12.90 p=0.12 Tree and neighbourhood variables 1 2 CRNDEN, SUBST2 , HDR, %LC, VRFETCH, DIREX30, TIMEHARV SPECIES, CRNDEN, SUBST2 , HDR, %LC, VRFETCH, DIREX30, POSTDENS 74.4 80.6 0.745 0.807 5.84 p=0.67 10.93 p=0.21 16.79 p=0.03 23.87 p<0.01 Neigbourhood and stand variables 1 2 VRFETCH, TIMEHARV, POSTDENS, SLOPE, DISTCOAST VRFETCH, SOILDEPTH, POSTDENS, ASPECT, WIND 70.8 70.2 0.712 0.706 5.06 p=0.75 2.71 p=0.95 14.24 p=0.07 14.02 p=0.08 Tree and stand variables 1 2 CRNDEN, SUBST2, HDR, TOPEX1K, DISTCOAST CRNDEN, SUBST2 , HDR, %LC, ASPECT 71.2 77.1 0.715 0.773 9.35p=0.31 14.06 p=0.08 9.99 p=0.27 23.93 p<0.01 Tree, neighbourhood and stand variables 1 2 CRNDEN, HDR, %LC, SUBST2, VRFETCH, POSTDENS, TIMEHARV, TOPEX1K, DISTCOAST SPECIES, CRNDEN, HDR, %LC, SUBST2, VRFETCH, POSTDENS, ASPECT, DIREX30, SOILDEPTH 77.1 81.9 0.773 0.821 10.20 p=0.25 4.62 p=0.80 7.75 p=0.46 28.31 p<0.01 'Variables described in Appendix I. 55 variables, using one of the classes as a reference level (see Appendix III). Substrate was re-coded as a binary variable (SUBST2; organic vs. other). Variables that were included but not selected in any of the models include retention type, spacing factor, elevation, and stand height. In general, similar variables were significant in the models for each of the two Clayoquot datasets. Most of the models showed no internal lack-of-fit when tested against model-building data, and but were slightly less accurate when tested against the other dataset. Of the models that included variables from a single attribute level (tree, neighbourhood, stand), the tree model had the highest concordance values (68.0 and 74.9%), followed by the neighbourhood-only model. The model with only stand variables had the lowest predictive ability. Concordance values increased as variables from other levels were added. Six variables were selected in both of the tree-neighbourhood-stand level models (HDR, %LC, SUBST2, CRNDEN, POSTDENS, and VRFETCH). These variables were therefore used to fit an overall model using the complete dataset. The same process of model-fitting using combinations of variables from tree, neighbourhood and stand levels was completed for the STEMS datasets (Table 11). All neighbourhood variables were highly correlated at the STEMS study site, therefore univariate analysis was used to choose a single neighbourhood variable to use in modelling (DIREX30). The variables crown class, aspect, and TOPEXlk were included but not selected in any models. Similar variables were significant in the models for each of the two STEMS datasets. The models that included only stand variables had the highest concordance values, but the worst internal fit. The rest of the models showed no internal lack-of-fit when tested against model-building data, and were accurate when tested against the other dataset. 56 Table 11: STEMS models using tree, neighbourhood and stand variables (Dataset 1 n=942, Dataset 2 n=949). Critical x2 value=l5.507. Model Dataset Variables8 % Concordant c-value H-L test yf statistic -fitting dataset H-L test x^  statistic -test dataset Tree variables only 1 DBH 60:5 0.620 7.73 p=0.46 11.78 p=0.16 2 SPECIES, %LC 71.7 0.725 6.06 p=0.64 15.42 p=0.05 Neighbourhood 1 DIREX30 63.6 0.772 2.90 p=0.23 15.60 p=0.05 variables only 2 DIREX30 71.4 0.820 2.79 p=0.25 23.45 p<0.01 Stand variables only 1 DISTCOAST, ELEV 75.0 0.762 32.83 pO.Ol 24.25 p<0.01 2 DISTCOAST, ELEV 76.6 0.776 35.05 p<0.01 22.84 pO.Ol Neighbourhood and tree 1 DBH, DIREX30 80.7 0.815 6.14p=0.63 4.46p=0.81 variables 2 DBH, %LC, DIREX30 84.7 0i854 6.04 p=0.64 10.11 p=0.26 Neighbourhood and 1 DIREX30, DISTCOAST 80.5 0.835 12.20 p=0.14 11.59 p=0.17 stand variables 2 DIREX30, DISTCOAST, SLOPE 87,2 0.880 15.49 p=0.05 27.70 p<0.01 Tree and stand variables 1 DBH, DISTCOAST, ELEV 76.7 0.773 6.97 p=0.54 19.73 p=0.01 2 %LC, DBH, DISTCOAST, SLOPE, ELEV 79.0 0.795 27.16 pO.Ol 8.33 p=0.40 Tree, neighbourhood 1 DBH, DIREX, DISTCOAST 83.9 0.845 5.64p=0.68 8.97p=0.34 and stand variables 2 %LC, DBH, DIREX30, DISTCOAST, 87.4 0.879 23.21 p<0.01 12.15 p=0.14 SLOPE "Variables described in Appendix I. 57 Three variables were selected in both tree-neighbourhood-stand level models (DIREX30, DBH, and DISTCOAST) and were used to fit an overall model using the complete dataset. Interactions between class variables and continuous variables were tested for the best-fit models at Clayoquot and STEMS, but were not found to be significant. Parameter estimates for the initial models at Clayoquot and STEMS can be found in Appendix IV. BEST-FIT MODELS • . . > „ ' The correlations between variables used in the models are low, allowing the coefficients to be interpreted with caution. In the best-fit model for Clayoquot (Table 12), the odds of windthrow increased with height-diameter ratio, crown density and VRfetch, and decreased with increasing percent live crown and post-harvest density. HDR was the most influential of the continuous variables, based on the odds-ratios. Having a full crown increased the odds of being windthrown 3.5 times over the odds for a tree with a sparse crown, while a tree growing on organic substrates had twice the odds of being windthrown as a tree growing other substrates (wood or mineral). The maximum predicted probability of windthrow for an individual tree was 75%, for a tree growing on organic substrates with a HDR of 115, %LC of 47%, a surrounding post-harvest density of200 sph, and VRFETCH of 1056 m. In the best-fit model for STEMS (Table 12), the odds of windthrow increased with increasing DIREX30 and DISTCOAST, and decreased with increasing tree diameter. Based on the odds ratios, DIREX30 was the most influential variable in the model. At STEMS, the maximum predicted probability of windthrow for an individual tree was 73%. This was for a tree with a DBH of 22 cm, located at a distance of eight kilometres from the coast, where DIREX30 was also eight (the maximum). A Hosmer-Lemeshow goodness of fit test for the 58 Table 12. Variables, coefficients, and odds-ratios for the best-fit logistic regression models at the Clayoquot and STEMS study site. Dependent variable is WTSTAT. Clayoquot full dataset STEMS full dataset Variable (n=1215) (n=1891) Coefficient Odds Coefficient Odds Ratio Ratio Intercept -3.5157 -6.0367 HDR 0.0300 1.306a Not included %LC -0.0245 0.740a Not included DBH Not included -0.0524 0.696a CRNDEN ( Full vs Sparse) 1.0007 2.720 Not included CRNDEN ( Moderate vs Sparse) 0.5599 1.750 Not included SUBSTRATE (Organic vs other) 0.7832 2.188 Not included VRFETCH 0.00128 1.192" Not included POSTDENS -0.00199 0.736 s Not included DIREX30 Not included 0.4617 1.447a DISTCOAST Not included 0.5622 1.119 a % concordant 75.3 85.2 c-value 0.755 0.858 Hosmer-Lemeshow test 3.27p=0.92 11.82 p=0.16 statistic % trees correctly classified b 72 94 % windthrown trees correctly 59 54 classified b a Odds-ratio estimates are for a 10% change in the given variable. b Cut-point = 0.2 best-fit models indicated no lack-of-fit in either dataset, although at STEMS, the 10 probability groups were not very well distributed (Figure 14). 59 • Clayoquot • STEMS Observed Figure 14. Observed windthrow versus predicted probability of windthrow for best-fit Clayoquot and STEMS models. Data are sorted by predicted probability of windthrow and divided into ten equal groups. EFFECT OF ADDING STAND-LEVEL WINDTHROW RISK Local models for predicting windthrow risk along clearcut edges have been developed for Weyerhaeuser's North Island Timberlands (Lanquaye 2003), and West Island Timberlands (Mitchell 2003). These models were used to calculate stand-level windthrow probabilities for the Clayoquot and STEMS areas (Figure 15; see Appendix V for details on NIT and WIT models). The mean predicted stand-level windthrow probability for the Clayoquot site was 0.34, with a minimum of 0.17 and a maximum of 0.74. At the STEMS site, mean stand-level windthrow risk was 0.12, and the range was smaller, with a minimum of 0.09 and a maximum of 0.45. Stand-level windthrow risk for each area was added as a variable to the best-fit models already developed for Clayoquot and STEMS, but was not significant in either case. 60 a) b) 0.2 0.3 0.4 0.5 0.6 0.7 Risk of windthrow at Clayoquot after clearcutting (WIT model) 0.1 0.2 0.3 0.4 Risk of windthrow at STEMS after clearcutting (NIT model) Figure 15. Proportion of trees windthrown (bar) and number of trees sampled (0) versus predicted risk of windthrow after clearcutting using stand-level models developed by Mitchell (2003) and Lanquaye (2003) at a) Clayoquot, n=1215, and b) STEMS, n=1891. 61 CHAPTER 5 - DISCUSSION Several hypotheses were formulated and tested in this study, with varying results. At the stand level, it was expected that windthrow would be greater in uniform than multi-storied stands, and that windthrow would increase with increasing topographic exposure and stand height. At the neighbourhood level, windthrow was anticipated to increase as retention levels decreased, and be greater in dispersed than aggregate retention. Windthrow was also expected to increase as the number of years since harvest increased. Finally, at the tree level, it was expected that trees with high height-diameter ratios and dense crowns would be more susceptible to windthrow, and that western redcedar would be less susceptible than other species. DIFFERENCES BETWEEN STANDS At the Clayoquot site, 16.5% of sampled trees were windthrown, while at the STEMS site only 5.3% of sampled trees were damaged by wind. This disparity is likely due to differences in wind climate, as both peak hourly winds and mean annual windspeeds were higher on the west coast. Mitchell et al. (2001) likewise found that the probability of windthrow was higher on the west coast of Vancouver Island than on the east coast. The stands on the east and west coast of the island also differed in age, pre-harvest stand density, and species composition. No significant difference in windthrow levels was found between uniform and multi-storied stands, or among stands dominated by different species at Clayoquot. Several other coastal studies have found stands dominated by western redcedar to be more wmdfirm than stands dominated by western hemlock; however, cedar stands were mainly multi-storied while 62 hemlock stands were uniform (Mitchell et al. 2001, Lanquaye 2003). More windthrow was expected in uniform stands in part because trees in these stands tend to be tall and slender. Although the trees found in uniform stands at Clayoquot were taller and more slender, on average, than trees in multi-storied stands, there was no difference in windthrow levels. This may be because the uniform stands sampled at Clayoquot were older Hemlock-Amabilis stands consisting of large diameter trees that had adapted to the local wind regime. In addition, the stands were harvested using retention systems rather than clearcutting, so that the increase in exposure due to harvesting would be less than that experienced by the uniform Hemlock-Amabilis stands on eastern Vancouver Island that were investigated in other studies (e.g., Mitchell et al. 2001). At STEMS, all plots were located in uniform stands and were dominated by Douglas-fir. Mean windspeed and topographic exposure were generally not significant variables in the models given the other variables. Mitchell et al (2001) found that stand-level windthrow was related to these variables, while another study found no consistent relationship between windthrow and windspeed (Ruel 2000). Tree-level studies have generally not included windspeed or topographic exposure as predictor variables, and it is possible that the windspeed data used in this study were too coarse to capture within-stand differences. Windthrow was highest in both the most sheltered and most exposed locations, possibly indicating that winds are being funneled along valley bottoms (Mitchell 2003, Ruel et al. 2003). Significant variables at the stand-level were mainly related to location. Windthrow increased with increasing slope and distance from the coast, and varied with elevation. More windthrow occurred on south-facing aspects at Clayoquot, which was expected given the prevailing south-easterly winds. These trends seem to indicate that geographic differences in 63 wind climate are important, but that these differences are not adequately captured by available mean windspeed estimates or TOPEX indices. Windthrow did not vary significantly with stand height in this study, although others have found a positive relationship (e.g., Lohmander and Helles 1987, Mitchell 2003). In multi-storied stands that contain trees of several heights, stand height may not be a useful indicator of the susceptibility of trees to windthrow. Stand-level risk of clearcut edge windthrow was not a significant variable when added to the overall Clayoquot or STEMS models. This implies that windthrow risk in variable retention is associated with a different set of variables than windthrow risk along clearcut edges. It is possible that more conservative retention treatments were used in high-risk stands, confounding the relationship between windthrow occurrence and stand-level windthrow risk; however, further examination of the data showed that there was no difference in mean retention levels between low-risk and high-risk stands. MODELS For both the Clayoquot and STEMS datasets, models were fit using variables from three different levels: tree, neighbourhood, and stand. Adding different levels to the models increased predictive ability, with tree variables providing the greatest increases at Clayoquot, and neighbourhood variables providing the greatest increases at STEMS. This difference may be due in part to the homogeneity of the trees within the uniform stands at STEMS. The best-fit model for Clayoquot included only tree and neighbourhood variables, while the model for STEMS included one variable from each of the three levels. Several of the variables selected by the models were also significant in other modelling studies (e.g., Valinger and Fridman 1999, Peterson 2004). Although fit was adequate for both the Clayoquot and STEMS models, the probability groups for the STEMS model were not well distributed, and included only one high-risk group, making the STEMS model less reliable for prediction. 64 The Clayoquot and STEMS models correctly classified 72 and 94% of the sample trees in the two study areas. This is comparable to what has been found in other modelling studies at both the tree and stand level. Peterson (2004) was able to correctly predict windthrow status of 68-90% of trees vrithin five stands in Minnesota using tree species and diameter. Stand-level models developed by Mitchell et al. (2001) accurately predicted the damage status of 71-76% of clearcut edge segments on northern Vancouver Island, while Kramer et al. (2001) obtained correct classification rates of 72% in a stand-level windthrow study of natural wind disturbance patches in Alaska. The models developed here were better able to classify undamaged trees than those that were windthrown. This is due, in part, to the relative rarity of windthrown trees in the datasets. Despite the many differences between the two study areas, in both cases similar tree and neighbourhood attributes were key predictors (DBH and HDR, and VRFETCH and DIREX30). This implies that windthrow risk after partial-cutting is controlled by the same underlying processes in both areas. In addition, neither stand-level variables nor stand-level windthrow risk were strongly related to tree:level windthrow risk in variable retention, which means that the extra cost and effort required to obtain individual tree and neighbourhood measurements in retention stands is warranted, if the goal is to predict windthrow risk at the tree-level. KEY TREE-LEVEL VARIABLES Since trees have been shown to adapt to increased wind exposure, individual tree characteristics should be indicative of tree-level windthrow risk. Several tree characteristics were related to windthrow, including diameter, height-diameter ratio, percent live crown, crown density, and substrate. At the STEMS study area, larger diameter trees were less likely to be windthrown. A similar, though non-significant trend occurred at the Clayoquot study area. This finding is 65 contrary to results from several other field studies, in which larger trees have been found to be more susceptible to windthrow (Rollerson 1981, Holmes 1985, Canham et al. 2001). On the other hand, wmching studies have found that the force required to uproot trees is linearly related to DBH 2 , which indicates that trees of larger diameter will be more resistant to windthrow, all other things being equal (Peltola et al. 2000). At the Clayoquot study area, wmdthrow risk increased with increasing height-diameter ratio, a trend that is generally supported in the literature (Cremer et al. 1982, M l and Sagar 2001), although Valinger and Fridman (1997) found that high height-diameter ratios in the largest sample tree indicated a low-risk, high-density site in Swedish boreal forests. While the relationship between height-diameter ratio and snapping is intuitive and straightforward, the mechanism which reduces the risk of uprooting in trees with lower HDR is not fully understood (Gardiner et al. 1997). Greater diameter growth in the lower stem may indicate that proportionally more resources have been allocated to the development of structural roots, or it may simply move the center of gravity of the tree downward, both of which would be expected to increase resistance to uprooting. Percent live crown was a significant variable in the Clayoquot model; trees with longer crowns were less likely to be windthrown. Dunham and Cameron (2000) similarly found that undamaged Sitka spruce had larger crown areas than damaged trees in a plantation study. Trees with larger crowns should be subjected to higher wind loads. However, trees with crowns extending further down the bole may also be those that have developed under low density conditions, and are therefore better adapted to the higher wind loads to which they are subjected after partial-cutting. Mean percent live crown was negatively correlated with pre-harvest density at Clayoquot (r=-0.19, p=0.003, n=234), lending some support to this theory. Trees with longer crowns are also often those with lower height-diameter ratios (Wang et al. 1998); 66 however, the two variables were not significantly correlated at Clayoquot and were only moderately correlated at STEMS. Windthrow risk at Clayoquot increased with increasing crown density, as expected. There is some uncertainty regarding what is captured by this variable, as the plots at Clayoquot were sampled at different times after harvesting. This time lag would have allowed the crowns of individual trees to adapt by shedding foliage and/or branches. It is therefore impossible to tell whether crowns classified as sparse were sparse before harvesting took place, or whether they became sparse as the tree adapted to the increased wind loads after harvesting. In the Clayoquot area many redcedar trees, in particular, have sparse crowns prior to harvest (personal observation). In addition, the relationship between crown density and windthrow was strongest in the more recently harvested plots, and less obvious in plots harvested several years previously, a trend which supports the value of pre-treatment crown density as a variable. Windthrow did not vary between species at the Clayoquot site, while at the STEMS site more western redcedar were windthrown than other species. This is contrary to findings from other coastal studies, where western redcedar has been noted as a windfirm species (Grizzel and Wolf 1998, Beese 2001). At the STEMS site, western redcedar was a minor component of the stand (<4% of trees), and most redcedar were in the intermediate crown class. KEY NEIGHBOURHOOD-LEVEL VARIABLES Several attributes describing the local stand, or neighbourhood, were related to windthrow. Retention level and post-harvest density appear to be more important to windthrow risk than the pattern or type of retention employed. In the Clayoquot study area, few trees were windthrown when more than 20% of the stems were retained, while almost 50% of trees were windthrown when retention levels were below 20%. A negative relationship also existed between windthrow and post-harvest density, although no obvious threshold values were 67 identified. N o trend was observed with pre-harvest stand density, although this has been a significant variable in other studies. Windthrow increased as the spacing factor within the stand increased. Despite being strongly related to windthrow, spacing factor was not significant in any of the Clayoquot model. However, spacing factor was highly correlated with post-harvest density, which was significant in the Clayoquot model. Windthrow was higher in the two dispersed retention types than in the aggregated types at Clayoquot. A t S T E M S , windthrow was highest in the dispersed retention treatment, which also had the lowest retention level of any treatment (~5%), meaning that the effects of retention pattern can not be separated from the effects of retention level. Other coastal windthrow studies have found more windthrow in dispersed than aggregated retention (Beese 2001, Rollerson et al. 2002). In a wind tunnel experiment, Gardiner et al. (1997) found that flow over the canopy was more strongly related to stand density than to the pattern of thinning, a finding supported by the Clayoquot model, which included post-harvest density as a variable, but not retention type. Windthrow was expected to increase with fetch distance, and several fetch variables were tested in this study. A t the Clayoquot site, windthrow was most strongly related to the two new fetch variables, S I M P L E F E T C H and V R F E T C H . A t the S T E M S site, all three fetch measurements were highly positively correlated with each other, and showed similar relationships to windthrow occurrence. A t both study areas, DIREX30 was more strongly related to windthrow than DIREX60 or DIREX90, demonstrating that even very small openings can lead to increased risk of windthrow. Given that this was the case, it may also be possible to calculate V R F E T C H over a much shorter distance than the 300m used in this study. V R F E T C H and DIREX30 were significant in the best-fit models for Clayoquot and S T E M S , respectively, 68 indicating that these variables capture some of the change in exposure of retained trees in variable retention stands. Similar exposure variables have been significant in studies of windthrow on clearcut edges (Lanquaye 2003, Mitchell 2003). Windthrow varied with year of harvest at Clayoquot, but did not increase with time since harvest, as has been found in some windthrow studies (e.g., Mitchell et al. 2001, Ruel et al. 2003). Holmes (1985) found no relationship between time since harvest and amount of windthrow on clearcut edges on north-eastern Vancouver Island, while Beese (2001) found that most windthrow occurred in a year with extreme damaging winds. It is possible that windthrow levels at Clayoquot are more dependent on local exposure to severe winter storms than the number of years since harvesting occurred; however, meteorological data to confirm this are not currently available. More windthrow occurred in plots harvested during 1998, 2002 and 2003 than in plots harvested in other years. Retention levels were similar across years, except for 1998, which had low average retention levels (34%). SOIL CHARACTERISTICS At the Clayoquot study site windthrow increased as permeable soil depth decreased, but did not vary significantly with soil moisture or nutrient regime. Substrate was the only significant soil-related variable in the Clayoquot models. Trees growing on organic substrates were more likely to be windthrown than trees growing on coarse woody debris or directly in mineral soil. Trees rooted on wood are thought to be less stable than trees rooted in mineral soil (Stathers et al. 1994), but it is possible that in the often saturated soils of the west coast, large pieces of woody debris provide stable rooting platforms. No other studies were found that assessed rooting substrate for individual trees; however, stand-level windthrow studies in coastal BC and Alaska found that more windthrow occurred on mineral soils than organic soils (Harris 1989, Mitchell etal. 2001). 69 LIMITATIONS OF METHOD The Clayoquot model presented here uses windthrow data collected in Clayoquot Sound, and is indicative of the wind climate that occurred from 1996 to 2002. A wide range of retention types and patterns were sampled in order for the models to be as useful as possible, but changes to wind climate due to factors such as global warming, or major changes in harvesting practices would invalidate these models. Hybrid models that incorporate both mechanistic and empirical modelling results may prove to be useful in the face of changing future conditions. The time-lag which occurred between harvesting and sampling is also of concern. Pre-harvest conditions were reconstructed from measurements taken after harvesting, which may have introduced some inaccuracies. The potential for post-harvest changes in crown density were noted above. Every attempt was made to identify and discount trees which were windthrown before harvesting occurred, but it is possible that some previously damaged trees were included in the sample. Also, aerial photos of individual blocks were taken in the same year but at different times after harvest, making it difficult to map canopy retention levels consistently. To improve sampling efficiency, it would have been preferable to obtain aerial photos and map retention polygons prior to sampling. This was done after sampling in this study, because overall retention patterns were not as easy to characterize during ground sampling as had been expected and silviculture prescription maps did not map internal retention very precisely. 70 CHAPTER 6 - CONCLUSIONS Adequately fitting tree-level windthrow risk models can be developed to predict the risk of windthrow following partial-cutting in coastal stands. Similar tree and neighbourhood variables were selected by the models for two study areas with different stand types and overall levels of windthrow, implying that the processes that result in windthrow after partial-cut harvesting are consistent. This suggests that the models may also work for other areas; however, the portability of the models has yet to be tested and it would be necessary to calibrate models for differences in regional wind regime. Geographic factors influence windthrow risk in variable retention stands, but the coarse windspeed estimates and TOPEX indices that are currently available do not provide good tree-level prediction. The development of finer-scale wind flow models and flow models for within retention canopies would likely help. While stand-level windthrow risk models are useful for identifying locations in the landscape where stand edges are more susceptible when exposed by clearcutting, tree and neighbourhood models are needed to evaluate the effects of harvest design and tree properties on risk to residual trees. The effects of different patterns and levels of retained trees on within-stand windspeeds and turbulence have yet to be quantified, but appear to be approximated by the fetch indices developed here. The extra cost and effort required to obtain the detailed measurements required to build these models would appear to be justified if predictions of tree-level windthrow risk are the aim. Retention levels of at least 20% of the original stand density (in sph) resulted in lower windthrow probabilities at the Clayoquot site, and at both study areas there was a linear relationship between post-harvest density and windthrow levels. Where fewer than 5% of stems 71 are retained, windthrow risk will likely be reduced by keeping openings small (less than 60m across). There are easily measured tree-level attributes that can be used to predict susceptibility to windthrow. The most windfirm trees within the coastal stands studied here tend to be trees with height-diameter ratios of less than 60 and percent live crown of greater than 40%. Targeting these lower-risk trees for retention should result in a more wmdfirm stand after harvest. At the same time, retaining only low-risk trees will have implications for the future value of the stand, as trees with low height-diameter ratios and high percent live crown may be less valuable than more slender, cylindrical, and windthrow-prone trees. The results of this study will be incorporated into other ongoing modelling efforts, and will contribute to new tools currently being developed which allow interactive assessment of the risk of different cutblock designs (S. Mitchell, personal communication). Initiating an on-going program of windthrow monitoring would also be worthwhile. Windthrow monitoring with randomly or systematically located plots will allow foresters to compare outcomes to goals, and assess whether planned retention levels are met or maintained. Monitoring will also identify risk factors and provide data for quantitative modelling. At the moment, the most efficient means of obtaining tree-level windthrow data is through field surveys. Combining windthrow sampling with other post-harvest surveys (e.g., regeneration surveys), may be one way of obtaining these data at little extra cost Low-elevation aerial photos are also valuable for mapping post-harvest retention levels and patterns, particularly in very complex variable retention stands. In this thesis, windthrow risk is defined as the probability of windthrow occurring, and should not be confused with the possible consequences that might result from the occurrence of windthrow (e.g. sedimentation, failure to maintain minimum retention levels). The findings 72 presented here should help managers identify high-risk trees. It remains the responsibility of forest managers to determine the impact of wmdthrow for a given tree, and decide what level of risk is tolerable. 73 CHAPTER 7 - RECOMMENDATIONS The following recommendations are for researchers or forest managers interested in furthering understanding of windthrow: 1. Monitoring of windthrow in variable retention stands should be undertaken which includes measurements of individual tree characteristics. Ideally, monitoring would occur before or immediately after harvest, and again two or three years after harvest. 2. Relating windthrow to retention patterns requires accurate maps of retention features. Low-elevation colour air photos (scale of 1:5000 or less) have proven to be useful for mapping complex retention patterns within cutblocks, and can also be used to estimate retention levels. Such photos have shown promise for windthrow monitoring in low levels of variable retention (Rollerson et al. 2002), but were not detailed enough to allow consistent identification of wmdthrown trees in the high levels of retention found at Clayoquot. 3. In the future, the use of retention systems in uniform and second growth stands will likely increase. More tree-level research should be undertaken in these stands. 4. Regional differences in tree and plot-level windthrow risk should be explored further, and the portability of the models developed here should be tested with data from other areas. 74 5. More accurate and finer resolution models of windflow over terrain and through canopies should be developed. 6. The calculation of VRFETCH and other fetch variables should be automated as much as possible, and the variables themselves should be incorporated into harvest-design tools currently being developed by the UBC windthrow research group. More wind tunnel work with model forests should be done to test the basis for the fetch indices developed here and to establish general equations for fetch after partial-cutting. 75 REFERENCES Alexander, R A . 1964. Mininiizing windfall around clearcuttings in spruce-fir forests. Forest Science 10:130-142. BC Hydro. 2001. Wind energy resource mapping consulting services for British Columbia. True Wind Solutions, LLC, Albany, New York. 8 pp. Beasley, B., R. Scott, D. Edwards, C. Jacobsen, M. Moeges, and K. McNutt. 2002. Monitoring and Demonstrating the Scientific Panel Recommendations for Sustainable Forestry in Clayoquot Sound. Report for Long Beach Model Forest Society and Iisaak Forest Resources Ltd., Ucluelet, BC. 180 pp. Beese, WJ. 2001. Windthrow monitoring of alternative silvicultural systems. In Proceedings of the Windthrow Researchers Workshop, January 31-February 1,2001, Richmond, BC. Compiled by S.J. Mitchell and J. Rodney, pp. 2-11. Bergerud, W A . 1996. Introduction to Logistic Regression Models: with worked forestry examples. Biometrics Information Handbook No.7. Research Branch, BC Ministry of Forests, Victoria, BC. Working Paper 26/1996.147 pp. Burton. P.J. 2001. Windthrow patterns on cutblock edges and in retention patches in the SBSmc. In Proceedings of the Windthrow Researchers Workshop, January 31-February 1, 2001, Richmond, BC. Compiled by S.J. Mitchell and J. Rodney, pp.19-31. Bull, G A.D. and E.R.C Reynolds. 1968. Wind turbulence generated by vegetation and its implications. Forestry (Supp.) 41:28-37. Canham, C D . and O.L Loucks. 1984. Catastrophic windthrow in the presettlement forests of Wisconsin. Ecology 65:803-809. Canham, CD. , M.J. Papaik, and E.F. Latty. 2001. Interspecific variation in susceptibility to windthrow as a function of tree size and storm severity for northern temperate tree species. Canadian Journal of Forest Research 31:1-10. Clayoquot Sound Scientific Panel. 1995. Report 5: Sustainable Ecosystem Management in Clayoquot Sound. Cortex Consultants Inc., Victoria, BC. 296 pp. Coates, K. D. 1997. Windthrow damage 2 years after partial-cutting at the Date Creek silvicultural systems study in the Interior Cedar-Hemlock forests of northwestern British Columbia. Canadian Journal of Forest Research 27:1695-1701. Coates, K. D. 2001. Do spatial patterns of retained canopy trees after partial-cutting affect the risk and extent of wind damage? In Proceedings of the Windthrow Researchers Workshop, January 31-February 1,2001, Richmond, BC. Compiled by S.J. Mitchell and J. Rodney, pp.157-165. Cremer, K.W., C J . Borough, F.H. McKinnell, and P.R. Carter. 1982. Effects of stocking and thinning on wind damage in plantations. New Zealand Journal of Forestry Science 12:244-265. 76 de Montigny, L. 2004. Silviculture Treatments for Ecosystem Management in the Sayward (STEMS): Establishment Report for STEMS 1, Snowden Demonstration Forest. Research Branch, BC Ministry of Forests. Technical Report 17. Victoria, BC. Dunham, R.A. and A.D. Cameron. 2000. Crown, stem and wood properties of wind-damaged and undamaged Sitka spruce. Forest Ecology and Management 135:73-81. Elling, A.E., and E.S. Verry. 1978. Predicting wind-caused mortality in strip-cut stands of peatland black spruce. Forestry Chronicle 54:249-252. Environment Canada. 1992. Marine weather hazards manual: a guide to local forecasts and conditions, West Coast edition. 2n d Ed. Gordon Soules Book Publishers Ltd., Environment Canada and the Canada Communication Group PiibUshing, Supply and Services Canada. West Vancouver, BC. Environment Canada. 2004. Canadian Climate Normals 1971-2000 for Campbell River A and Tofino A. Last accessed September 21,2004 at: htto://www.climate.weameroffice.ec.gc.ca/climate normals ESRI. 1996. Using Arc View GIS. Environmental Research Systems International Inc. Redlands, CA. Everham, E.M. and N.V.L. Brokaw. 1996. Forest damage and recovery from catastrophic wind. Botanical Review 62:113-185. Fleming, R.L., and R.M. Crossfield. 1983. Ship-cutting in shallow-soil upland black spruce near Nipigon, Ontario. III. Windfall and Mortality in the leave-strips: preliminary results. Great Lakes Forest Research Centre. Canadian Forestry Service. 28 pp. Flesch T.K. and J.D. Wilson. 1998. Wind and remnant tree sway in forest cutblocks. I. Measured winds in experimental cutblocks. Agricultural and Forest Meteorology 93:229-242. Forest Practices Code of BC Act. 1999. Operational Planning Regulations. S.B.C., c.41. Foster, D.R. 1988. Species and stand response to catastrophic wind in central New England, U.S.A. Journal of Ecology 76:135-151. Foster, D.R., and E.R. Boose. 1992. Patterns of forest damage resulting from catastrophic wind in central New England, USA. Journal of Ecology 80:79-98. Gardiner, B.A., CP. Stacey, R.E. Belcher, and C.J. Wood. 1997. Field and wind tunnel assessments of the implications of respacing and thinning for tree stability. Forestry 70:233-252. Gardiner, B.A. and CP. Quine. 2000. Management of forests to reduce the risk of abiotic damage - a review with particular reference to the effects of strong winds. Forest Ecology and Management 135:261-277. Gratkowski, HJ . 1956. Windthrow around staggered settings in old-growth Douglas-fir. Forest Science 2:60-74. Green, R.N. and K. Klinka. 1994. A field guide to site identification and interpretation for the Vancouver Forest Region. Research Branch, BC Ministry of Forests. Land Management Handbook No. 28. Victoria, B C 77 Grizzel, J.D. and N. Wolff. 1998. Occurrence of windthrow in forest buffer strips and its effect on small streams in Northwest Washington. Northwest Science 72:214-223. Hannah, P., J.P. Palutikof and C.P. Quine. 1995. Predicting windspeeds for forest areas in complex terrain. In Wind and Trees. Edited by M.P. Coutts and J. Grace. Cambridge University Press, Cambridge, U.K. pp. 113-129. Harris, A.S. 1989. Wind in the forests of southeast Alaska and guides for reducing damage. USDA Forest Service. General Technical Report PNW-GTR-244. 63 pp. Holmes, S.R. 1985. An analysis of windthrow along clearcut boundaries in the Tsitika watershed. B. A. Thesis. University of British Columbia. 104pp. Hosmer, D.W. and S. Lemeshow. 1989. Applied Logistic Regression. Wiley. Toronto, Ont. 307 PP-Huggard, D.J., W. Klennar, and A. Vyse. 1999. Windthrow following four harvest treatments in an Engelmann spruce - subalpine fir forest in southern interior British Columbia, Canada. Canadian Journal of Forest Research 29:1547-1556. Hutte, P. 1968. Experiments on windflow and wind damage in Germany; site and susceptibility of spruce forests to storm damage. Forestry (Supp.) 41:20-26. Jalkanen A. and U. Mattila. 2000. Logistic regression models for wind and snow damage in northern Finland based on the National Forest Inventory data. Forest Ecology and Management 135:315-330. Jull, M. and B. Sagar. 2001. Wind-resistant silvicultural systems research project: Identification of wind risk factors and mitigation measures in central and northern Interior partial cuts. Five-year final report to FRBC. Prince George, BC. 32 pp. Kramer, M.G., A J . Hansen, M.L. Taper and E J Kissinger. 2001. Abiotic controls on long-term windthrow disturbance and temperate rainforest dynamics in Southeast Alaska. Ecology 82:2749-2768. Lanquaye, C O . 2003. Empirical modelling of windthrow risk using GIS. Master's Thesis. University of British Columbia, Vancouver, BC. Lee, X. 2000. Air motion within and above forest vegetation in non-ideal conditions. Forest Ecology and Management 135:3-18. Lekes, V. and I. Dandul. 2000. Using airflow modelling and spatial analysis for defining wind damage risk classification (WINDARC). Forest Ecology and Management 135:331-344. Lertzman, K., G.D. Sutherland, A. Inselberg and S.C Saunders. 1996. Canopy gaps and the landscape mosaic in a coastal temperate rainforest. Ecology 77:1254-1270. Lohmander, P. and F. Helles. 1987. Windthrow probability as a function of stand characteristics and shelter. Scandinavian Journal of Forest Research 2:227-238. MacCurragh R.S. 1991. Spacing: an option for reducing storm damage. Scottish Forestry 45:285-297. Meunier, S., J-C. RueL G. Laflamme and A. Achim. 2002. Resistance compare de l'epinette blanche et du sapin baumier au renversement. Canadian Journal of Forest Research 32:642-652. 78 Ministry of Sustainable Resources Management. 2002. Vegetation resources inventory: ground sampling procedures. Version 4.3. Terrestrial information branch, Resources Inventory Committee. 230 pp. Minore, D. 1990. Thuja plicata. In Silvics of North America: 1. Conifers. R.M. Burns and B.H. Honkala, tech. coords. Agriculture Handbook 654, US Department of Agriculture, Forest Service, Washington, DC. pp.590-600. Mitchell, SJ. 1995. A synopsis of windthrow in British Columbia: occurrence, implications, assessment and management. In Wind and Trees. Edited by M.P. Courts and J. Grace. Cambridge University Press, Cambridge, U.K. pp. 448-459. Mitchell, S J . 1998. The windthrow triangle: A relative windthrow hazard assessment procedure for forest managers. Forestry Chronicle 71:446-450. Mitchell, S J . 2000. Stem growth responses in Douglas-fir and Sitka spruce following thinning: implications for assessing wind-firmness. Forest Ecology and Management 135:105-114. Mitchell, S.J., T. Hailemariam and Y. Kulis. 2001. Empirical modelling of cutblock edge windthrow on Vancouver Island, Canada, using stand level information. Forest Ecology and Management 154:117-130. Mitchell, S J . , and WJ. Beese. 2002. The retention system: reconciling variable retention with the principles of silvicultural systems. Forestry Chronicle 78:397-404. Mitchell, SJ. 2003. Windthrow hazard mapping using GIS, Weyerhaeuser West Island Timberlands. Final Report prepared for Bill Beese, Weyerheauser Ltd., Nanaimo, BC. 30pp. Mitchell, S J . , A. Pearson, J.P. Kimmins, N. Lanquaye and S. Ortlepp. 2004. Investigation of windthrow as a natural disturbance agent in coastal BC forests. FII Technical Report, Project R04-074. Vancouver, BC. 53 pp. Moore, M.K. 1977. Factors contributing to blowdown in streamside leave strips on Vancouver Island. Ministry of Forests Land Management Report No. 3. Victoria, BC. 34 pp. Murphy, B. and P.L. Jackson. 1997. Extreme value analysis: return intervals of severe wind events in the central interior of British Columbia. MacGregor Model Forest Association, Prince George, BC. 38 pp. Neilsen, C.C.N. 1995. Recommendations for stabilisation of Norway spruce stands based on ecological surveys. In Wind and Trees. Edited by M.P. Courts and J. Grace. Cambridge University Press, Cambridge, U.K. pp. 424-435. Nicoll B.C. and D. Ray. 1996. Adaptive growth of tree root systems in response to wind action and site conditions. Tree Physiology 16:891-898. Novak, M.D., J.S. Warland, A.L. Orchansky, R. Ketler and S. Green. 2000. Wind tunnel and field measurements of turbulent flow in forests. Part I: Uniformly thinned stands. Boundary-Layer Meteorology 95:457-495. Nowacki, G J . and M.G. Kramer. 1998. The effects of wind disturbance on temperate rain forest structure and dynamics of southeast Alaska. U.S. Dept. of Agriculture, Forest Service, Pacific Northwest Research Station. General Technical Report PNW-GTR-421. 25 pp. 79 Oliver, C H . and B.C. Larson. 1990. Forest Stand Dynamics. McGraw-Hill Inc., New York. 467 pp. Pearson, A.F. 2000. Natural disturbance patterns in a coastal temperate rain forest watershed, Clayoquot Sound, British Columbia (PhD). University of Washington, Seattle, WA. 168 pp. Peltola, H. and S. Kellomaki. 1993. A mechanistic model for calculating windthrow and stem breakage of Scots pines at stand edges. Silva Fennica 27:99-111. Peltola, H., S. Kellomaki, A. Hassinen and M. Granander. 2000. Mechanical stability of Scots pine, Norway spruce and birch: an analysis of tree-pulling experiments in Finland. Forest Ecology and Management 135:143-153. Peterson, C J . 2004. Within-stand variation in windthrow in southern boreal forests of Minnesota: Is it predictable? Canadian Journal of Forest Research 34:365-375. Pojar, J. and A. MacKinnon (Eds.) 1994. Plants of the Pacific Northwest Coast. Lone Pine Publishing, Vancouver, BC. 526 pp. Quine, CP. 1994. An improved understanding of windthrow - moving from hazard towards risk. Research Information Note 257. The Forestry Authority, Research Division, Forestry Commission, Edinburgh, Scotland. 6pp. Quine, CP. and I.M.S. White. 1994. Using the relationship between rate of tatter and topographic variables to predict site windiness in upland Britain. Forestry 67:245-256. Quine, CP. and I.M.S. White. 1998. The potential of distance-limited topex in the prediction of site windiness. Forestry 71:325-332. Rees, D.J., and J. Grace. 1980. The effects of wind on the extension growth of Pinus contorta Douglas. Forestry 53:145-153. Rollerson, T.P. 1981. Queen Charlotte Woodlands Division Windthrow Study. MacMillan-Bloedel Limited Woodland Services, Nanaimo, B C 32 pp. Rollerson, TP. and K. McGourlick. 2001. Riparian Windthrow - Northern Vancouver Island. In Proceedings of the Windthrow Researchers Workshop, January 31-February 1,2001, Richmond, BC. Compiled by S.J. Mitchell and J. Rodney, pp. 139-156. Rollerson, TP., WJ. Beese and C M . Peters. 2002. Variable Retention Windthrow Monitoring Pilot Project. Forest Renewal BC Progress Report. Nanaimo, BC. 71 pp. Rowan, C , S J . Mitchell and T. Hailemariam. 2001. Edge wmdfirming treatments in coastal British Columbia. In Proceedings of the Windthrow Researchers Workshop, January 31-February 1,2001, Richmond, BC. Compiled by S J . Mitchell and J. Rodney, pp.205-222. Ruel, J.-C. 1995. Understanding windthrow: silvicultural implications. Forestry Chronicle 71:434-445. Ruel, J-C. 2000. Factors influencing windthrow in balsam fir forests: from landscape studies to individual tree studies. Forest Ecology and Management 135:169-178. Ruel, J-C, CP. Quine, S.Meunier and J. Suarez. 2000. Estimating windthrow risk in balsam fir stands with the ForestGales model. Forestry Chronicle 76:329-337. 80 Ruel, J-C, SJ. Mitchell, and M. Dornier. 2002. A GIS based approach to map wind exposure for windthrow hazard rating. Northern Journal of Applied Forestry 19:183-187. Ruel, J-C, P. Raymond, and M. Pineau. 2003. Windthrow after shelterwood cutting in balsam fir stands. Northern Journal of Applied Forestry 20:5-13. Rullie, J. and N. Shah. 1998. Keep it simple! Orthorectification alternatives save time and money. GIS World 11:58-62. Rudnicki, M., S J . Mitchell, and M.D. Novak. 2004. Wind tunnel measurements of crown streamlining and drag relationships for three conifer species. Canadian Journal of Forest Research 34:666-676. SAS Institute Inc. 2001. SAS/STAT User's Guide, Version 8.2. SAS Institute, Cary, NC. Scott, R. and B. Beasley. 2001. Variable retention windthrow monitoring report. Long Beach Model Forest Society Report. Ucluelet, BC. 42 pp. Sinton, D.S., J.A. Jones, J.L. Ohmann and FJ . Swanson. 2000. Windthrow disturbance, forest composition, and structure in the Bull Run Basin, Oregon. Ecology 81:2539-2556. Smith, V.G., M. Watts and D.F. James. 1987. Mechanical stability of black spruce in the clay belt region of northern Ontario. Canadian Journal of Forest Research 17:1080-1091. Stacey, G.R., R.E. Belcher and C J . Wood. 1994. Wind flows and forces in a model spruce forest. Boundary-Layer Meteorology 69:311-334. Stathers, R.J., T.P. Rollerson and S J . Mitchell. 1994. Windthrow handbook for British Columbia forests. BC Ministry of Forests, Victoria, BC. Working Paper 9401. 31 pp. Stokes, A., A.H. Fitter and M.P. Courts. 1995. Responses of young trees to wind: effects on root growth. In Wind and Trees. Edited by M.P. Courts and J. Grace. Cambridge University Press, Cambridge, U.K. pp. 264-275. Tabachnick, B.G. and L.S. Fidell. 2001. Using multivariate statistics. 4th edition. Allyn and Bacon, Toronto. Talkkari, A., H. Peltola, S. Kellomaki, and H. Strandman. 2000. Integration of component models from the tree, stand and regional levels to assess the risk of wind damage at forest margins. Forest Ecology and Management 135:303-313. Telewski, F.W. 1995. Wind-induced physiological and developmental responses in trees. In Wind and Trees. Edited by M.P. Courts and J. Grace. Cambridge University Press, Cambridge, U.K. pp. 237-263. Valinger. E. and J. Fridman. 1997. Modelling probability of snow and wind damage in Scots pine stands using tree characteristics. Forest Ecology and Management 97:215-222. Valinger, E. and J. Fridman. 1999. Models to assess the risk of snow and wind damage in pine, spruce, and birch forests in Sweden. Environmental Management 24:209-217. Veblen, T.T., D. Kulakowski, K.S. Elsenhart, W.L. Baker. 2001. Subalpine forest damage from a severe windstorm in northern Colorado. Canadian Journal of Forest Research 31:2089-2097. Wang, Y., S J . Titus and V.M. LeMay. 1998. Relationships between tree slenderness coefficients and tree or stand characteristics for major species in boreal mixedwood forests. Canadian Journal of Forest Research 28:1171-1183. 81 Waring, R.H. and W.H. Schlesinger. 1985. Forest Ecosystems: Concepts and Management. Academic Press, Inc. Orlando, Florida. 340 pp. Whiteman, C D . 2000. Mountain Meterology: Fundamentals and applications. Oxford University Press. Oxford. 355pp. 82 A P P E N D I C E S 83 APPENDIX I: DESCRIPTION OF VARIABLES Table I. A . Description of variables. Variable Label Description Units Tree height HT Height to live HTLC crown (bottom) Height to top of HTTOPLC crown Diameter at breast DBH height Percent live crown %LC Height-diameter HDR ratio Crown density CRNDEN Crown class CRNCLASS Response variable WTSTAT Species SPP Direction of fall DIROFFALL Substrate SUBST Tree variables Height of tree to nearest decimeter; measured with an m impulse laser gun, or using a clinometer and tape measure Height at which branches take up 75% of m circumference of tree; for trees with sparse or asymmetrical crowns, height to first substantial live branch Height to the top of the live crown, if different from tree height (e.g. snag-top cedars) Diameter at breast height (1.3m); measured with a cm DBH tape to nearest half-centimetre = [(height to top of crown - height to bottom of % crown) / total height of tree] X 100 HT (m) / DBH (m) Visual estimate of crown density (full, moderate, sparse) Crown class in relation to position in canopy (dominant, codominant, intermediate, suppressed) Windthrow response variable, l=windthrown, 0=standing Tree species CW= Western redcedar HW= Western hemlock BA=Amibilis fir YC=y el low-cedar Direction of fall for windthrown trees; measured with degrees a compass to the nearest degree Substrate that tree is immediately growing on; 0=organic, M=mineraL W=wood 84 Table I. A. (Continued). Description of variables. Variable Label Description Units Fetch Simple fetch VR fetch DIREX30 DIREX60 DIREX90 Percent stems retained Percent basal area retained Percent canopy retained Retention type Time since harvest ended Permeable soil depth FETCH SIMPLEFETCH VRFETCH DIREX30 DIREX60 DIREX90 Neighbourhood variables Sum of continuous distance of 0% canopy cover surrounding each plot in eight cardinal directions, to a maximum of300m in any direction Sum of distance of 0% canopy cover surrounding each plot in eight cardinal directions, to a maximum of 300m in any direction Distance multiplied by removal level every 30m for 300m in eight cardinal directions surrounding the plot Number of cardinal directions with at least 30m of clear fetch min=0, max=8 • Number of cardinal directions with at least 60m of clear fetch Number of cardinal directions with at least 90m of clear fetch STEMSRET percent stems retained within plot BARET CANOPYRET RETN TYPE percent basal area retained within plot % of canopy cover retained to the nearest 10%, mapped from low-elevation air photos in Arc View retention type interpreted from air photos, based on both the type of retention within the polygon, and the type of retention adjacent to it (aggpatch, aggbdry, dispatch, dispbdry, harvested) TIMEHARV 2003 - year harvested % % % years SOILDEPTH Depth of soil in plot, measured by pusing a 50cm long cm piece of 6mm diameter steel rod into the ground; averaged for five locations within the plot Post-harvest density POSTDENS Post-harvest density, calculated for each plot PREDENS Pre-harvest density Pre-harvest basal PREBA area Post-harvest basal POSTBA area Pre-harvest density, calculated for each plot Pre-harvest basal area, calculated for each plot Post-harvest basal area, calculated for each plot sph sph m2/ha m2/ha 85 Table I. A . (Continued). Description of variables. Variable Label Description Units Spacing factor S/H S/H = distance between trees (m) divided by average tree height (m) in the plot Post-harvest density (sph) was used to estimate the distance between trees. Soil moisture class MOIST Relative soil moisture based on site series. CWHvhl: DRY=3; FRESH=4, 5; WET=1, 6, 7,11, 12,13 CWHvml, vm2: Dry=3,4 ; FRESH=T, 5; WET=6, 7, 13 Soil nutrient class NUTRIENTS Relative soil nutrient class based on site series. CWHvhl: POOR=3,4, 1,11,12; RICH=5, 6, 7, 13 CWHvml, vm2: POOR= 3,1, 6,13; RICH=4 ,5 ,7, 11, 14 Stand structure STANDTYPE M=Multi-storied (closed forest with all crown classes well represented) S=single-storied (closed forest comprised mainly of dominants and codominants) Lead species VOL_LEADSPP Lead species in the stand, by volume Stand height S T D H T Average stand height from cruise data at Clayoquot and from forest cover at STEMS m Stand volume STD_VOL Average stand volume, from cruise data at Clayoquot, from forest cover at STEMS m3/ha Stand density STD_SPH Average stand density, from cruise data at Clayoquot from forest cover at STEMS sph Stand age S T D A G E Average stand age, from cruise data at Clayoquot from forest cover at STEMS years Site Index SI Site index (in metres at age 50), from forest cover data m Topex lkm,Topex 2 km, Topex 3 km TOPEX1K, TOPEX2K, TOPEX3k Distance-limited topographic exposure calculated at 1000m, 2000m and 3000m; derived from DEM in ArcView degrees Mean wind speed WIND Mean annual windspeed km/hr Elevation ELEV Elevation, derived from DEM in ArcView m Slope SLOPE Slope, derived from DEM in ArcView degrees Aspect ASPECT Direction that site is facing, derived from DEM in ArcView degrees Distance from coast DISTCOAST Distance from coast in 0.5 km classes km 86 APPENDIX n : VARIABLE SUMMARIES OVER ALL PLOTS Table II. A. Summary of variables at Clayoquot over all plots (11=234). Variable Label Mean Std Dev Min Max Tree variables Tree height (m) MEANHT 21 6 10 44 Diameter at breast height (cm) MEANDBH 43 19 17 134 Height-diameter ratio MEANHDR 57 12 24 104 Live crown ratio MEAN%LC 47 11 23 75 Neighbourhood variables Percent stems windthrown PCTSTEMSWT 19 25 0 100 VRFetch VRFETCH 760 304 105 1644 DIREX30 DIREX30 3 2 0 8 Percent stems retained STEMSRET 66 26 4 100 Percent basal area retained BARET 56 33 1 100 Percent canopy retained CANOPYRET 63 37 0 100 Time since harvest ended (years) TIMEHARV 2 1 0 6 Permeable soil depth (cm) SOILDEPTH 31 9 9 50 Post-harvest density (sph) POSTDENS 376 240 25 1400 Stand variables Stand height (m) STD HT 25 6 13 43 Stand volume (m2) STD VOL 577 206 230 1392 Stand density (SPH) STD SPH 455 124 100 903 Topex 1km (°) TOPEX1K 35 34 -26 136 Mean wind speed (m/s) WIND 5 1 4 7 Elevation (m) ELEV 184 164 22 648 Slope (°) SLOPE 11 10 0 43 Aspect (°) ASPECT 201 100 3 358 Distance from coast (km) DISTCOAST 3 1 1 7 Windthrow probability as predicted by WIT model WITWTRISK 0.34 0.09 0.17 0.74 87 Table II. B. Summary of variables at STEMS over all plots (n=l 15). Std Variable Label Mean Dev Minimum Maximum Tree variables Mean tree height MEANHT 29 5 12 42 Mean tree dbh MEANDBH 34 6 20 67 Mean HDR MEANHDR 83 9 58 112 Mean %LC MEAN%LC 45 11 25 85 Neighbourhood variables Percent stems windthrown PCTWT 17 29 0 100 Fetch (m) FETCH 302 459 0 1470 Vrfetch VRFETCH 659 478 30 1686 Direx DIREX30 3 3 0 8 Percent canopy retained CANOPYRET 69 44 0 100 Post-harvest density (sph) POSTDENS 359 240 10 1020 Stand variables Stand height (m) S T D H T 23 2 21 32 Site Index SI 33 3 29 38 Topexlk(°) TOPEX1K 17 11 -6 44 Mean wind speed (m/s) MEANWIND 4 0 4 4 Slope (°) SLOPE 7 3 1 12 Elevation (m) ELEV 212 18 166 247 Aspect (°) ASPECT 157 126 7 356 Distance to the coast (km) DIST COAST 7 1 6 8 Windthrow probability as predicted by NIT model NITWTPROB 0.2 0.01 0.19 0.23 88 APPENDIX III: DUMMY VARIABLE CODING FOR CATEGORICAL VARIABLES Table III. A. Coding for dummy variables used in Clayoquot modelling. Dummy Class Variable Values variable 1 2 3 4 Crown density Full 1 0 (CRNDEN) Moderate 0 1 Sparse 0 0 Species Ba 1 0 0 (SPP) Cw 0 1 0 Hw 0 0 1 Yc 0 6 0 Substrate Organic 1 (SUBST2) Other 0 Aspect North 1 0 0 (ASPECT) East 0 1 0 West 0 0 1 South 0 0 0 Retention Aggbdry 1 0 0 0 Type Aggpatch 0 1 0 0 (RETNTYPE) Dispbdry 0 0 1 0 Disppatch 0 0 0 1 Harv 0 0 0 0 Table III. B. Coding for dummy variables used in STEMS modelling. Class Variable Values 1 2 3 Species Fd 1 0 Cw 0 1 Hw 0 0 Crown class Dominant 1 0 0 Codominant 0 1 0 Intermediate 0 0 1 Suppressed 0 0 0 89 APPENDIX IV: INTERCEPT AND PARAMETER ESTIMATES FOR INITIAL MODELS Table IV. A. Parameter estimates for initial Clayoquot models. Variable Tree only Neighbourhood only Stand only Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Intercept -3.2299 -3 7582 -1.5775 -1.157*5 -2.3482 -3.2817 HDR 0 0284 0 0314 %LC -0.0161 -0.0262 CRNDEN F 0.8270 1.0891 CRNDEN M 0.2435 0.7336 SUBSTRATE 0.5452 1 1782 VRFETCH 0.00133 0 0007^ 2 POSTDENS -0.00164 -0.00243 TIMEHARV -0.2099 PREBA -0.00329 DIREX30 0.1006 TOPEXlk , 0.00684 DISTCOAST 0.1667 WIND . 0.3891 ASPCLNvsS -0.9289 ASPCL hvsS , ' ' -0.8932 ASPCLWvsS -0.3752 90 Table IV. A. (Continued). Parameter estimates for initial Clayoquot models. • Tree and Neighbourhood and Tree, Neighbourhood Variable Neighbourhood Tree and Stand Stand and Stand Model 1 Model 2 Model.l Model 2 Model 1 Model 2 Model 1 Model 2 Intercept -3.2967 -3.4625 -4.3487 -3.541 -2.8039 -2.3428 -4.4598 ' ' -2.2848 HDR 0.0288 0.0463 0.0307 0 0328 0.0316 0.0492 %LC -0.0192 -0.0308 -0.0140 -0 0253 -0.0 73 -0.0312-CRNDEN 1 1 0241 1.4267 0.9302 1.1241 1.064 1.4682 CRNDEN 2 0.3665 0.8663 0.2012 0.7669 0.3058 0.8933 SUBSTRATE 0.5075 1.2373 0.4564 1.2622 0.4172 1.2915 SPECIES 1 -1.8695 13113911 " -1.4642 SPECIES 2 -0.5806 -0.3203 SPECIES 3 -1.0327 lllljllljf^  ^ ^ ^ ^ ^ ^ -0.7875 VRFETCH 0.0011 0.00083 0.00147 0.00103 0 0014 0.00104 POSTDENS -0.0017 -0.0033 -0 0023 -0.0025 -0.00241 - -0.0033 TIMEHARV -0.2009 -0.1879 -0.1965 DIREX30 0 0973 0.1245 §|§§ltH81§fl 0.1144 SOILDEPTH -0.0266 -0.0381 TOPEX1K 0.00577 Jl§tll|l||t|ll 0.00549 DISTCOAST 0.2338 0.339 0.3651 WIND J8fltlt§lfllSl8§i 0 4201 ASPCL 1 -1.021 -0.9986 - -1.1714 ASPCL 2 ^ ^ ^ ^ ^ ^ -0.8991 -0.903 -1.0684 ASPCL 3 -0.1876 -0.5159 -0.5794 SLOPE ilBlllllB 0.0266 91 Table IV. B. Parameter estimates for initial STEMS models. Neighbourhood Variable Tree only only Stand only Model Model Model Model Model Model 1 2 1 2 1 2 Intercept -13167 1 4389 -3 8418 -4.0673 -2 1502 -3.8323 DBH -0.0466 SPECIES 1 0.1464 1111311! SPECIES 2 -2.2389 % L C -0.0559 DIREX30 0.4409 0.5152 DISTCOAST J|t|§|||§§|^  iiiitfitfit 1 2946 ' , 1.617 ELEV -0.0448 -0.0472 Table IV. B. (Continued). Parameter estimates for initial STEMS models. Tree, Tree and Neighbourhood Neighbourhood Variable Neighbourhood Tree and Stand and Stand and Stand Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Model 1 Model 2 Intercept -1 9912 -1 117 1 1771 2.0488 -7 3457 -10 176 -5.2569 ' -6.8139 DBH -0.056 -0.0579 -0.0552 -0.0431 -0.054 -0.0512 SPECIES 1 lltlllttlttftlt SPECIES 2 % L C -0 0268 -0.0277 tl|llf§lll| 181111 fflltlSillSf TO.0371 DIREX30 0.4564 0.5592 0.4007 0.4995 0.4196 0.5606 DIS'I COAST 1 3109 1 91 05195 0.9825 0 4745 > 6.9718 ELEV -0.0521 -0.068 S L O P E -0.1337 -0 1113 • -0.1446 92 APPENDIX V: STAND-LEVEL WINDTHROW RISK MODEL CALCULATIONS WIT WINDTHROW MODEL CALCULATIONS (from Mitchell 2003) LogitWiT= (-5.0134) + (STDAGE x 0.00231) - (TOPEX3K x 0.00645) + (CBRGCN x 0.00810) + (SI x 0.0211) + (MEANWIND x 0.2987) + (FETCH x 0.000200) + (TIMEHARV x 0.0344) + (DIREX x 0.2452) WITwtprob= exp (Logitwrr ) exp (1+ Logitwrr) Variables: STD AGE = stand age in years. This value was unknown for some stands, so a mean value of 300 years was used (from forest cover data). TOPEX3k = topex calculated at a distance of 3000m CBRGCN = the bearing from a given edge segment to the centroid of the block. This could not be calculated for the Clayoquot plots, as they were not located on edges. Instead, this variable was held constant at 180° (assumes each plot is exposed towards the south). SI = site index in metres at age 50. This value was unknown for some stands, so a mean value of 28 was used (from forest cover data) MEANWIND = mean annual windspeed from BC Hydro data (km/hr) TIMEHARV = time since harvest in years DIREX= Number of the 8 cardinal directions with at least 100m of fetch, DIREX90 was used as the closest substitute. 93 NIT WINDTHROW MODEL CALCULATIONS (from Lanquaye 2003) LogitNrr = (-6.1280) + (DIREX x 0.250) - (TOPEX1K x 0.0028) + (MEANWIND x 0.4792) + (STDHT x 0.0165) + (CBRGCN x 0.0083) - (CATTACK x 0.0039) NITwtprob= exp (LogitNrr) exp (1+ LogitNrr) Variables: DIREX = number of 8 cardinal directions with at least 100m of fetch, DIREX90 was used as the closest substitute. TOPEXlk = topographic exposure calculated at a distance of 1000m MEANWiND= mean annual windspeed from BC Hydro data (km/hr) S T D H T = stand height in metres. CBRGCN = the bearing from a given edge segment to the centroid of the block. This could not be calculated for the STEMS plots, as most were not located on edges. Instead, this variable was held constant at 180° (assumes each plot is exposed towards the south). CATTACK - Cosine of the angle between the boundary bearing and the prevailing wind direction. Held constant at 1 (greatest exposure). 94 APPENDIX VI: DETAILS OF IMAGES USED TO MAP RETENTION POLYGONS Table VI. A. Details of aerial photos and digital images used in retention mapping Category Clayoquot photos STEMS aerial photos Type of imagery IKONOS satellite imagery Very low-elevation aerial photos Low-elevation aerial photos Orthophotos Orthophotos Orthophotos UTM Coordinates left 275027 various various 92E049, Left 1038887 Left 1039138 or mapsheets bottom 5455892 right 288612 top 5469788 92E058,92F021 92F003, 92F012 Bottom 560232 Right 1042642 Top 563288 Bottom 60353 Right 1042012 Top 562772 Number of images 1 64 12 5 1 1 Colour/B&W Colour Colour Colour Colour B&W Colour Date taken 2001 October 14,2003 2000 1996 Pre-harvest Post-harvest Scale of photos -1:3400 to 1:4000 1:5000-1:10000 1:20000 1:5000 unk Pixel resolution Im pixel resolution <1 m <1 m ~5 m 6 m 2m Source Interfor Bazett Aerial Photography Interfor Ministry of Forests Ministry of Forests Ministry of Forests 95 APPENDIX VII: EXAMPLE OF FETCH CALCULATIONS The following table shows the actual values used to calculate the fetch variables for one plot, HE2021-10. Each row in the table corresponds to one of the points surrounding the plot (Figure 6). Each point is assigned a retention level based on the amount of canopy retained within the retention polygon where the point is located (CANOPYRET). The distance between points is 30m, and the retention level at each point is assumed to be representative of retention level over the distance between points. Values for each of the fetch variables were calculated for each point surrounding the plot, as follows: VRFETCH = 30m x (1-CANOPYRET/l 00) SIMPLEFETCH = 30 IF CANOPYRET<5, otherwise SIMPLEFETCH =0 FETCH = 30 if CANOPYRET for the given point and all points closer to the plot <5%, otherwise FETCH = 0 VRFETCH, SIMPLEFETCH and FETCH values for each point were then summed to obtain fetch values for the plot (see bottom of spreadsheet). Table VTJ. A. Values used for fetch calculations for plot HE2021-10. POINT DIR DIST (m) X-COORD Y-COORD CANOPYRET VRFETCH SIMPLE FETCH FETCH E30 E 30 964700 499140 0 30 30 30 E60 E 60 964730 499140 100 0 0 0 E90 E 90 964760 499140 0 30 30 0 E120 E 120 964790 499140 0 30 30 0 E150 E 150 964820 499140 0 30 30 0 E180 E 180 964850 499140 0 30 30 0 E210 E 210 964880 499140 0 30 30 0 E240 E 240 964910 499140 0 30 30 0 E270 E 270 964940 499140 0 30 30 0 E300 E 300 964970 499140 0 30 30 0 N30 N 30 964670 499170 0 30 30 30 N60 N 60 964670 '.- 499200 100 0 0 0 N90 N 90 - s - 964670 499230 100 0 0 0 N120 N 120 964670 499260 100 0 0 0 N150 N 150 964670 499290 100 0 0 0 N180 N . 180 964670 499320 100 0 0 0 N210 N 210 964670 499350 100 o' 0 0 N240 N 240 964670 499380 100 0 0 0 N270 N 270 964670 499410 100 0 0 0 N300 N 300 964670 499440 100 0 0 0 NE30 NE 30 964691 499161 50 15 0 0 NE60 NE 60 964713 499182 0 30 30 0 NE90 NE 90 964734 499203 100 0 0 0 NE120 NE 120 964755 499225 100 0 0 0 NE150 NE 150 964776 499246 100 0 0 0 NE180 NE 180 964797 499267 100 0 0 0 96 Table VII. (Continued) Values used for fetch calculations for plot HE2021-10. DIST X- Y- SIMPLE POINT DIR (m) COORD COORD CANOPYRET VRFETCH FETCH FETCH NE210 NE 210 964819 499288 100 0 0 0 NE240 NE 240 964840 499309 100 0 0 0 NE270 NE 270 964861 499331 100 0 0 0 NE300 NE 300 964882 499352 100 0 0 0 NW30 NW 30 964691 499118 100 0 0 0 NW60 NW 60 964713 499097 100 0 0 0 NW90 NW 90 964734 499076 100 0 0 0 NW120 NW 120 964755 499055 100 0 0 0 NW150 NW 150 964776 499034 100 0 0 0 NW180 NW 180 964797 499012 0 30 30 0 NW210 NW 210 964819 498991 0 30 30 0 NW240 NW 240 964840 498970 0 30 30 0 NW270 NW 270 964861 498949 0 30 30 0 NW300 NW 300 964882 498928 0 30 30 0 S30 S 30 964670 499110 0 30 30 30 S60 S 60 964670 499080 0 30 30 30 S90 S 90 964670 499050 100 0 0 0 S120 S 120 964670 499020 100 0 0 0 S150 S 150 964670 498990 100 0 0 0 S180 S 180 964670 498960 100 0 0 0 S210 S 210 964670 498930 100 0 0 0 S240 S 240 964670 498900 100 0 0 0 S270 S 270 964670 498870 100 0 0 0 S300 S 300 964670 498840 100 0 0 0 SE30 SE 30 964649 499161 100 0 0 0 SE60 SE 60 964628 499182 100 0 0 0 SE90 SE 90 964607 499203 100 0 0 0 SE120 SE 120 964585 499225 100 0 0 0 SE150 SE 150 964564 499246 100 0 0 0 SE180 SE 180 964543 499267 100 0 0 0 SE210 SE 210 964522 499288 100 0 0 0 SE240 SE 240 964500 499309 100 0 0 0 SE270 SE 270 964479 499331 100 0 0 0 SE300 SE 300 964458 499352 100 0 0 0 SW30 SW 30 964649 499118 0 30 30 30 SW60 SW 60 964628 499097 0 30 30 30 SW90 SW 90 964607 499076 50 15 0 0 SW120 SW 120 964585 499055 100 0 0 0 SW150 SW 150 964564 499034 100 0 0 0 SW180 SW 180 964543 499012 100 0 0 0 SW210 SW 210 964522 498991 100 0 0 0 SW240 SW 240 964500 498970 100 0 0 0 SW270 SW 270 964479 498949 100 0 0 0 SW300 SW 300 964458 498928 100 0 0 0 W30 w 30 964640 499140 100 0 0 0 W60 w 60 964610 499140 100 0 0 0 W90 w 90 964580 499140 0 30 30 0 97 Table VTI. (Continued) Values used for fetch calculations for plot HE2021-10. DIST X- Y- SIMPLE POINT DIR (m) COORD COORD CANOPYRET VRFETCH FETCH FETCH W120 W 120 964550 499140 100 0 0 0 W150 W 150 964520 499140 100 0 0 0 W180 W 180 964490 499140 100 0 0 0 W210 W 210 964460 499140 0 30 30 0 W240 W 240 964430 499140 70 9 0 0 W270 W 270 964400 499140 70 9 0 0 W300 W 300 964370 499140 100 0 0 0 708 660 180 98 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0074983/manifest

Comment

Related Items