@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Science, Faculty of"@en, "Botany, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Bjorkman, Anne Donahey"@en ; dcterms:issued "2009-01-09T16:42:21Z"@en, "2008"@en ; vivo:relatedDegree "Master of Science - MSc"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description "Early land survey records can be used to reconstruct the historical distribution and abundance of tree species prior to the large-scale impact of industrialized societies. Comparing these records to current vegetation patterns enables an examination of the shifts that have occurred in plant communities since the arrival of European settlers in North America. I used presettlement (1859-1874) land survey records from southeastern Vancouver Island and Saltspring Island, British Columbia, Canada to reconstruct the relative abundance and density of tree species in these areas. I then collected equivalent vegetation data from the same points in the modern landscape, which enabled me to compare the two points in time and identify the changes in large-scale vegetation patterns that have occurred since European settlement. My results show a significant increase in the relative abundance of maple (Acer macrophyllum) and cedar (Thuja plicata), and a corresponding decrease in Douglas-fir (Pseudotsuga menzeisii). Furthermore, there has been a considerable increase in tree density in undeveloped areas. The 1859 records indicate that at least one third of the land surveyed was made up of prairies or open “plains,” while a combination of open woods and forests made up the remaining two thirds. Based on comparable density measures from 2007, prairies and plains now represent less than 5% of the undeveloped landscape, while forests comprise nearly 90%. These changes are likely due to a combination of factors that have been influenced by European settlement, most notably logging and fire suppression. The suppression of fire has led to an infilling of trees into previously open areas and has led to the rapid decline of the open prairie and savanna habitat types once common in this area. The results of this study can inform conservation efforts throughout the study area, particularly those involving the restoration of prairie or savanna habitats."@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/3421?expand=metadata"@en ; dcterms:extent "28248738 bytes"@en ; dc:format "application/pdf"@en ; skos:note " CHANGES IN THE LANDSCAPE AND VEGETATION OF SOUTHEASTERN VANCOUVER ISLAND AND SALTSPRING ISLAND, CANADA SINCE EUROPEAN SETTLEMENT by ANNE DONAHEY BJORKMAN B.A., Cornell University, 2004 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES (Botany) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2008 © Anne Donahey Bjorkman, 2008 ii ABSTRACT Early land survey records can be used to reconstruct the historical distribution and abundance of tree species prior to the large-scale impact of industrialized societies. Comparing these records to current vegetation patterns enables an examination of the shifts that have occurred in plant communities since the arrival of European settlers in North America. I used presettlement (1859-1874) land survey records from southeastern Vancouver Island and Saltspring Island, British Columbia, Canada to reconstruct the relative abundance and density of tree species in these areas. I then collected equivalent vegetation data from the same points in the modern landscape, which enabled me to compare the two points in time and identify the changes in large-scale vegetation patterns that have occurred since European settlement. My results show a significant increase in the relative abundance of maple (Acer macrophyllum) and cedar (Thuja plicata), and a corresponding decrease in Douglas-fir (Pseudotsuga menzeisii). Furthermore, there has been a considerable increase in tree density in undeveloped areas. The 1859 records indicate that at least one third of the land surveyed was made up of prairies or open “plains,” while a combination of open woods and forests made up the remaining two thirds. Based on comparable density measures from 2007, prairies and plains now represent less than 5% of the undeveloped landscape, while forests comprise nearly 90%. These changes are likely due to a combination of factors that have been influenced by European settlement, most notably logging and fire suppression. The suppression of fire has led to an infilling of trees into previously open areas and has led to the rapid decline of the open prairie and savanna habitat types once common in this area. The results of this study can inform conservation efforts throughout the study area, particularly those involving the restoration of prairie or savanna habitats. iii TABLE OF CONTENTS Abstract.................................................................................................................................... ii Table of Contents ................................................................................................................... iii List of Tables .......................................................................................................................... iv List of Figures...........................................................................................................................v Acknowledgements ................................................................................................................ vi Introduction..............................................................................................................................1 Land survey records.................................................................................................................3 Study Area ...............................................................................................................................5 Historical accounts of Vancouver Island .................................................................................8 Methods...................................................................................................................................12 Historical records ...................................................................................................................12 Addressing uncertainties in the historical data ......................................................................17 Field Methods ........................................................................................................................19 Data analysis ..........................................................................................................................21 Results .....................................................................................................................................27 Discussion ...............................................................................................................................36 Patterns of habitat distribution...............................................................................................37 Inferring process from pattern ...............................................................................................39 Implications for Conservation ..............................................................................................47 References...............................................................................................................................49 Appendix A.............................................................................................................................60 Appendix B .............................................................................................................................62 iv LIST OF TABLES Table 1: Species names in historic surveys with the current common and Latin names............19 Table 2: Number of historically surveyed and resurveyed points in each part of the study area...............................................................................................................................................28 Table 3: Chi-squared analysis of the change in tree species composition, past-to-present ........29 Table 4: Transition matrix showing the proportion of sites that changed from a low tree density to a higher density and vice versa ...................................................................................34 Table 5: Kolmogorov-Smirnov tests for differences in tree size distributions between 1874 and 2007 on Salt Spring Island ....................................................................................................35 v LIST OF FIGURES Figure 1: Map of Vancouver Island, the Gulf Islands, and the British Columbia coastline.....6 Figure 2: Portion of map of Cowichan District, created from the 1859 survey notes by Oliver Wells ........................................................................................................................................13 Figure 3: Portion of notes from Cowichan District and translation of the shorthand.............14 Figure 4: Three datasets in the Cowichan Valley, Chemainus district, and Saltspring Island, within the study area ................................................................................................................15 Figure 5: Proportions of tree species in each dataset, past and present ..................................29 Figure 6: Comparison of overall line tree and bearing tree data from the 1874 Saltspring Island surveys ......................................................................................................................... 30 Figure 7: Changes in density past to present, Cowichan Valley and Saltspring Island..........31 Figure 8: Relative proportion of each habitat type described by Oliver Wells in the Cowichan Valley, 1859............................................................................................................33 Figure 9: Tree density by habitat type in the Cowichan Valley, 1859 ...................................33 Figure 10: Tree size frequency histograms for Saltspring Island ...........................................36 vi ACKNOWLEDGEMENTS First and foremost, I thank my supervisor, Mark Vellend for his knowledge, wisdom, and encouragement. I thank my committee members, Sarah Gergel, Lori Daniels, and Andrew MacDougall, for their many helpful insights throughout the research process. Steen Magnussen provided assistance with the GamPoi density estimates. The other members of the Vellend Lab (Patrick Lilley, Emily Drummond, Hiroshi Tomimatsu, Will Cornwell, Heather Kharouba, Tanis Gieselman, and Jenny McCune) have provided countless hours of advice, support and amusement. I was exceedingly lucky to have the most wonderful field assistant possible, Jen Muir, whose cheerful companionship made even the thorniest thickets enjoyable. Simon Zappia and Irene Glenn also provided lively and helpful assistance in the field. I thank Brenda Guild and Don Gillespie for generously providing us with their beautiful cabin on Saltspring Island, and I thank the hundreds of property owners throughout the Cowichan Valley and Saltspring Island who allowed me to traipse through their backyards in search of property markers and bearing trees. Finally, I thank Hannes Dempewolf for providing love, support, and nutritionally balanced meals, Lu and Susie for their feline affection, and my friends and family for their motivation and encouragement. My parents, especially, have been unwavering in their support and have always inspired me to follow my dreams. 1 INTRODUCTION Understanding the processes underlying present-day ecological patterns in North America depends fundamentally on our ability to separate the past and present influences of natural variation in environmental conditions and of native peoples from the profound changes brought about by European settlers. Much of today’s ecological research is concerned with understanding how environmental change (e.g. habitat fragmentation, climate change) will affect ecosystem biodiversity and functioning in the future (McCarty 2001). Before we can successfully predict the future, however, we must first understand the processes that were important in creating and maintaining ecosystems both historically and in the present (Foster & Motzkin 1998). Ecological research often involves studies of “natural” landscapes and the role of “natural” processes in creating them. However, these studies often overlook the ubiquitous role of humans – both pre- and post-European settlement – in impacting the landscape (Foster & Motzkin 1998, Foster 2000a, Lunt & Spooner 2005). These impacts can potentially last for hundreds of years, and studies that do not take them into account often overlook an important influence on current ecological patterns (Denevan 1992, Gómez-Pompa & Kaus 1992, Foster & Motzkin 1998). Historical records provide important insights into past landscape conditions, the processes that maintained them, and reasons for the changes that have occurred into the present (Foster 2000a, Foster 2000b, Whitney & Decant 2005). Historical records come in many forms, ranging from ice and lake sediment cores to historical journals, newspapers, and photographs. These records have been used in North America and around the world to reconstruct the historical distribution of species, ecosystems, or biomes (e.g., Williams et al. 2001, Brown & Hebda 2002, Cogbill et al. 2 2002). The interpretation of these data can inform an understanding of the processes that led to the creation and dissolution of these historical systems. For example, paleoecological information about shifts in species ranges can be compared to historical disturbance regimes or climatic conditions in order to gain insight into the effect of each of these factors on species distributions (Davis 1988, Foster et al. 1990). Additionally, understanding landscape conditions in more recent history can be used to set baselines for conservation and restoration efforts that attempt to restore natural areas to a pre-industrial state (Hobbs & Cramer 2008). In North and South America, especially, historical records can be indispensible for understanding composition and abundance of plant and animal species, the frequency of natural disturbances, and the effect of native peoples on the landscape prior to the arrival of European settlers (Cronon 1983, Whitney & Decant 2005). Qualitative accounts of the North American presettlement landscape abound in the form of reports by early adventurers and the diaries of the first settlers to arrive in the New World. For example, early accounts of the New England landscape include general descriptions of the forest as well as information about the identity and location of particular species: “The timber of the country grows straight and tall, some trees being twenty, some thirty foot high, before they spread forth their branches.” (Wood 1634) “As for Trees the Country…Vines, Cedars, Okes, Ashes, Beeches, Birch trees, Cherie trees bearing fruit whereof wee did eate, Hasels, Wichhasels… walnut-trees, Maples…and a kinde of tree bearing a fruit like a small red Peare-plum” (Pring in Cronon 1983). Although these written records are valuable in providing a broad description of the presettlement landscape, they do not allow quantification of temporal landscape changes. 3 Quantitative records, such as lake sediment cores, early climate records or land survey data, are often more challenging to obtain than qualitative records, but are invaluable in their ability to provide a source of historical data that can be compared directly to the modern landscape to identify the direction and extent of change that has occurred. Often, a combination of quantitative and qualitative records from a variety of sources and time periods can be used together to present a compelling history of a region across time (Schulte & Mladenoff 2001, Egan & Howell 2005). Land survey records One valuable form of historical data in North America originates from early land survey records. Studies of historic land surveys in eastern and midwestern North America have provided invaluable information about vegetation patterns across the landscape prior to European settlement (Manies & Mladenoff 2000, Cogbill et al. 2002, Whitney & Decant 2005). In most cases, these surveys were commissioned by the government to document the quality and potential uses of the land, and to divide these areas into lots for future settlement (Bourdo 1956, North et al. 1977, Almendinger 1996). The survey records include information about the dominant vegetation and habitat types within each parcel of land. In particular, the surveyor recorded the species, diameter, and distance from a survey point to one or more “bearing” or “witness” trees at regular intervals along the survey lines (Moore 1871, Bourdo 1956). These trees were then marked and used as semi-permanent indicators to delineate the boundary of that survey section (Bourdo 1956, Whitney 1994, Schulte & Mladenoff 2001). Some surveyors also recorded information about disturbances, such as the location and extent of burned areas or fallen trees, and the general appearance of certain habitat types (Whitney & Decant 2005). 4 The wealth of data in these early survey records can provide information about a wide range of historical conditions and landscape patterns. Records of bearing tree species allow for large-scale quantitative reconstructions of tree species’ relative abundances in presettlement forests (e.g., Abrams & Ruffner 1995, Brown 1998, Radeloff et al. 1999, Cogbill et al 2002). This, in turn, can be used to look at shifts over time in the location of various forest types (Cogbill et al. 2002). The recorded distance from each survey corner to the nearest bearing tree allows an estimation of tree density, though contemporary studies rarely collect field data using methods that enable directly comparable density calculations (Zhang et al. 2000). The inclusion of dominant understory species and general habitat descriptors such as soil type provide a record of the location and extent of less common habitat types, such as wetlands and prairies (Whitney & Decant 2005). For records that contain two or more bearing trees at each point, an analysis of historic species associations is possible (i.e. which species tend to occur together) and also historical species-environment associations (Batek et al. 1999, Whitney & Decant 2005). Finally, some historic records contain information about the location of indigenous villages and the impact of indigenous peoples on the surrounding landscape, thereby providing a unique record of the role of anthropogenic disturbance in maintaining historic ecosystems (Foster II et al. 2004). By recording information about the appearance of the landscape before European influence, these early land surveys and the information they contain provide a critical baseline for understanding subsequent changes in vegetation and underlying factors affecting the modern landscape. Although historical land surveys exist for the Pacific Northwest, these records have rarely been exploited to aid in reconstructing the pre-settlement landscape in this area. 5 My research uses land surveys conducted prior to or just after European settlement (1859-1874) from the Cowichan Valley of southeastern Vancouver Island and from Saltspring Island to reconstruct the vegetation patterns and landscape characteristics that dominated in these areas prior to the arrival of Europeans. I then compare these historical data to field data that I collected from the same locations in order to identify what changes have occurred since European settlement. These comparisons, combined with published historical and experimental studies, allow me to make inferences about what processes were likely important in determining patterns in the presettlement landscape, and how these processes may have changed over time. The study area: Vancouver Island and Saltspring Island Vancouver Island and the Gulf Islands lie off the southwestern coast of British Columbia, Canada (Figure 1). Although the southwestern coast is largely dominated by the Coastal Western Hemlock biogeoclimactic (BEC) zone (Egan 1999a), the southeastern coast of Vancouver Island and the Gulf Islands (including Saltspring Island) are located in the much smaller Coastal Douglas-fir (CDF) zone (Egan 1999b). The CDF zone is characterized by warmer temperatures (9.2 to 10.5 ºC mean annual temperature) and a drier climate (647 to 1263 mm mean annual precipitation; Meidinger & Pojar 1991) than the rest of Vancouver Island due to a rainshadow effect produced by the Olympic and Vancouver Island mountains (Egan 1999b, Meidinger & Pojar 1991), with less than half the annual precipitation of Coastal Western Hemlock sites just 100 km away (Meidinger & Pojar 1991). Two major ecosystem types dominate the CDF zone. The first of these, the dry Douglas-fir forests that give the CDF zone its name, is by far the largest. These forests are dominated by Douglas-fir (Pseudotsuga menziesii), with minor components of western red- 6 cedar (Thuja plicata), grand fir (Abies grandis), red alder (Alnus rubra) and big-leaf maple (Acer macrophyllum). The understory of these forests is often dominated by salal (Gaultheria shallon), Oregon-grape (Mahonia spp.) and sword fern (Polystichum munitum). In the driest forests along the eastern coast, Pacific madrone (Arbutus menziesii) grows interspersed with Douglas-fir and grasses are abundant in the understory (Egan 1999b, Flynn 1999). Due to its accessibility, the coastal Douglas-fir ecosystem has been heavily impacted by logging over the past 150 years. The Ministry of Environment, Land, and Parks estimates that only 0.5% of the original 220,000 hectares of coastal Douglas-fir forest remains as relatively undisturbed old forest (Flynn 1999). Figure 1: Map of Vancouver Island, the Gulf Islands, and the British Columbia coastline. 7 The second distinct ecosystem type that occurs within the CDF zone is sometimes known as the “Garry oak ecosystem” (Fuchs 2001). These areas are usually characterized by an open grassland with scattered Garry oak (Quercus garryana) trees. Many native wildflowers, especially camas (Camassia leichtlinii and C. quamash), shooting star (Dodecatheon spp.), and satinflower (Olsynium douglasii) are abundant among the grasses (Erickson 1996, Erickson & Meidinger 2007). Modern oak savanna habitats usually occur on very shallow, rocky soils, which prevent Douglas-fir establishment due to soil moisture limitation and shallow rooting depths (MacDougall et al. 2004, MacDougall & Turkington 2005). However, a few remaining patches of oak savanna habitat occur in areas with deeper, richer soil, where several previous studies have suggested that frequent fires were historically responsible for keeping these areas in their open state (Tveten 1997, Turner 1999, MacDougall et al. 2004, Pellatt et al. 2007). Garry oak savanna was previously much more abundant than today; by some estimates, more than 90% of the historical habitat has been converted to agriculture or destroyed (Lea 2006, Vellend et al. 2008). Due to environmentally biased fragmentation of this habitat, current oak savanna is now more likely to be found in steeper areas and at higher elevations with more rainfall than the historical average (Vellend et al. 2008). As a result of this extensive fragmentation, the Garry oak ecosystem is one of the most endangered ecosystems in Canada and is of high conservation concern (Egan 1999b, Fuchs 2001). There are over 100 nationally and provincially listed species-at-risk associated with Garry oak ecosystems, with at least 12 of these threatened at a global scale, while several others have already been extirpated (Fuchs 2001). Many active restoration efforts are underway to preserve and restore the few remaining patches of oak savanna habitat (GOERT 2008, 8 GOMPS 2008, GORP 2008), but these efforts are hampered by several factors, including some uncertainty about the former distribution, composition, and abundance of the GOE, and the historical processes important in maintaining this ecosystem. Historical accounts of Vancouver Island Victoria, the first European settlement on Vancouver Island, was initially surveyed by the Hudson’s Bay Company in 1839 and officially founded in 1849. Settlers quickly spread up the coast, reaching the Cowichan Valley and many of the Gulf Islands by the early 1860’s (Lillard 1986). However, European influence began long before settlement began, with the discovery of the island in 1770 by Captain James Cook and the rapid spread of small-pox to many First Nations communities (Arnett 1999). Prior to European contact, the population of First Nations people on Vancouver Island and surrounding areas likely numbered in the tens of thousands, but by the 1900’s this number had declined by nearly 90% (Harris 1994). With a relatively large historical native population, it is likely that First Nations people had a profound impact on the pre-settlement landscape (Boyd 1999). Likewise, the dramatic decline in the native population and cessation of many cultural practices indicates that this impact has been all but eliminated in the present day. Qualitative evidence indicates that First Nations peoples influenced the Vancouver Island landscape in various ways. Historic records in the form of letters and journals describe vast forested areas interspersed with large patches of open oak savanna or prairie, and suggest that fire, set by First Nations people, may have played an important role in maintaining these open areas (MacDougall et al. 2004). In an 1849 article in the Times newspaper entitled “Colonization of Vancouver Island,” the anonymous author wrote of the land around Victoria: 9 “Miles of the ground were burnt and smoky, and miles were still burning. The Indians burn the country in order to [promote]…the roots which they eat. The fire runs along at a great pace, and it is the custom here if you are caught to gallop right through it; the grass being short, the flame is very little; and you are through in a second…” Considerable evidence suggests that periodic burning of open areas attracted deer and improved the growth of several important plant species, including a variety of fruit-bearing bushes and edible bulbs such as camas (Camassia leichtlinii and C. quamash) and wild onions (Allium spp.; Turner 1999, Beckwith 2004). This evidence is corroborated by the oral tradition of many First Nations communities (Marshall 1999, Turner 1999). Several studies have attempted to use historical data to determine the frequency of fires in these areas (Gedalof et al. 2006, McCoy 2006, Pellatt et al. 2007). Estimates of fire frequency at Rocky Point, on the southern tip of Vancouver Island, were attempted by analyzing tree-rings and residual fire scars. Although the results did reveal significant evidence of encroachment of Douglas-fir (Pseudotsuga menziesii) trees into the previously open prairie, no conclusive evidence about the historical frequency of fire was found (Gedalof et al. 2006). Charcoal analyses conducted from lake sediment cores at two locations on Vancouver Island and one on North Pender Island revealed a mean fire return interval of 27 to 41 years (McCoy 2006, Pellatt et al. 2007). However, these analyses are unlikely to pick up the frequency of small- scale and low-intensity fires such as would have been used by First Nations people to maintain areas of open prairie (McCoy 2006), in part because small grass fires are often less intense than forest fires, and therefore may be less likely to leave the lacustrine charcoal deposits used for determining fire return intervals (Duffin et al. 2008). As a result, charcoal analyses within the study area do not present a complete picture of the historical fire regime (McCoy 2006, Duffin et al. 2008). 10 Although these studies have been important in suggesting the importance of First Nations activity in influencing historic vegetation patterns, their scope is limited by the availability of historical data. Qualitative accounts, like those mentioned above, describe the landscape of Vancouver Island prior to extensive European disturbance and are some of the most widely available data, but these accounts are not amenable to quantification. Charcoal and pollen analyses from lake sediment cores can provide information on fire and vegetation over a long time scale (McCoy 2006), but are often limited by the location of lakes and the size and intensity of fires (Duffin et al. 2008). One form of historical datum that has been largely overlooked in this region is pre- settlement land survey records. As mentioned previously, the original purpose of these surveys was to identify areas of valuable land – for agriculture, forestry, or mineral extraction – and to delineate property boundaries for settlement. At the same time, these records provide a valuable source of data about the pre-settlement landscape, including the abundance of certain tree species, the location and extent of certain habitat types, and the tree density and tree size structure of wooded areas. Furthermore, many of the property boundaries delineated by surveyors in the 1800’s are still in place today and are easily relocated, thus allowing equivalent contemporary data to be collected. The goal of my research was to use historical survey records to gain insights into the appearance of the pre- settlement landscape and, through comparisons with current landscape conditions, to better understand some of the important ecological processes at play in these areas. In particular, these data allow me to test several predictions based on the commonly presented hypothesis of a frequent anthropogenic historical fire regime by comparing historical vegetation patterns of Vancouver Island to the present day landscape and to other historical, theoretical, and 11 experimental studies of the effects of fire on forest structure and composition. The objectives of my research were as follows: (1) To quantitatively characterize the vegetation of this region prior to European settlement in terms of: a) relative tree species abundances, b) tree size distributions, c) forest density, and d) the relative abundance and location of habitat types (prairies, forests, etc). (2) To identify the extent and direction of change since European settlement to the present day in each of the variables mentioned above. (3) To use the results of my research, combined with previous historical studies and experimental data, to identify some potentially important ecological processes that maintained vegetation structure in the presettlement landscape. The central hypothesis is that low-intensity fires occurred at high frequency in the presettlement landscape, with important consequences for the vegetation. This hypothesis makes three specific predictions about historical vegetation patterns as they compare to the current landscape. Firstly, forest composition should show a shift over time, from fire- sensitive species being relatively rare in the past and increasing in abundance to the present. An increase in early-successional, disturbance-associated species in the present is also likely due to the increase in logging and agriculture-related disturbances. Secondly, frequent fires should result in lower tree density and a relatively high proportion of open habitat types (prairie, plains, savanna) in the pre-settlement period compared to the present. In the current landscape, fire suppression would therefore lead to an increase in forest density and a decrease in the proportion open habitat types. This increase in tree density should result from an increase in seedling recruitment and survival due to fire suppression, with young, fire-sensitive saplings filling in the gaps between the 12 mature trees more resistant to burning (Agee 1998). Finally, frequent fires should lead to a unimodal tree size (diameter) distribution in the presettlement landscape, with few small trees and many medium to large trees. This prediction is based on experimental data depicting the expected tree size frequency distributions in a forest prone to frequent, low-intensity fires (Peterson & Reich 2001), which produces a unimodal distribution of tree sizes. Alternately, a monotonically decreasing size distribution, often referred to as a negative exponential or “inverse J-shaped” curve, is the expected tree size distribution in mature forests not prone to a low-intensity fire regime (Leak 1965, Lorimer 1980, Rouvinen & Kuuluvainen 2005, Westphal et al. 2006). I therefore expect the 2007 data to reveal a loss of the unimodal distribution formerly maintained by frequent burning and instead exhibit the typical inverse J-shaped curve. By comparing presettlement and current vegetation patterns, I can test these predictions about the important processes at play in the historic landscape. Although historical land survey data have been used in several studies of presettlement vegetation in North America, this study represents the first time, to my knowledge, that the original survey points have been resurveyed to allow a direct comparison of changes in species composition, tree density, and forest structure. METHODS Historical records I used historical land survey records completed between 1859 and 1874 to reconstruct the presettlement landscape of southeastern Vancouver Island and Saltspring Island. These original records are available in the BC Land Title & Survey archives in Victoria, British Columbia. The survey maps (e.g., Figure 2) and corresponding field notes (e.g., Figure 3) 13 contain information about vegetation and landscape features in these areas as they appeared before European settlement. The maps were lithographed in 1860 at the Topographical Depot of the War Office under the direction of Major A.C. Cooke, R.E., and Col. H. James, Director. Copies are available in the BC Land Title & Survey office in Victoria, the University of British Columbia Rare Books and Special Collections library in Vancouver, and at the National Archives in Surrey, England. The total area of historically surveyed land included in this study was over 266 square kilometers, or 26,600 hectares. Figure 2: Portion of map of Cowichan District, created from the 1859 survey notes by Oliver Wells; horizontal and vertical lines represent range and section survey lines. Maps are available in the archives of the BC Land Title & Survey office, Victoria, BC, Canada. The scale bar and north arrow have been added to the original map. 14 Figure 3: Portion of notes from Cowichan District (left) and translation of the shorthand with the original text bold and underlined (right). Notes are available in the archives of the BC Land Title & Survey office, Victoria, BC, Canada. Historical surveys were divided into three separate datasets, based on the surveyor, location, and the year in which the survey was conducted. The first dataset, which includes the districts of Shawnigan, Cowichan, Quamichan, Somenos, and Seymour (176 km 2 in total), is located on the southeastern coast of Vancouver Island, near Duncan, British Columbia (Figure 4). This area will henceforth be referred to as the “Cowichan Valley” dataset. These districts were surveyed in 1859 by Oliver Wells. One additional district on Vancouver Island, the Chemainus district (26 km 2 ), was surveyed in 1864 by William Ralph, and will be referred to as the “Chemainus” dataset. The final set of data includes parts of Saltspring Island (64 km 2 ), which is just off the southeastern coast of Vancouver Island and practically due east of the Cowichan Valley. These areas on Saltspring were surveyed in 1874 by Ashdown Green. This survey took place after some limited European settlement had already occurred (settlement began in 1859). Therefore, these data were reviewed and 15 any areas indicating European influence at the time of the survey (roads, fields, houses) were removed from the dataset. Figure 4: Three datasets in the Cowichan Valley, Chemainus district, and Saltspring Island, within the study area. Presettlement surveys were conducted along a series of east-west and north-south lines, referred to as “section” and “range” lines, respectively. Each of the resulting rectangles (the area between two “range” and two “section” lines) was approximately 1 km 16 by 400 meters in size (see Fig. 2). At the intersection of every section and range line, the surveyor planted a post as an indication of the exact location of the property corner. At every post, the nearest tree was marked as a “witness” or “bearing” tree, so that the property corner could be found should the original post be lost. The species of tree and the distance from the post to the tree was recorded in the notes. In the Chemainus and Saltspring Island surveys (but not in the Cowichan Valley) the diameter of the bearing tree was also recorded. We used the survey data to estimate species composition (bearing tree species frequencies), forest density (based on the distance from a post to the nearest tree), and forest size structure (bearing tree diameter distribution) in the presettlement landscape. In addition to these specific bearing tree data, surveyors also recorded descriptions of habitat types along the survey lines and at each point, such as “prairie,” “open pine and oak plains” or “thickly wooded forests”. Although general habitat characteristics were described in the historical survey notes of every dataset, only the Cowichan Valley surveyor described the habitat surrounding each point using consistent terminology. Most of these descriptions were repeated frequently throughout the surveys, and analyzing these habitat descriptions can allow estimation of the proportion of different habitat types across the landscape. I identified six reoccurring terms for this analysis of the Cowichan Valley data: “forest” (n=20 points), “open woods” (n=32), “pine and oak plains” (n=53), “bottom land” (n=13), “swamp” (n=10), and “prairie” (n=27). These descriptions were used by the surveyor to describe 155 out of the 225 total points in the dataset (about 70%). The remaining 30% of points either had no description or were described using terms such as “rocky, worthless land” (9 locations), “tolerably good land” (8 locations), or “large pine timber” (15 locations), that do not allow straightforward, consistent classification. 17 The Saltspring Island surveys include one additional type of historical datum that the other datasets do not: line trees. Every time a section or range survey line directly intersected a tree, that tree was marked as a “line” tree, and the species and diameter of that tree recorded in the notes. There are a total of 645 line trees in the Saltspring Island data along 126 km of survey lines. Line tree data have been used in other studies of historic survey data as a check for bias with respect to tree species and size in the bearing tree records, as it presumably represents an unbiased record of forest composition along those lines (Janke et al. 1978, Zhang et al. 2000, Whitney & Decant 2005). I will address the specific uses of line tree data in checking for bearing tree bias following the description of my analyses. Addressing uncertainties in the historical data As with many historical studies, some uncertainty comes with the translation of historic species nomenclature to that of the present (Table 1). Although this process was made somewhat easier by the relatively low diversity of major tree species in this area, some grouping of species was necessary. In particular, the 1859 Cowichan Valley surveys did not distinguish between hemlock and Douglas-fir, and refer to both species as “pine.” This was determined because hemlock was never recorded in the historical survey data, despite being present in other regions historically and in this region currently. Furthermore, a later 1864 survey at some of the same points re-identified one of the original “pine” bearing trees as a hemlock. Therefore, hemlock and Douglas-fir were grouped in the 1859 and 2007 Cowichan Valley datasets. As Douglas-fir dominates both past and present forests, it is probable that most trees identified as “pine” were in fact Douglas-fir. Only two other pine species, western white pine (Pinus monticola) and shore pine (Pinus contorta), are present on Vancouver Island. These are unlikely to have occurred in the 1859 survey as western white pine is rare 18 along the coast and neither species occurred in any of the 2007 surveys at the same survey locations. The Chemainus and Saltspring Island surveyors did distinguish between hemlock and Douglas-fir species, so no grouping was necessary in these datasets. Another concern is the absence in the historic data of certain species that do occur in the modern landscape. For example, cherry (Prunus spp.) occurs occasionally in wet or disturbed areas in the present, but was not recorded in the past. It is unclear whether this was due to a low abundance of cherry in the past, or whether it represents surveyor bias against choosing cherry as a bearing tree, possibly because it is small and short-lived relative to other species. As a result of this, all analyses were done both with and without cherry, and in all cases the inclusion or exclusion of cherry made no difference in the statistical outcome of the analysis. As a result, only the analyses without cherry are reported here. Cottonwood (Populus balsamifera) also did not occur in the historic dataset but was included in the analysis because it is unlikely that its absence represents surveyor bias, as it is as large as other species in the historical data and typically lives 100-200 years (Steinberg 2001). Cottonwood is relatively uncommon in the current landscape and strictly limited to riparian areas – a small proportion of the total study area – and was therefore likely to have been missed due to its rarity. Further details of species nomenclature are provided in Appendix A. 19 Table 1: Species names in historic surveys with the current common and Latin names. Years refer to the survey dataset (1859 = Cowichan Valley, 1864 = Chemainus, 1874 = Saltspring Island). See Appendix A for explanation of naming. Historical names without an accompanying year apply to all three datasets. Species names follow The Illustrated Flora of BC (Douglas et al. 1998-2000). Field methods In order to characterize changes in vegetation patterns across time, I compared the historical patterns revealed in the land survey data with current field data. During the summer of 2007, I collected vegetation data at historical survey points using methods similar to those employed in the historical surveys. A nearly perfect past-to-present point match with respect to spatial position was possible because the historical survey corners still represent property boundary lines to the present day. I verified the location of these points 20 by georeferencing the historical maps in ArcGIS 9.2 (ESRI, Redlands, California, USA), and found that the corner locations matched a 2006 map of property parcels provided by the local district governments (Cowichan Valley Regional District, North Cowichan District, and Capital Regional District). We could then easily obtain the geographic coordinates of these corners, and used a GPS unit to identify the location of the points in the field (+/- 15m). In many cases, the exact location was immediately obvious due to the presence of property markers, flagging tape, and occasionally by recently marked bearing trees. At each survey point, the species of, distance to, and diameter at breast height (1.3m from ground, DBH) of each of the ten trees closest to the stake was recorded. The farthest recorded distance from a post to a bearing tree in the historical data was 26 meters (sites with no tree close by were marked as “no bearing tree near”). As a result, I used 26 meters as the maximum distance in the 2007 data as well. I recorded the distance to the 10 closest trees, or all trees within a 26 meter radius if there were fewer than 10 trees. Only trees with DBH > 10 cm were recorded, as this corresponds to the smallest bearing trees recorded in the historic data and is slightly larger than the minimum size of 6.35 cm designated in the manual of instructions to surveyors (Moore 1871). Distance was measured with a Leica DISTO A5 laser distance meter (St. Gallen, Switzerland) to an accuracy of +/- 5 cm, except in areas of high shrub density, where I used a meter tape. All trees were identified to species except for willow (Salix spp.), which was identified to genus. In some cases, the survey point was on developed land, including recent clear-cuts, agricultural or recreational fields, houses, or parking lots; one survey line now runs directly through the city of Duncan. None of these points had any natural habitat left to survey, and were excluded from the study. Included in my surveys were any points with a contiguous 21 area of forest or natural prairie (i.e. not an agricultural field or artificial lawn) large enough to survey at least ten trees (or 26 meters) in any direction before encountering a developed area. Areas that had been selectively logged but had some trees remaining (uncommon) were surveyed, but clear-cut sites (more common) were excluded. If the area immediately around the survey point was heavily disturbed, I traveled a maximum of 50 meters to the closest patch of contiguous forest or prairie. If there were no such areas within 50 meters, a general description of the landscape was recorded but no formal survey was completed. In some cases, the point occurred exactly at the edge of a natural habitat (e.g. the point was located at the boundary between an agricultural field and a forest). In this case, only the ten closest trees within the forested area were measured, and the proportion of the area around each point (from 0-360 o ) that was natural habitat was recorded. In total, approximately 40% of historically surveyed points were not surveyable due to development or logging of the natural habitat. Of the surveyed sites, approximately 70% were resurveyed in their original location, while the other 30% of surveys were conducted a maximum of 50 meters away from the original point. Data analysis The data were analyzed to test for changes from the historic period to the present day in species composition, forest size structure, and forest density. In addition, in the Cowichan Valley only, habitat descriptions were used to look at changes in the representation of particular habitat types. Species Abundance. Changes in species frequencies were determined using a chi- squared analysis (SAS 9.1, Cary, North Carolina) of past and present frequencies of each species across all points in each dataset. Rare deciduous trees (maple, apple, willow, oak, 22 and cottonwood) are reported individually but were grouped for the chi-square analysis, as the chi-squared test is more likely to produce a type I error if categories contain fewer than 5 observations (Sokal & Rohlf 1995). An exact test can sometimes be used in situations of very small sample sizes, but this was not an option for my data because of the large number of observations in some categories (notably Douglas-fir). In addition, in the Cowichan Valley data, grand fir (N=1) was grouped with Douglas fir and hemlock for the chi-squared analysis due to the rarity of grand fir in the historical dataset. For all of these analyses I used only those points where data were collected in both time periods. Forest density. Forest density was calculated in both the Cowichan Valley and Saltspring Island datasets. A density estimate for Chemainus was not calculated due to the small number of surveyed points in this district and the large margin of error that results. Density was calculated using density estimators based on the distance from each point to the centre of the n-th closest tree (Morisita 1957, Eberhardt 1967). Many density estimators have been proposed for this type of datum (e.g., Cottam & Curtis 1956, Persson 1964, Pollard 1971, Byth & Ripley 1980). However, most of these estimators make unrealistic assumptions about the spatial distribution of organisms in the landscape, potentially producing highly biased density estimates (Engeman et al. 1994, Bouldin 2008). The Byth and Ripley (1980) procedure, for example, assumes that all organisms are randomly distributed in space (i.e. density is constant across space). This assumption is clearly violated in my study area due to the presence of habitat types ranging from densely wooded forests to open prairies. Other equations, such as those developed by Persson (1964), assume an evenly spaced distribution of organisms – again not likely unless the site is a plantation (Lynch & Rusydi 1999). For this study, I used two density estimators that allow for variable 23 tree density across the landscape. The first of these, known as the GamPoi method (Magnussen et al. 2008), allows for variation in density among points across the study area according to a (very flexible) gamma distribution, and assumes a random (Poisson) distribution of trees within areas of different density. In other words, this method explicitly accounts for variable densities, and it performs well on simulated datasets (Magnussen et al. 2008). GamPoi was the primary density estimation method used for both presettlement and present-day data, in part because, unlike many other estimators, it can calculate densities with n=1 (only one tree at each point, as in the historic data). However, the GamPoi estimator cannot account for datasets with varying values of n across sample points, such as the 2007 data where n is sometimes <10, or where the surveyed area is <360 o around each point (i.e. those sites that occurred at habitat edges). As a result, the GamPoi estimator was used only for a subset of present-day sites that met both of these criteria, as well as for the matching set of historical points. I used a second method of estimating forest density that allowed use of the complete set of present-day sites. This method could not be used to calculate presettlement density estimates because it requires n > 1, and was therefore only useful in estimating 2007 density measures. Because the same density estimation methods could not be used on the entire set of both presettlement and present-day sites, a present-day subset of suitable sites was also analyzed using both methods to assess potential differences between methods. Density ( ) was estimated using a variation of an equation first introduced by Morisita in 1957 and later adapted by Keuls in 1963, 24 with a variance adapted from Cochran (1977), where t is the number of points, n is for the n-th closest tree to the point, A is the area within which trees were recorded around that point (for 360° sites, this is equivalent to !r 2 ), and the summation is across points. The (n-1)/n correction factor was shown to be unbiased for clumped spatial distributions by Eberhardt in 1967 (the n in the denominator is implicit in the equation shown above). This method allows density to vary from point to point, assuming a negative binomial (clumped) distribution of trees. Bouldin (2008) found that this equation was more accurate in estimating forest density than the point-centered quarter method (Cottam & Curtis 1956) formerly used to estimate density from land survey data. It should also be noted that A is calculated based on the distance from the point to the n-th tree, which is assumed to be the distance to the center of the tree rather than the edge. Historical distances were measured to the edge of the tree. Because the presettlement Cowichan Valley dataset did not include tree diameters to allow calculation of distance to tree center, I used a sampling method to assign a random diameter value to each tree in the Cowichan Valley dataset based on the presettlement distribution of tree sizes on Saltspring Island, where diameter measurements were recorded. This random assignment of DBH values was repeated ten times, and the final density and standard error estimates for the Cowichan Valley were based on the average of these ten measures. Using distances to the tree edge in all datasets produced slightly higher density estimates for all datasets, as expected, but the qualitative differences among datasets were identical. 25 Tree density in different habitat types. Tree density was also calculated for the historical Cowichan Valley data within each described habitat type using the GamPoi estimator. Only four habitat types (“prairie”, “plains”, “open woods” and “forest”) had a large enough sample size to estimate densities (n=20 or greater). This was done both to verify the surveyor’s descriptions (e.g. prairies should be less dense than forests) and to estimate the degree of openness in each habitat type. Because we do not have equivalent habitat descriptions for the 2007 data, we cannot conduct an exactly comparable analysis for the 2007 data. To provide an approximate assessment of the proportional representation of comparable habitat types in the present day, we calculated the number of 2007 sites within each 1874 density category. Density was calculated at each point by dividing the number of trees (generally, n=10) by the area within which trees were recorded (!r 2 at a 360 o point, where r is the distance to the n-th tree). Density categories were based on the approximate midpoint between the density estimates of two consecutive presettlement habitat types. Size structure. Pre-European forest size structure was estimated from the diameter measurements of bearing trees on Saltspring Island. Because historical diameters were only recorded in two-inch intervals, I rounded all measurements to the midpoint of the nearest 10 cm category (10-19 cm becomes 15 cm, 20-29 cm becomes 25 cm, etc.). Differences in the size distributions in 1874 and 2007 were then tested using a Kolmogorov-Smirnov test. Two comparisons were made between past and present sites: 1) 1874 bearing trees vs. 2007 bearing trees (the first tree only at each site), and 2) 1874 line trees vs. all 2007 trees (the ten closest trees at each site). In the second analysis comparing 1874 line trees and all ten 2007 trees at each site, I corrected for the varying plot sizes within which trees were observed in the 2007 surveys by weighting the diameter measurements by plot area. This was done to 26 ensure that the line tree data (collected along a line rather than in a plot) were directly comparable to the bearing tree data by removing the effect of plot size. Because the distance to the 10 th tree is likely to be much longer in a plot with ten large trees than ten small trees, large trees may be overrepresented in the 2007 dataset if no corrections are made. The same bias may apply to the single bearing tree datasets, but weighting by the distance to a single tree introduces large amounts of variance, and the datasets are exactly comparable across time periods, so no corrections were made. In addition, I compared the 1874 bearing tree data with the 1874 line tree data, to check for any size-bias in the bearing tree data. Finally, I compared the 2007 bearing tree dataset (using only the first tree at each site) to all 2007 trees (the closest ten trees at each site) to ensure that these two datasets were comparable. Assessment of potential biases. One of the strongest criticisms of using historic bearing tree data is the potential for bias, including a surveyor’s preference for larger, longer- lived, or smooth-barked trees, which could affect estimates of species composition, size structure, and density in the past (Almendinger 1996, He et al. 2000, Whitney & Decant 2005). Many previous studies using bearing tree data have found very little, if any, evidence of bias in terms of species composition, but the degree of bias can vary from survey to survey (Maines & Mladenoff 2000, Manies et al. 2001, Schulte & Mladenoff 2001, Whitney & Decant 2005). Manies et al. (2001) found some evidence for bias in tree species composition and size based on the preferences of individual surveyors, but the biases were not consistent across surveyors. Therefore, I conducted several analyses to test for and correct any evidence of bias in the historical survey data. The line tree data included in the 1874 Saltspring Island records provide a source of data that are highly unlikely to contain any biases with respect to tree species or size, and 27 which can therefore be used to check for bias in the bearing tree data. Line trees were marked and recorded whenever the survey line directly intersected a tree, and were not chosen by the surveyor with any specific qualification in mind (Moore 1871, Almendinger 1996, Zhang 2000). I compared species frequencies in the line tree data and the bearing tree data in order to detect any potential bias in species composition in the latter. I should note here that due to the different methods of collecting these two sources of data, a perfect match is not expected. For example, tree species that tend to occur in very open landscapes, such as Garry oak trees, would be overrepresented in the bearing tree dataset because a surveyor would have to travel farther away from the post to mark that tree, but in a line transect those trees would not likely be hit. Nonetheless we expect this comparison to point to any major biases in favor of, or against, particular tree species in the historical data. Line tree data can also allow an assessment of bias in bearing tree diameter measurements, for example if surveyors preferentially chose large trees. Because line tree data also contain diameter measures, a simple process of weighting the line trees by their inverse diameter (the probability of a line hitting a tree is proportional to its diameter) provides an unbiased estimate of the distribution of line tree diameters. Line tree data thus allow a direct analysis of the degree of bias in bearing tree species composition and size in the historical survey data. RESULTS During the summer of 2007, I resurveyed a total of 259 out of 428 historic survey points (Table 2). Nearly all sites that occurred in areas of natural habitat were surveyed (only three sites were excluded due to lack of permission, and 15 sites were excluded due to time 28 constraints). The rest of the sites that were not resurveyed represent developed or cleared sites that had no natural habitat remaining. Table 2: Number of historically surveyed and resurveyed points in each part of the study area. Species abundance. Significant changes in species frequencies were observed in all three datasets (Table 3; Fig. 5). There was no significant difference in species frequencies between line trees and bearing trees on Salt Spring Island (Table 3; Fig. 6), which suggests little to no bias in bearing tree species selection by surveyors in the past. The main trends visible across all datasets are a large increase in the proportion of maple and cedar and a decrease in Douglas-fir. There were some further minor changes in species composition within each dataset, such as an increase in grand fir in the Cowichan Valley dataset and an increase in alder in the Chemainus dataset, but these were not uniform across all datasets. A more detailed picture of compositional changes across time can be depicted in a transition matrix, showing the number of points where the closest tree has remained the same or changed to particular other species (Appendix B), although the preponderance of very low frequencies in this table precludes statistical analysis. 29 Table 3: Chi-squared analysis of the change in tree species composition, past-to-present a) b) 30 Figure 5: Proportions of tree species in each dataset, past and present: a) Cowichan Valley b) Chemainus district c) Saltspring Island. “Pine” in the Cowichan Valley dataset refers to Douglas-fir and hemlock combined; the category “other” includes apple, cottonwood, arbutus, yew, and willow trees, which occurred at very low frequencies. Figure 6: Comparison of overall line tree and bearing tree data from the 1874 Saltspring Island surveys. Approximately equal proportions of each species in each data type indicate little to no surveyor bias in bearing tree selection. Forest Density. All tree density estimates showed an increase in forest density from the presettlement time period to the present (Figure 7). In the Cowichan Valley, average density increased from 335.6 stems/ha (+/- 125.3) in 1859 to 829.8 stems/ha (+/- 171.4) in 2007. On Saltspring Island, average density increased from 258.6 stems/ha (+/- 196.2) in c) 31 1874 to 729.5 stems/ha (+/- 169.0) in 2007. Although the 95% confidence intervals are fairly large they do not overlap in any case, and I therefore conclude that there are significant differences in mean density between past and present in both datasets. The data subsets that were analyzed using both the GamPoi and the Morisita/Keuls density estimation methods show almost identical means and confidence intervals, which suggests that the density results are not affected by the use of two different estimation methods. The data subsets show average densities similar to those from the entire dataset from which they were taken. Figure 7: Changes in density past to present, Cowichan Valley and Saltspring Island. Density estimators used are the GamPoi method (Magnussen 2008) and Keuls’ density equation (1963). Error bars represent 95% confidence intervals. The 2007 data subset for Saltspring showed a slightly lower density than the overall dataset, which could be due to the fact that the <360º sites (i.e. near disturbed edges), which were 32 excluded from the data subset, likely have higher density. However, the mean density of these subsets is still within the confidence interval of the overall dataset. Habitat type. A review of the habitat types described in the 1859 survey notes reveals a landscape composed of a mosaic of prairie, plains, open woods, and forest habitats (Figure 8). Of the 70% of points described by Oliver Wells, “forest” sites comprised 12.9% of described points, “open woods” comprised 20.6%, “plains” (described either as “open pine plains” or “pine and oak plains”) encompassed 34.2%, and “prairie” made up 17.4% of points in the Cowichan Valley. Habitat described as “swamp” and “bottom land” made up 7% and 8%, respectively. The density estimates for each of these habitat types correspond as expected with the surveyors description, albeit with large confidence intervals (Figure 9). Sites labeled as “forest” had an average density of 404.3 stems/ha (+/- 296.6), “open woods” 227.2 stems/ha (+/- 187.9), “plains” 61.3 stems/ha (+/- 41.0), and sites described as “prairie” 8.8 stems/ha (+/- 3.5). The density estimate for the “prairie” habitat is a maximum estimate, as nearly half of the sites in this habitat type had no bearing tree recorded at all (the 1859 notes state “no bearing tree near”). In this case, I used the maximum distance measured from a point to a bearing tree in 1859 (26 meters) as a substitute for the distance measurement when no bearing tree was marked. Undescribed sites or sites described as “bottom land” or “swamp” had an average density of 416 stems/ha, suggesting that these sites were largely (though not entirely) forests of some kind. 33 Figure 8: Relative proportion of each habitat type described by Oliver Wells in the Cowichan Valley, 1859. In total, 70% of all survey points were described by Wells. Figure 9: Tree density by habitat type in the Cowichan Valley, 1859. All estimates use the GamPoi method (Magnussen 2008), error bars represent 95% confidence intervals. “Other” includes those sites described as “bottom land” and “swamp,” and all undescribed points. 34 In 2007, 1.2% of sites had a density of less than 35 stems/ha (roughly corresponding to the “prairie” habitat type), 2.0% of sites had a density between 35 and 150 stems/ha (roughly corresponding to “plains”), 17.6% of sites had a density between 150 and 315 stems/ha (corresponding to “open woods”), and 79.2% of sites had a density of 315 stems/ha or greater (“forest”). Furthermore, a site-by-site comparison of past and present density based on the these categories shows that nearly all sites increased in density from past to present. There were no previously forested sites that became prairie or plains, but 60% of the prairie sites and 97% of plains sites became open woods or forest (Table 4). Although these clearly represent approximate categories of tree density, these analyses make it clear that open, sparsely treed habitat types decreased greatly between 1850 and 2007 and were replaced by forest. Table 4: Transition matrix showing the proportion of sites that changed from a low tree density to a higher density (bold numbers) and vice versa; 2007 habitat categories are based on equivalent density estimates from each presettlement habitat type. “Forest” also includes sites designated as “bottom land” and “swamp.” All rows sum to 1. The “total” column indicates the number of sites in each historical category. 35 Size structure. The Kolmogorov-Smirnov tests revealed a significant difference in the size distribution of trees between 1859 and 2007 (Table 5). Tree sizes (Figure 10) show a unimodal distribution in 1874, with few of the smallest trees and many medium-sized trees, while the 2007 trees show a monotonically decreasing pattern, with many small trees and fewer medium and large trees. This pattern is clear in both the bearing tree data and in the line tree data, which suggests that the bearing tree data are not highly size-biased. There was no significant difference between the size distribution of 1874 bearing trees and line trees, nor was there a difference when ten bearing trees were used instead of one in the 2007 dataset (Kolmogorov-Smirnov test results; Table 5). Table 5: Kolmogorov-Smirnov tests for differences in tree size distributions between 1874 and 2007 on Salt Spring Island; \"BT\" is the one bearing tree closest to each point, \"2007 All\" incorporates the ten closest trees at each site, and \"1874 Line\" is the set of line trees weighted by inverse DBH. 36 Figure 10: Tree size frequency histograms for Saltspring Island; the black, dashed line represents the bearing tree data from 1874, the grey dashed line represents the bearing tree data from 2007 (the one closest tree at each site), the black solid line represents the line tree diameters (weighted by inverse diameter) in 1874, and the solid grey line is all 2007 trees (up to 10 at each site). DISCUSSION My results show significant changes over the past 150 years in the composition, density, and size distributions of trees on southeastern Vancouver Island and Saltspring Island. I found a decrease in the abundance of Douglas-fir and an increase in maple and cedar across all three datasets. Forest density increased nearly two-fold between time periods, and forest size structure has shifted from primarily medium-large trees (35-45 cm in diameter) to dominance by trees of 15 cm in diameter or less. Finally, the proportion of the 37 landscape composed of open habitat (150 stems/ha or less) has decreased dramatically since pre-settlement times, from more than a third of the landscape to <5%. Patterns of habitat distribution The results of my study support several conclusions of previous research on the historical landscape of southeastern Vancouver Island. These studies have primarily focused on the rapid decline of oak savanna habitat in these areas (MacDougall et al. 2004, Gedalof et al. 2006, Lea 2006). This decline has been attributed in large part to development and conversion of oak prairie to agriculture, but even those areas of natural habitat that remain are threatened by encroachment of Douglas-fir and consequent conversion to forest (Gedalof et al. 2006). These findings are supported by the results of my study that show an increase in tree density across the landscape as well as a dramatic decline in the proportion of open habitat types. While this decline and degradation of oak savanna habitat had been well documented by previous studies, my research contributes novel information in several ways. The unique historical dataset has allowed quantitative characterization of tree density and composition, not only in particular habitats or localities but across a broad landscape. A particularly interesting result was the discovery of what might be considered a previously unrecognized historical habitat “type,” namely what Oliver Wells referred to as “pine plains”. Most of the “plains” in Fig. 6 were “pine plains,” although there were some references to “pine and oak plains”. The density estimate for this habitat type indicates a very open landscape with few trees, but where trees did occur, they were most often “pine” (Douglas-fir), in marked contrast to present day savannas where oaks predominate. At the front of the survey notebook for the Shawnigan district, Oliver Wells (1859) writes: 38 “In the Field Notes and Plan of Survey, the term 'Pine Plain' is only applied to land of the best quality, - open, - and little wood upon it, which usually grows in clumps with an occasional isolated tree. The picturesque effect being very similar to that of an extensive cultivated Park.” This description is consistent with the density estimate (61 stems/ha) and indicates a relatively open landscape. In the current landscape, however, habitat of this description is almost nonexistent, and there is no explicit mention of a habitat type fitting this description in other historical studies of this area, despite the fact that it occupied a substantial portion of the landscape. My results also indicate that the concern surrounding the decline of Garry oak trees per se may be somewhat misdirected. Although Garry oak trees are commonly assumed to be the cornerstone species of the threatened prairie and savanna habitats in B.C. (Hosten et al. 2006, Cavers 2008) – thus, “Garry oak ecosystems” – the trees themselves do not show any obvious decline in relative abundance despite the fact that the amount of savanna habitat has decreased by 90% or more. In fact, in the Chemainus district oak trees may have actually increased in relative abundance. In addition, the presettlement survey notes rarely indicate oak bearing trees; even in open areas of prairie and plains, where oaks now dominate, the bearing tree was usually either Douglas-fir or nonexistent (i.e., there is no tree within 26 meters of the post). This indicates that the “Garry oak” prairies and savannas of the presettlement landscape may have been characterized more by their open nature and likely by the understory plant community than by the presence of the Garry oak trees themselves. In other words, the large number of threatened herbaceous species within oak savanna habitats today probably have little relationship to or dependence on oak trees per se, but rather on areas of open, treeless habitat. This conclusion is supported by the work of Gedalof et al. 39 (2006), who found that oak recruitment at Rocky Point on Vancouver Island spiked during the late 1800’s, coincident with European settlement, and proposed that oaks may have been less abundant prior to this period. These results indicate that the areas of dense oak woodland that often characterize GOE site remnants today may actually be artifacts of European influence rather than a true reflection of the presettlement landscape. Inferring process from pattern The considerable changes in landscape and vegetation patterns that have occurred from the 1800’s to the present-day can be attributed at least in part to the large-scale disturbances initiated by European settlement. Although the most obvious landscape changes are due to the complete conversion of natural habitat to houses, parking lots, and agriculture, my study makes it clear that even the areas of “natural” habitat still remaining have changed dramatically since presettlement times. By comparing the results of my study to those of historical and experimental studies elsewhere in North America, I am able to identify some of the likely processes that have led to these changes. Possibly the largest influence of European settlement has been the ubiquitous logging of old-growth forests. In the low coastal area where this study occurs, less that one-half of one percent of the forest remains in a relatively undisturbed old-growth state (Flynn 1999). Clear-cut logging often leads to an increase in early successional species (Halpern 1989, Abrams & Nowacki 1992, Foster et al. 1998, Fuller et al. 1998), as it removes the commercially valuable, late successional species like Douglas-fir and cedar and provides an abundant light source for fast-growing early successional species such as alder and maple (Kimmins 1997). This is the most likely explanation for the large proportional increase in 40 maple. Maple is commonly found in disturbed areas due to its rapid growth and ability to sprout from the stump after logging (Uchytil 1989). The prevalence of logging likely also plays a role in the observed increases in forest density. After logging occurs, forest density will increase as forests of low density and large trees are replaced by denser forests with many smaller trees (Kaufmann et al. 2000). As the forest matures, density dependent mortality will cause thinning of the forest and an consequent decrease in forest density (He & Duncan 2000). Estimates of forest density in current old-growth Douglas-fir forests generally range from 250-500 stems per hectare, depending on habitat type, elevation, and aspect (Spies & Franklin 1991). The average historical density observed in my study is approximately 300 stems/hectare, which falls within the lower range of old-growth forest and is consistent with other studies of presettlement forest density (Zhang 2000). However, it should be noted that this density value actually represents the average of a wide range of habitat types and does not indicate that the entire landscape consists of forests of uniform density. Widespread logging also likely accounts for some of the observed decrease in average tree diameter, as logging removes the largest trees and allows small trees to grow up in their place. However, it does not explain the unimodal distribution of tree sizes, as even old- growth forests generally exhibit the common inverse J-shaped distribution (McCarthy and Weetman 2006, Westphal et al. 2006). Furthermore, logging does not explain the disappearance of open habitat types (i.e., “prairie” and “plains”) which once accounted for a combined 35% of the presettlement landscape. The tree density of these habitat types is much lower than a typical old growth forest, which indicates that other factors must have played a role in keeping these areas open. 41 Finally, an increase in logging fails to account for the observed increase in red cedar from past to present. Cedar is a shade-tolerant and highly valued species, and it is therefore unlikely that logging in the present day would cause an increase in its abundance. However, some form of preferential harvesting may explain the low relative abundance of cedar in the past. Red cedar was an important species in many western First Nations cultures, and was used extensively for building houses, canoes, totem poles, and more (Turner 1979, Deur & Turner 2005). Wide-spread harvesting of red cedar by First Nations could have affected the abundance of this species in the pre-settlement landscape, especially in areas of high population density. A study of lake sediment pollen on Anthony Island, north of Vancouver Island, revealed a significant decline in the abundance of cedar coincident with human occupation of the island approximately 1000 years BP. This trend could not be explained by any changes in climate, and was therefore taken as evidence of First Nations harvesting (Lacourse et al. 2007). It seems likely that a similar effect of harvesting could be evident on Vancouver Island, where First Nations populations were some of the largest in North America (Harris 1994, Boyd 1996). The relative increase of cedar in the present-day could then be attributed to a decline in First Nations activity following European settlement. Although industrial logging now removes far more trees overall than were harvested by First Nations peoples, it may not select as strongly for cedar. Frequent low intensity fires historically and subsequent widespread fire suppression appear to be critical in explaining some of the observed changes from presettlement times to the present. The suppression of fire can have a profound affect on community composition, and has been linked with shifts in species composition in many areas (Barrett & Arno 1982, Radeloff et al. 1999, Nowacki & Abrams 2008). The increase in cedar trees from past to 42 present could also be explained as a consequence of fire suppression. Cedar trees have very thin bark and shallow roots, and are highly susceptible to fire (Minore 1983, Tesky 1992, Agee 1993). In addition, cedar foliage is highly flammable, so even low-intensity fires can kill small cedar trees (Agee 1993). As a result, frequent fires in some areas of the past landscape could have restricted the range of red cedar to wet areas less prone to fire and reduced its abundance overall. This hypothesis is supported by the changes in species composition at each site (Appendix B). In all datasets combined, 42% of sites changed from a bearing tree species of high fire tolerance to a species of low fire tolerance. In contrast, only 10% of sites changed from a species of low fire tolerance to high fire tolerance, while 48% of sites remained the same. This indicates an overall shift in forest composition toward species of low fire resistance coincident with the suppression of fire. The increase in cedar could also be an indirect result of the suppression of fire. As cedar trees are highly shade tolerant (Tesky 1992), the observed expansion of forest into formerly open areas may have facilitated a corresponding increase in cedar. However, as the expansion of forest is itself a likely result of fire suppression, the increase in cedar would therefore be an indirect result of fire suppression. In either of these scenarios, cedar trees may have been able to increase in abundance and possibly spread to areas where they were formerly excluded by fire. The shift in forest size structure from a unimodal diameter distribution in the past to a monotonically decreasing distribution in the present is also likely due in large part to the suppression of fire. The unimodal pattern visible in the historical Saltspring Island data has been observed in forests with a regime of frequent fires (Peterson & Reich 2001), and was described in at least one other study of presettlement tree size distributions (Zhang et al. 43 2000). The experimental study by Peterson & Reich (2001) of the effects of prescribed burning in oak savannas of the Midwestern United States showed that fires of high frequency (three or more per decade) produced the same unimodal distribution of sizes seen in the presettlement Vancouver Island data. In the same study, a regime of infrequent fires (less than two per decade) or no fires produced a monotonically decreasing distribution of tree sizes more similar to that seen in the 2007 Saltspring data. A regime of frequent but low- intensity fire will kill most young seedlings and limit successful establishment. Older trees, particularly fire-resistant species such as Douglas-fir and Garry oak, will not be killed by a low-intensity fire and will continue to grow unharmed (Agee 1993). Occasionally, the fire interval will be longer than normal or a seedling will survive by chance until it reaches an age where it is no longer susceptible to fire (Agee 1993). This process creates a unimodal distribution of tree sizes, with very few small trees surviving each year but occasional pulses of recruitment (Peterson & Reich 2001). Fire suppression is also likely implicated in the increase of forest density from past to present. The suppression of fire has been linked to increases in forest density throughout North America, even in unlogged forests (Covington & Moore 1994). In the southwestern United States, frequent fires played an important role in the presettlement ecology of ponderosa pine (Pinus ponderosa) forests (Covington & Moore 1994, Mast et al. 1999). Historical records of these ponderosa pine forests indicate an “open, park-like” habitat, with a maximum canopy cover of 25-30% and an understory dominated by grasses (Covington & Moore 1994, Mast et al. 1999). Density estimates in these forests reveal an average tree density of approximately 60 stems per hectare – a number very similar to that for the “pine plains” of Vancouver Island (61 stems/ha). Frequent, low intensity fires were responsible for 44 keeping these forests in their open state, and since fire suppression began in the early 1900’s, forest density has increased to 800 stems/ha or more. Another study in the Wisconsin pine barrens found a high proportion of relatively open barrens (less than 375 stems/ha) in the presettlement landscape and concluded that these areas had been kept open by a frequent fire regime in the past (Radeloff 1999). Many studies of fire regimes throughout North America have found similar increases in forest density (e.g. Taylor 2000, Shumway et al. 2001). In all of these studies, fire suppression has led to a dramatic increase in forest density and near disappearance of open habitat types in the modern landscape. Depending on environmental conditions, the frequency of fire can determine whether a particular area is prairie, savanna or forest (Nowacki & Abrams 2008). Encroachment of forest into areas of prairie following fire suppression has been documented in several locations (Agee 1993) and has been observed in the oak savannas of Vancouver Island as well (Gedalof et al. 2006, Pellatt et al. 2007). In these sites, the soil is often deep and moist enough to support a forest, but frequent fires had previously prevented seedlings from establishing. With the suppression of fire, seedlings are able to establish and gradually grow to form a closed canopy over the formerly open prairie. In sum, fire suppression not only leads to an increase in density within the forests themselves, it also contributes to the conversion of prairie to forest. This process is especially apparent in the rapid decline of open habitat types of Vancouver Island from presettlement times to the present day. Density estimates for each habitat type indicate that more than a third of the landscape (“prairie” and “plains”) averaged <150 stems per hectare, far less than the observed density of contemporary old-growth forests. In contrast, less than 4% of the current landscape has a density of <150 stems per hectare. Furthermore, this trend holds true regardless of whether 45 all historical sites or only resurveyed historical sites are included in the analysis, which means the observed reduction in open habitat is not just a byproduct of preferential agriculture or development in more open habitat types. This result again is consistent with a regime of frequent fire in the presettlement landscape, which acted to keep these habitats in their open state. Although my research did not look specifically at the role of anthropogenic vs. natural burning, my results provide indirect evidence for a fire regime strongly influenced by First Nations communities. A study of the natural fire return interval in the Douglas-fir forests of Vancouver Island estimates a fire every 96-392 years, based on the frequency of lightning strikes in this region (Pew & Larsen 2001). The more localized study of lake charcoal sediment in the Cowichan Valley revealed a fire return interval of 27-41 years (McCoy 2006, Pellatt et al. 2007). However, experimental studies indicate that the unimodal tree size distribution requires more frequent fires, often of three year intervals or less (Peterson & Reich 2001). This supports the idea that more fires occurred than would be expected given the “natural” fire return interval, and were therefore likely related to anthropogenic activity. Logging and fire suppression are by no means the only two factors which have influenced the profound changes from presettlement to present. Exotic species, climate change, and herbivory have also been found to have large impacts on plant communities in this region (Hamann & Wang 2006, Best 2007, Gonzales 2007, MacDougall 2008). Invasive plant species have had a particularly dramatic impact on the composition of the oak savanna understory plant community (Fuchs 2001, Alverson 2005, Lilley 2007). However, there are very few invasive tree species in natural habitats in this region, and so any substantial 46 influence of exotics on tree species composition is unlikely. An abundance of herbivores can have an effect similar to that of fire – reducing the cover of shrubs and trees through herbivory and contributing to the maintenance of open grassland habitat (van Langevelde et al. 2003, Côté et al. 2004). While this may indeed have played a role in the presettlement landscape, it is unlikely that the past abundance of herbivores is larger than today. Due to the extirpation of natural predators on Vancouver Island and the reduced role of hunting, it is probable that current levels of herbivores, especially deer, are higher now than at any time in recent history (MacDougall 2008). However, it is possible that the large population of Roosevelt Elk, now extirpated from the southern part of the island, played a role in limiting tree recruitment through herbivory and thus helped to maintain an open landscape (Johnston & Cushman 2007, MacDougall 2008). Finally, several studies have addressed the impact of climate change on ecosystems in southwestern British Columbia (e.g.,Walker & Pellatt 2001, Mote 2003, Hamann & Wang 2006). Within the CDF zone, there have been modest increases in mean annual temperature (~0.7 ºC) and precipitation (13%), especially winter precipitation, over the past century (Mote 2003, Hamann & Wang 2006). While these changes in climate undeniably will have an impact on landscapes throughout the Pacific Northwest (Taylor & Taylor 1997, Hamann & Wang 2006), it seems unlikely that they have led to the specific patterns observed in my study. For example, climate change models predict an expansion of the CDF zone in general (Hamann & Wang 2006) and an increase in open woodlands and prairies (Hebda 1997) coincident with warming temperatures, rather than the dramatic decline in open habitat types that have been observed. Furthermore, although some climate models do predict an increase in cedar (+14% by 2025), the same models predict an even greater increase in Douglas-fir 47 (+32% by 2025; Hamann & Wang 2006). Therefore, it is unlikely that the reduction in open habitat types and the relative decrease in Douglas-fir that I observed are the result of recent changes in climate. IMPLICATIONS FOR CONSERVATION Historical data can provide valuable information about the processes underlying present landscape patterns. My study provides strong evidence of dramatic changes in species composition, forest size structure, and forest density in natural areas across the landscape since presettlement times. The direction of these changes and the vegetation patterns evident in the presettlement landscape likely derive from a frequent fire regime that was influenced by anthropogenic burning. Current vegetation patterns likely reflect the results of extensive logging and fire suppression. Restoration efforts are often prone to some uncertainty about the target conditions they seek to create, especially in terms of species composition, abundance, and management regimes (Higgs 1997, Lindenmeyer et al. 2007, Hobbs & Cramer 2008). This is particularly true in those areas where historical landscape conditions are not known because no reference sites remain or those that do have been greatly altered from their presettlement state. Often, land managers will follow a “do nothing” approach, and allow the land to return to its “natural” state (Hobbs & Cramer 2008). However, my study and previous studies in this area indicate that the open nature of much of the endangered oak savannas on Vancouver Island is largely anthropogenic in origin, maintained by frequent fires purposefully set by First Nations peoples. As a result, any attempt to restore these habitats to their pre-European state can not be accomplished by following the “do-nothing” approach. A restoration scheme with the goal of maintaining an open savanna habitat would almost certainly have to involve the 48 active removal of encroaching trees and shrubs, either through frequent burning or alternative strategies with similar outcomes such as mowing or manual tree removal (MacDougall et al. 2004, Gedalof et al. 2006). The challenges that restoration efforts often face illustrate a common debate in the realm of restoration ecology. Many ecologists note the futility of attempting to restore an ecosystem to a particular presettlement condition, because ecosystems are constantly shifting (Foster 2000b, Sinclair & Byrom 2006). Furthermore, there is no guarantee that the prevalent vegetation patterns of 1859 would still be visible today, even in the absence of European influence (Foster and Motzkin 1998, Foster 2000b). However, the extensive loss of oak savanna habitat over a short time period is undeniably due, at least in part, to the dramatic impact of European settlement, and cannot be attributed solely to changes in natural environmental cycles. Furthermore, the large number of species-at-risk in this habitat means that allowing this ecosystem to disappear will also lead to the loss of most of these species. Garry oak habitat has therefore been listed as a conservation priority by several national, regional, and local institutions (Erickson 1993, Fuchs 2001, GOERT 2008), and restoration attempts are already underway. My research provides evidence of the presettlement processes that maintained these ecosystems, which can be used in whatever restoration attempts are deemed appropriate by these institutions. The results of my study reinforce the importance of understanding historical patterns and processes when making decisions about the conservation and restoration of threatened ecosystems. 49 REFERENCES Abrams, M. and C. Ruffner. 1995. Physiographic analysis of witness-tree distribution (1765- 1798) and present forest cover through north central Pennsylvania. Canadian Journal of Forest Research 25:659-668. Abrams, M. and G. Nowacki. 1992. Historical variation in fire, oak recruitment, and post- logging accelerated succession in central Pennsylvania. Bulletin of the Torrey Botanical Club 119:19-28. Agee, J.K. 1993. Fire Ecology of Pacific Northwest Forests. Island Press: Washington, D.C. Agee, J.K. 1998. The landscape ecology of western forest fire regimes. Northwest Science 72:24-34. Almendinger, J.C. 1996. Minnesota's Bearing Tree Database. Minnesota Department of Natural Resources: St. Paul, Minnesota. Alverson, E. 2005. Preserving prairies and savannas in a sea of forest: a conservation challenge in the Pacific Northwest. Plant Talk 40:23-27. Anonymous. 1849. Article in the Times newspaper. University of British Columbia Archives, Vancouver, Canada. Arnett, C. 1999. The Terror of the Coast. Talonbooks: Burnaby, BC, Canada. Barrett, S. and S. Arno. 1982. Indian fires as an ecological influence in the northern Rockies. Journal of Forestry 80:647-651. Batek, M.J., A.J. Rebertus, W.A. Schroeder, T.L. Haithcoat, E. Compas, and R.P. Guyette. 1999. Reconstruction of early nineteenth-century vegetation and fire regimes in the Missouri Ozarks. Journal of Biogeography 26:397-412. Beckwith, B.R. 2004. “The queen root of this clime”: ethnoecological investigations of blue camas (Camassia leichtlinii (Baker) Wats., C. quamash (Pursh) Greene; Liliaceae) and its landscapes on southern Vancouver Island, British Columbia. Ph.D. Dissertation. University of Victoria, Victoria, British Columbia. Best, R.J. 2007. Exotic herbivores facilitate the exotic grasses they graze: mechanisms for an unexpected \"invasional meltdown\". M.Sc. Thesis, University of British Columbia, Canada. Bouldin, J. 2008. Some problems and solutions in density estimation from bearing tree data: a review and synthesis. Journal of Biogeography 35:2000-2011. Bourdo, E.A. 1956. A review of the General Land Office survey and of its use in quantitative studies of former forests. Ecology 37:754–768. 50 Boyd, R. 1996. Commentary on early contact-era smallpox in the Pacific Northwest. Ethnohistory 43:307-328. Boyd, R. 1999. Indians, Fire and the Land in the Pacific Northwest. Oregon State Univ. Press, Corvallis, Oregon. Brown, D. 1998. Mapping historical forest types in Baraga County Michigan, USA as fuzzy sets. Plant Ecology 134:97-111. Brown, K.J. and R.J. Hebda. 2002. Origin, development, and dynamics of coastal temperate conifer rainforests of southern Vancouver Island, Canada. Canadian Journal of Forest Research 32:353-372 Byth, K. and B.D. Ripley. 1980. On sampling spatial patterns by distance methods. Biometrics 36:279-284. Cavers, M.R. 2008. Sub Quercu Felicitas: Place, knowledge, and Victoria's Garry oaks, 1843-2008. M.Sc. Thesis, Queen's University. Cochran, W.G. 1977. Sampling Techniques, 3rd ed. Wiley, New York. Cogbill, C.V., J. Burk, and G. Motzkin. 2002. The forests of presettlement New England, USA: spatial and compositional patterns based on town proprieter surveys. Journal of Biogeography 29:1279-1304. Cook, Major A.C., and Col. H. James. 1860. Vancouver Island Colony: Comiaken District. Topographical Depot of the War Office, England. Cook, Major A.C., and Col. H. James. 1860. Vancouver Island Colony: Cowichan District. Topographical Depot of the War Office, England. Cook, Major A.C., and Col. H. James. 1860. Vancouver Island Colony: Quamichan District. Topographical Depot of the War Office, England. Cook, Major A.C., and Col. H. James. 1860. Vancouver Island Colony: Shawnigan District. Topographical Depot of the War Office, England. Cook, Major A.C., and Col. H. James. 1860. Vancouver Island Colony: Somenos District. Topographical Depot of the War Office, England. Côté, S.D., T.P. Rooney, J. Tremblay, C. Dussault, and D.M. Waller. 2004. Ecological impacts of deer overabundance. Annual Review of Ecology and Systematics 35:113- 147. Cottam, G. & J.T. Curtis. 1956. The use of distance measures in phytosociological sampling. Ecology 37:451–460. 51 Covington, W.W. and M.M. Moore. 1994. Southwestern ponderosa forest structure: changes since Euro-American settlement. Jornal of Forestry 92:39-47. Cronon, W. 1983. Changes in the land: Indians, Colonists, and the ecology of New England. Hill and Wang, New York. Davis, M.B. 1988. Ecological Systems and Dynamics [In Toward an Understanding of Global Change: Initial Priorities for U.S. Contributions to the International Geosphere-Biosphere Program, chairman Harold A. Mooney]. National Academy Press: Washington, D.C. Denevan, W. 1992. The pristine myth: the landscape of the Americas in 1492. Annals of the Association of American Geographers 82:369-385. Deur, D. and N.J. Turner. 2005. Keeping it living: traditions of plant use and cultivation on the Northwest Coast of North America. University of Washington Press: Seattle, Washington. Douglas, G.W., G.B. Straley, D.V. Meidinger, and J. Pojar (editors). 1998. Illustrated Flora of British Columbia. Volume 1: Gymnosperms and Dicotyledons (Aceraceae Through Asteraceae). B.C. Ministry of Environment, Lands & Parks and B.C. Ministry of Forests, Victoria. Douglas, G.W., G.B. Straley, D.V. Meidinger, and J. Pojar (editors). 1998. Illustrated Flora of British Columbia. Volume 2: Dicotyledons (Balsaminaceae Through Cucurbitaceae). B.C. Ministry of Environment, Lands & Parks and B.C. Ministry of Forests, Victoria. Douglas, G.W., D.V. Meidinger, and J. Pojar (editors). 1999. Illustrated Flora of British Columbia. Volume 3: Dicotyledons (Diapensiaceae Through Onagraceae). B.C. Ministry of Environment, Lands & Parks and B.C. Ministry of Forests, Victoria. Douglas, G.W., D.V. Meidinger, and J. Pojar (editors). 1999. Illustrated Flora of British Columbia. Volume 4: Dicotyledons (Orobanchaceae Through Rubiaceae). B.C. Ministry of Environment, Lands & Parks and B.C. Ministry of Forests, Victoria. Douglas, G.W., D.V. Meidinger, and J. Pojar (editors). 2000. Illustrated Flora of British Columbia. Volume 5: Dicotyledons (Salicaceae Through Zygophyllaceae) And Pteridophytes. B.C. Ministry of Environment, Lands & Parks and B.C. Ministry of Forests, Victoria. Duffin, K., L. Gillson, and K. Willis. 2008. Testing the sensitivity of charcoal as an indicator of fire events in savanna environments: quantitative predictions of fire proximity, area and intensity. The Holocene 18:279-291. Eberhardt, L. 1967. Some developments in distance sampling. Biometrics 23:207-216. 52 Egan, B. 1999a. The ecology of the coastal western hemlock zone. Ministry of Forests, British Columbia, Canada. Egan, B. 1999b. The ecology of the coastal Douglas-fir zone. Ministry of Forests, British Columbia, Canada. Egan, D. and E.A. Howell. 2005. Historical Ecology Handbook. Island Press: Washington, D.C. Engeman, R.M., R.T. Sugihara, L.R. Pank, and W.E. Dusenberry. 1994. A comparison of plotless density estimators using Monte Carlo simulation. Ecology 75:1769-1779. Erickson, W. 1993. Ecosystems in British Columbia at risk: Garry oak ecosystems. British Columbia Ministry of Environment, Lands and Parks. Erickson, W. 1996. Classification and interpretation of Garry oak (Quercus garryana) plant communities and ecosystems in southwestern British Columbia. M.Sc. thesis, Department of Geography, University of Victoria. Erickson, W. and D. Meidinger. 2007. Garry oak (Quercus garryana) plant communities in British Columbia: a guide to identification. British Columbia Ministry of Forests and Range Forest Science Program. Technical Report 040. Flynn, S. 1999. Ecosystems in British Columbia at risk: coastal Douglas-fir ecosystems. Ministry of Environment, Lands and Parks, British Columbia, Canada. Foster, D.R. 2000a. From bobolinks to bears: interjecting geographical history into ecological studies, environmental interpretation, and conservation planning. Journal of Biogeography 27: 27-30 Foster, D.R. 2000b. Conservation lessons & challenges from ecological history. Forest History Today Fall:1-11. Foster, D.R. and G. Motzkin. 1998. Ecology and conservation in the cultural landscape of New England: lessons from nature's history. Northeastern Naturalist 5: 111-126. Foster, D.R., G. Motzkin, and B. Slater. 1998. Land-use history as long-term broad-scale disturbance: regional forest dynamics in central New England. Ecosystems 1:96-119. Foster, D.R., P.K. Schoonmaker, and S.T.A. Pickett. 1990. Insights from paleoecology to community ecology. Trends in Ecology & Evolution 5:119-122. Foster II, H.T., B. Black, and M.D. Abrams. 2004. A witness tree analysis of the effects of Native American Indians on the pre-European settlement forests in east-central Alabama. Human Ecology 32:27-47. 53 Fuchs, M.A. 2001. Towards a recovery strategy for Garry oak and associated ecosystems in Canada: ecological assessment and literature review. Technical Report GBEI/EC-00- 030. Environment Canada, Canadian Wildlife Service, Pacific and Yukon Region. Fuller, J.L., D.R. Foster, J.S. McLachlan, and N. Drake. 1998. Impact of human activity on regional forest composition and dynamics in central New England. Ecosystems 1:76- 95. Garry Oak Ecosystems Recovery Team. Website accessed November, 2008. http://www.goert.ca/ Garry Oak Meadow Preservation Society (GOMPS). Website accessed November, 2008. http://www.geocities.com/RainForest/8132/gomps.html Garry Oak Restoration Project (GORP). Website accessed November, 2008. http://www.gorpsaanich.com/ Gedalof, Z., M.G. Pellat, D.J. Smith. 2006. From prairie to forest: three centuries of environmental change at Rocky Point, Vancouver Island, British Columbia. Northwest Science 80:34-46. Gómez-Pompa, A. and A. Kaus. 1992. Taming the wilderness myth. BioScience 42:271-279. Gonzales, E.K. 2007. The effects of herbivory, competition, and disturbance on island meadows. Ph.D. Dissertation. University of British Columbia, Canada. Green, A. 1874. Field Book, Saltspring Island. [Survey notebooks in the B.C. Land Title & Survey archives, Victoria, BC, Canada] Halpern, C.B. 1989. Early successional patterns of forest species: interactions of life history traits and disturbance. Ecology 70:704-720. Hamann, A. and T. Wang. 2006. Potential effects of climate change on ecosystem and tree species distribution in British Columbia. Ecology 87:2773-2786. Harris, C. 1994. Voices of disaster: smallpox around the Strait of Georgia in 1782. Ethnohistory 41:591-626. He, F. and R.P. Duncan. 2000. Density-dependent effects on tree survival in an old-growth Douglas fir forest. Journal of Ecology 88:676-688. He, H.S., D.J. Mladenoff, T.A. Sickley, and G.G. Guntenspergen. 2000. GIS interpolations of witness tree records (1839-1866) for northern Wisconsin at multiple scales. Journal of Biogeography 27:1031-1042. Hebda, R. 1997. Impact of climate change on biogeoclimatic zones of British Columbia and Yukon. Pages 194-208 in E. Taylor and B. Taylor, editors. Responding to global 54 climate change in British Columbia and Yukon. BC Ministry of Environment, Lands and Parks. Higgs, E.S. 1997. What is Good Ecological Restoration? Conservation Biology 11:338-348. Hobbs, R.J. and V.A. Cramer. 2008. Restoration ecology: interventionist approaches for restoring and maintaining ecosystem function in the face of rapid environmental change. Annual Review of Environment and Resources 33:1-23. Hosten, P.E., O.E. Hickman, F.K. Lake, F.A. Lang, and D. Vesely. 2006. Oak Woodlands and Savannas. [Pages 63-92 in D. Apostal and M. Sinclair, editors. Restoring the Pacific Northwest: the art and science of ecological restoration in Cascadia]. Society for Ecological Restoration International. Island Press, Washington, D.C. Janke, R.A., D. McKaig, and R. Raymond. 1978. Comparison of presettlement and modern upland boreal forests on Isle Royale National Park. Forest Science 24:115-121. Johnson, B. and J. Cushman. 2007. Influence of a large herbivore reintroduction on plant invasions and community composition in a California grassland. Conservation Biology 21:515-526. Kaufmann, M.R., C.M. Regan, P.M. Brown. 2000. Heterogeneity in ponderosa pine/Douglas-fir forests: age and size structure in unlogged and logged landscapes of central Colorado. Canadian Journal of Forest Research 30:698-711. Keuls M., H.J. Over, and C.T. De Wit. 1963. The distance method for estimating densities. Statistica Neerlandica 17:71–91. Kimmins, J.P. 1997. Forest Ecology: A Foundation for Sustainable Management, 2 nd ed. Prentice-Hall, Inc: New Jersey. pp. 399-448. Lacourse, T., R.W. Mathewes, and R.J. Hebda. 2007. Paleoecological anlyses of lake sediments reveal prehistoric human impact on forests at Anthony Island UNESCO World Heritage Site, Queen Charlotte Islands (Haida Gwaii), Canada. Quaternary Research 68: 177-183 Lea, T. 2006. Historical Garry oak ecosystems of Vancouver Island, British Columbia, pre- European contact to the present. Davidsonia 17:34-50. Leak, W.B. 1965. The J-shaped probability distribution. Forest Science 11:405-409. Lillard, C. 1986. Seven Shillings a Year: The History of Vancouver Island. Horsdal & Schubart Publishers Ltd: Ganges, BC, Canada. Lilley, P.L. 2007. Determinants of native and exotic plant diversity and composition in remnant oak savannas on Southeastern Vancouver Island. M.Sc. Thesis, University of British Columbia. 55 Lindenmayer, D., R.J. Hobbs, R. Monague-Drake, J. Alexandra, A. Bennett, M. Burgman, P. Cale, A. Calhoun, V. Cramer, P. Cullen, D. Driscoll, L. Fahrig, J. Fischer, J. Franklin, Y. Haila, M. Hunter, P. Gibbons, S. Lake, G. Luck, C. MacGregor, S. McIntyre, R. Mac Nally, A. Manning, J. Miler, H. Mooney, R. Noss, H. Possingham, D. Saunders, F. Schmiegelow, M. Scott, D. Simberloff, T. Sisk, G. Tabor, B. Walker, J. Wiens, J. Woinarski, E. Zavaleta. 2007. A checklist for ecological management of landscapes for conservation. Ecology Letters 10:1-14. Lorimer, C.G. 1980. Age Structure and Disturbance History of a Southern Appalachian Virgin Forest. Ecology 61:1169-1184. Lunt, I.D. and P.G. Spooner. 2005. Using historical ecology to understand patterns of biodiversity in fragmented agricultural landscapes. Journal of Biogeography 32:1859- 1873. Lynch, T. and R. Rusydi. 1999. Distance sampling for forest inventory in Indonesian teak plantations. Forest Ecology and Management 113:215-221. MacDougall, A.S. 2008. Herbivory, hunting, and long-term vegetation change in degraded savanna. Biological Conservation 141:2174-2183. MacDougall, A.S., B.R. Beckwith, and C.Y. Maslovat. 2004. Defining conservation strategies with historical perspectives: a case study from a degraded oak grassland ecosystem. Conservation Biology 18:455-465. MacDougall, A.S. and R. Turkington. 2005. Are invasive species the drivers or passengers of change in degraded ecosystems?. Ecology 86:42-55. Magnussen, S., C. Kleinn, and N. Picard. 2008. Two new density estimators for distance sampling. European Journal of Forest Research 127:213-224. Manies, K.L. and D.J. Mladenoff. 2000. Testing methods to produce landscape-scale presettlement vegetation maps from the U.S. public land survey records. Landscape Ecology 15:741-754. Manies, K.L., D.J. Mladenoff, and E.V. Nordheim. 2001. Assessing large-scale surveyor variability in the historic forest data of the original U.S. Public Land Survey. Canadian Journal of Forest Research 31:1719-1730. Marshall, D.P. 1999. Those who fell from the sky: a history of the Cowichan peoples. Sandhill Books, Kelowna, British Columbia. Mast, J.N., P.Z. Fulé, M.M. Moore, W.W. Covington, and A.E.M. Waltz. 1999. Restoration of presettlement age structure of an Arizona ponderosa pine forest. Ecological Applications 9:228-239. McCarthy, J.W. and G. Weetman. 2006. Age and size structure of gap-dynamic, old-growth boreal forest stands in Newfoundland. Silva Fennica 40:209-230. 56 McCarty, J. 2001. Ecological consequences of recent climate change. Conservation Biology 15:320-331. McCoy, M.M. 2006. High resolution fire and vegetation history of Garry oak ecosystems in British Columbia. M.Sc. Thesis. Meidinger, D. and J. Pojar. 1991. Ecosystems of British Columbia. BC Ministry of Forests, Victoria, British Columbia. Minore, D. 1983. Western redcedar: a literature review. USDA Forest Service, Pacific Northwest Forest and Range Experiment Station, Portland, Oregon. Gen. Tech. Rep. PNW-150. Moore, John M. 1871. Instructions to the surveyors general of public lands of the United States for those surveying districts established in and since the year 1850; containing, also, a manual of instructions to regulate the field operations of deputy surveyors. Government Printing Office, Washington. Morisita, M. 1957. A new method for the estimation of density by the spacing method, applicable to non-randomly distributed populations. Physiology and Ecology – Kyoto 7: 134–144. [English translation (1960) Division of Range Management and Wildlife Habitat Research, US Forest Service, M-5123.] Mote, P.W. 2003. Trends in temperature and precipitation in the Pacific Northwest during the twentieth century. Northwest Science 77:271-282. North, M., D. Holdsworth and J. Teversham. 1977. A brief guide to the use of land surveyors' notebooks in the Lower Fraser Valley, B.C., 1859-1890. BC Studies 34:45-60. Nowacki, G., and M. Abrams. 2008. The demise of fire and “mesophication” of forests in the eastern United States. BioScience 58:123-138. Pellatt, M.G., Z. Gedalof, M. McCoy, K. Bodtker, A. Cannon, S. Smith, B. Beckwith, R. Mathewes, and D. Smith. 2007. Fire history and ecology of Garry oak and associated ecosystems in British Columbia: final report for the Interdepartmental Recovery Fund Project 733. Parks Canada Western and Northern Service Centre, British Columbia, Canada. Persson, O. 1964. Distance methods: the use of distance measurements in the estimation of seedling density and open space frequency. Skogshögskolan. 15:68. Peterson, D.W. and P.B. Reich. 2001. Prescribed fire in oak savanna: fire frequency effects on stand structure and dynamics. Ecological Applications 11:914-927. Pew, K.L. and C.P.S. Larsen. 2001. GIS analysis of spatial and temporal patterns of human- caused wildfires in the temperate rain forest of Vancouver Island, Canada. Forest Ecology and Management 140:1-18. 57 Pollard, J.H. 1971. On Distance Estimators of Density in Randomly Distributed Forests. Biometrics 27:991-1002. Radeloff, V.C., D.J. Mladenoff, H.S. He, and M.S. Boyce. 1999. Forest landscape change in the northwestern Wisconsin Pine Barrens from pre-European settlement to the present. Canadian Journal of Forest Research 29:1649-1659. Ralph, W. 1864. Survey of part of the Chemainus District. [Survey notebooks in the B.C. Land Title & Survey archives, Victoria, BC, Canada] Rouvinen, S. and T. Kuuluvainen. 2005. Tree diameter distributions in natural and managed old Pinus sylvestris-dominated forests. Forest Ecology and Management 208:45-61. Schulte, L.A. and D.J. Mladenoff. 2001. The original US Public Land Survey records: their use and limitations in reconstructing presettlement vegetation. Journal of Forestry October:5-10. Shumway, D., M. Abrams, and C. Ruffner. 2001. A 400-year history of fire and oak recruitment in an old-growth oak forest in western Maryland, U.S.A. Canadian Journal of Forest Research 31:1437-1443. Sinclair, A.R.E, and A.E. Byrom. 2006. Understanding ecosystem dynamics for conservation of biota. Journal of Animal Ecology 75:64-79. Sokal, R.R. and F.J. Rohlf. 1995. Biometry. W.H. Freeman and Company, New York. Spies, T.A. and J.F. Franklin. 1991. The structure of natural young, mature, and old-growth Douglas-fir forests in Oregon and Washington. USDA Forest Service, Pacific Northwest Research Station. General Technical Report PNW-GTR-285. Steinberg, P.D. 2001. Populus balsamifera. In: Fire Effects Information System, [Online]. USDA Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory. Taylor, A. 2000. Fire regimes and forest changes in mid and upper montane forests of the southern Cascades, Lassen Volcanic National Park, California, USA. Journal of Biogeography 27:87-104. Taylor, E. and B. Taylor, editors. 1997. Responding to global climate change in British Columbia and Yukon. BC Ministry of Environment, Lands and Parks. Tesky, J.L. 1992. Thuja plicata. In: Fire Effects Information System, [Online]. USDA Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory. Turner, N.J. 1979. Plants in British Columbia Indian Technology. British Columbia Provincial Museum, Victoria, Canada. Turner, N.J. 1999. “Time to burn”: traditional use of fire to enhance resource production by Aboriginal Peoples in British Columbia. Pages 185-218 in R. Boyd, editor. Indians, 58 Fire and the Land in the Pacific Northwest. Oregon State Univ. Press, Corvallis, Oregon. Tveten, R. 1997. Fire effects on prairie vegetation Fort Lewis, Washington. Pages 123-130 in P.V. Dunn and K. Ewing, editors. The Ecology and Conservation of the South Puget Sound Landscape. Nature Conservancy, Seattle. Uchytil, R.J. 1989. Acer macrophyllum. In: Fire Effects Information System, [Online]. USDA Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory. van Langevelde, F.V., C.A.D.M. van de Vijver, L. Kumar, J. van de Koppel, N. de Ridder, J. van Andel, A.D. Skidmore, J.W. Hearne, L. Stroosnijder, W.J. Bond, H.H.T. Prins, and M. Rietkerk. 2003. Effects of fire and herbivory on the stability of savanna ecosystems. Ecology 84:337-350. Vellend, M., A.D. Bjorkman, A. McConchie. 2008. Environmentally biased fragmentation of oak savanna habitat on southeastern Vancouver Island, Canada. Biological Conservation 141:2576-2584. Walker, I.R., and M.G. Pellatt. 2001. Climate change and impacts in southern British Columbia - a palaeoenvironmental perspective. Unpublished report to the Climate Change Action Fund, Natural Resources Canada, Ottawa. Wells, O. 1859. Cowichan Valley Surveys: Cowichan District. [Survey notebooks in the B.C. Land Title & Survey archives, Victoria, BC, Canada] Wells, O. 1859. Quamichan District Field Notes. [Survey notebooks in the B.C. Land Title & Survey archives, Victoria, BC, Canada] Wells, O. 1859. Cowichan Valley: Shawnigan or Southerly District. [Survey notebooks in the B.C. Land Title & Survey archives, Victoria, BC, Canada] Westphal, C., N. Tremer, G. Oheimb, J. Hansen, K. von Gadow, and W. Härdtle. 2006. Is the reverse J-shaped diameter distribution universally applicable in European virgin beech forests?. Forest Ecology and Management 223:75-83. Whitney, G.G. 1994. From coastal wilderness to fruited plain: a history of environmental change in temperate North America, 1500 to the present. Cambridge University Press, Cambridge, England. Whitney, G.G. and J.P. DeCant. 2005. Government Land Office Surveys and Other Early Land Surveys. Pages 147-172 in D. Egan and E.A. Howell, editors. Historical Ecology Handbook. Island Press, Washington, D.C. Williams, J., B. Shuman and T. Webb III. 2001. Dissimilarity analyses of late-quaternary vegetation and climate in eastern North America. Ecology 82:3346-3362. 59 Wood, W. 1634. New England’s Prospect. Tho. Cotes, London, England. [Reprinted in 1865 by John Wilson & Son, Boston, Massachusettes]. Zhang, Q., K.S. Pregitzer, and D.D. Reed. 2000. Historical changes in the forests of the Luce District of the Upper Peninsula of Michigan. American Midland Naturalist 143:94- 110. 60 APPENDIX A Table 1 is reproduced here for reference. Table A.1. Species names in historic surveys with the current common and Latin names. Years refer to the survey dataset (1859 = Cowichan Valley, 1864 = Chemainus, 1874 = Saltspring Island). See Appendix A for explanation of naming. Historical names without an accompanying year apply to all three datasets. Species names follow The Illustrated Flora of BC (1998-2001). Historical species nomenclature was determined based on the most likely identity of each species within each dataset. The present-day equivalents of “alder,” “apple,” “arbutus,” “cedar,” “maple,” “oak,” “willow,” and “yew” bearing tree species were easily determined due to the similarity of nomenclature across all historical surveys and in the present. In addition, there is only one native tree species of alder, apple, arbutus, cedar, maple, and oak that grows in the study area. There are several species of willow (Salix spp.) and as a result all willows were identified only to genus. 61 The identity of species in the Pinaceae family (Douglas-fir, Hemlock, and Grand fir) was more difficult, as the nomenclature for these species was different for each survey. “Balsam” (1864 and 1874) was presumed to refer to grand fir (Abies grandis) due to its similarity to “balsam fir” (Abies balsamea) on the eastern coast of North America. This is supported by the occurrence of a few locations in the notes where “balsam fir” is written out in full. In addition, balsam was commonly noted to be growing in the same stands as Douglas-fir, a pattern often observed in the present as well. “Pine” in the 1859 Cowichan Valley surveys was determined to refer both to Douglas-fir and hemlock based on subsequent surveys (in 1864) that re-identified several of these same trees as “Douglas pine” or “hemlock”. “Douglas pine” (Chemainus district, 1864) was determined to be the equivalent of Douglas-fir based on the similarity of the names and the dominance of Douglas-pine in the landscape. In addition, hemlock and “balsam” (grand fir) were recorded separately, and species of true pine are rare in this area. On Saltspring Island (1874), “fir” was determined to be the historical name for Douglas-fir, again based on the abundance of this species in the landscape, the presence of hemlock and balsam described separately in the notes, and the absence of any other pine species in the landscape. 62 APPENDIX B Table B.1. Transition matrix showing proportions of sites where the species of the closest tree remained the same or changed from past to present across all datasets. “Pine” includes both Douglas-fir and hemlock. “None” indicates that there were no trees within 26 meters of the point. The highlighted row shows the proportion of bearing trees designated as “pine” in the presettlement surveys that were still “pine” in the 2007 surveys (50%) or changed to cedar (18%), maple (13%) or other species (18%). The total number of trees within each category is shown in the last column. This indicates that, as expected, presettlement Douglas-fir trees have shifted toward cedar and maple in the present. All rows sum to one; total numbers at the end of each row are provided for reference. Figures B.1 and B.2 are a visual representation of this table. 63 Table B.2. Transition matrix showing changes in bearing tree composition at each site, categorized by fire resistance. Species of low fire tolerance include apple, arbutus, cedar, and willow; species of medium tolerance include alder, maple and grand fir; species of high fire tolerance include Douglas-fir and oak. Bold numbers indicate sites which have changed from a species of high fire tolerance to a species of lower fire tolerance. All rows sum to 1. 64 Figures B.1 and B.2 are a visual representation of Table B.1. Figure B.1. Map showing the species of bearing tree at each survey point, 1859-1874. 65 Figure B.2. Map showing the species of bearing tree (the single closest tree) at each survey point, 2007. "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2009-05"@en ; edm:isShownAt "10.14288/1.0066874"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Botany"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivatives 4.0 International"@en ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/4.0/"@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Changes in the landscape and vegetation of southeastern Vancouver Island and Saltspring Island, Canada since European settlement"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/3421"@en .