UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Natural disturbance and climate variability in the dry cool sub-boreal spruce ecosystem of the central… Campbell, Kirstin Anne 2006

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2007-266915.pdf [ 14.04MB ]
Metadata
JSON: 831-1.0074961.json
JSON-LD: 831-1.0074961-ld.json
RDF/XML (Pretty): 831-1.0074961-rdf.xml
RDF/JSON: 831-1.0074961-rdf.json
Turtle: 831-1.0074961-turtle.txt
N-Triples: 831-1.0074961-rdf-ntriples.txt
Original Record: 831-1.0074961-source.json
Full Text
831-1.0074961-fulltext.txt
Citation
831-1.0074961.ris

Full Text

N A T U R A L D I S T U R B A N C E A N D C L I M A T E V A R I A B I L I T Y IN T H E D R Y C O O L S U B - B O R E A L S P R U C E E C O S Y S T E M OF T H E C E N T R A L I N T E R I O R OF BRITISH C O L U M B I A by KLIRSTIN A N N E C A M P B E L L M . S c , University of Northern British Columbia, 2002 B.Sc . , University of Victoria, 1998 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E STUDIES (Forestry) T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A D E C E M B E R 2006 © Kirstin Anne Campbell, 2006 I A B S T R A C T This research presents an evaluation of current sustainable forest management policies in British Columbia within the context of past, current and future disturbance regimes, and including the possibility of a changing climate, for a 1 million ha study area in the dry cool sub-boreal spruce ecosystem in the central interior of the province. In British Columbia, policy mandates that management should be consistent with the temporal patterns of natural disturbance processes. This research critically assesses such a policy from a number o f perspectives, including an historical analysis, a modelling exercise of future conditions, and an examination of the relationship between climate and disturbances. Spatial disturbance data for fires and mountain pine beetle (1904-2004) were combined with vegetation data to create a comprehensive database o f historical disturbances and landscape conditions in the study area; these data were used in all subsequent analyses. The historical and current range in variability in natural disturbances was characterized in order to make recommendations of refinements in B C ' s sustainable forest management policies. Compared to the time period 1904-1953, disturbance frequency was lower, return intervals were longer, and early serai stage forests have declined in more recent time periods (1954-2004). In most cases the current landscape does not meet the policy guidelines. There is a need, however, to make these policies species-specific rather than ecosystem-specific. A forest management model was used to examine the potential impact on the landscape o f harvesting forests at the same rate as historical disturbances. Only after 130 years o f such a harvesting or disturbance regime, did the landscape return to the historical range o f variability in condition. The relationship between annual area disturbed and the Pacific Decadal Oscillation, temperature and precipitation) climate variables over an 89-year time period were modelled using Poisson regression. Although broad-scale variables such as temperature and precipitation were important, so was the multidecadal cycling of finer scale processes such as the Pacific Decadal Oscillation. The main conclusion of this dissertation is that any analysis o f the historical range o f variability in disturbances, especially i f it is being used to guide management policy, must include the effects o f climate on disturbances. T A B L E O F C O N T E N T S ABSTRACT ii TABLE OF CONTENTS iii LJST OF TABLES : vii LIST OF FIGURES viii ACKNOWLEDGEMENTS x CO-AUTHORSHIP STATEMENT xi CHAPTER 1 INTRODUCTION, LITERATURE REVIEW, STUDY AREA & DATA 1 1.1 Introduction 1 1.2 Literature Review 2 1.2.1 Range of variability and application to SFM 3 1.2.2 A review of the definition of disturbance, and descriptors of disturbance regimes ...6 1.2.3 Climate change, range of variability and disturbance regimes 13 1.3 Description of Study Area 14 1.3.1 Environment and vegetation 15 1.3.2 Natural disturbance 15 1.4 Description of Data 16 1.4.1 Disturbance data 17 1.5 Tables 19 1.6 Figures 21 1.7 References 22 CHAPTER 2 RANGE OF VARIABILITY IN DISTURBANCES IN A SUB-BOREAL ECOSYSTEM IN BRITISH COLUMBIA, WITH IMPLICATIONS FOR POLICY 30 2.1 Introduction 30 2.2 Methods 31 2.2.1 Study area description 31 2.2.2 Characterize the RV of disturbances using indicators within the past 100 years 33 2.2.3 Evaluate indicators changes in time, and compare with SFM policy 38 2.3 Results 38 2.3.1 Historical time period 38 2.3.2 Comparison of indicators over time 39 2.3.3 Comparison with policy 41 2.4 Discussion 42 2.4.1 Historical time period , 42 2.4.2 Changes in disturbance over 100 years 42 2.4.3 Range of variability and management 44 2.4.4 Fire suppression 45 2.4.5 Assumptions of this study and future improvements 45 2.4.6 Implications for management 46 2.5 Conclusion 47 2.6 Tables 48 2.7 Figures 50 2.8 References 62 C H A P T E R 3 A HIERARCHICAL SIMULATION-THROUGH-OPTIMIZATION A P P R O A C H TO FOREST DISTURBANCE M O D E L L I N G 68 3.1 Introduction 68 3.2 Materials and Methods 70 3.2.1 Study area : 70 3.2.2 The model 70 3.2.3 Parameterization of the model 72 3.2.4 Scenarios 76 3.3 Results 78 3.3.1 Scenario 1, forest level sensitivity analysis 78 3.3.2 Scenario 2, forest level historical disturbance targets 79 3.3.3 Scenario 3, habitat type level historical disturbance Targets 79 3.3.4 Scenario 4, landscape unit level historical disturbance Targets 80 3.3.5 Comparison of scenarios 2, 3, & 4 80 3.4 Discussion 81 3.4.1 Assessment of the model 81 3.4.2 Implications of modelling approach 83 3.4.3 Scenarios 85 3.4.4 Implications for management 87 3.4.5 Future direction 89 3.5 Tables 90 3.6 Figures 96 3.7 References 100 iv CHAPTER 4 INTERACTIONS OF BROAD-SCALE CLIMATE PATTERNS AND DISTURBANCE IN SUB-BOREAL FORESTS 104 4.1 Introduction 104 4.1.1 A review of climate impacts at multiple temporal scales : 104 4.1.2 A review of short-term cycling mechanisms, Pacific Decadal Oscillation 105 4.1.3 Examples of climate impacts at multiple scales 105 4.1.4 Context and objectives 106 4.2 Methods 107 4.2.1 Study area description 107 4.2.2 Disturbance data 107 4.2.3 Climate data 108 4.2.4 Statistical model 109 4.2.5 Predicting mountain pine beetle disturbance with the statistical model I l l 4.3 Results I l l 4.3.1 Summary of climate variability 1910-1999 112 4.3.2 Statistical model 112 4.3.3 Predicting mountain pine beetle disturbance .- 115 4.4 Discussion 115 4.4.1 Effect of short-term climate cycling- Pacific Decadal Oscillation 115 4.4.2 Extreme disturbance events 117 4.4.3 Fire suppression and climate impacts 118 4.4.4 Mountain pine beetle 118 4.4.5 Conclusions and implications 119 4.5 Tables 121 4.6 Figures 122 4.7 References 132 CHAPTER 5 DATA ACCURACY 137 5.1 Stand Age 138 5.2 Cover Type 140 5.3 Mountain Pine Beetle Data 140 5.4 Fire Data 140 5.5 Recommendations Regarding Error 141 5.6 Tables 143 v 5.7 Figures 144 CHAPTER 6 SUSTAINABLE FOREST MANAGEMENT, NATURAL DISTURBANCES AND CLIMATE CHANGE 145 6.1 Introduction 145 6.2 Summary of Methods 146 6.3 Key Findings 147 6.4 Concluding Remarks 150 vi LIST OF T A B L E S Table 1-1 Biogeoclimatic subzone and variant environmental and vegetation characteristics 19 Table 1-2 Source and description o f spatial and non-spatial data 20 Table 2-1 Description of the strata used in this study 48 Table 2-2 Description of indicators used to characterize the R V in disturbance 49 Table 3-1 Description and function o f the five required data layers in model 90 Table 3-2 Description of the data layer values used in this study 90 Table 3-3 Indicators and their attributes used in this study 91 Table 3-4 Disturbance targets calculated from historical wildfire disturbance rates 91 Table 3-5 Comparison of indicator values between Scenarios 2, 3, & 4 93 Table 3-6 Mean (SE) indicator values at time period 15 (150 years) for Scenarios 2-4, and the % change in the serai stage indicators from initial conditions 95 Table 4-1 Climate variable abbreviations and description 121 Table 5-1 Error quantification for stand age, cover type and mountain pine beetle data. 143 Table 5-2 Error quantification for fire 143 vii L I S T OF F I G U R E S Figure 1-1 Location o f study area in B C . 21 Figure 2-1 Location o f study area in B C 50 Figure 2-2 Cumulative % of study area by year of last disturbance 51 Figure 2-3 Percent of the study area by cover type for the current landscape 52 Figure 2-4 Percent o f the study area by habitat types for the current landscape 52 Figure 2-5 Distribution of cover types amongst the 4 vegetated habitat types for the current landscape 53 Figure 2-6 Percent area disturbed historically (1904-1953) and currently (1954-2004) ..54 Figure 2-7 Cause o f disturbances historically (1904-1953) and currently (1954-2004)... 54 Figure 2-8 Distribution of % area disturbed over time, by disturbance type 55 Figure 2-9 Percent area disturbed (y axis) by cover type (x axis) 1904-1953, and 1954-2004 55 Figure 2-10 Percent area disturbed (y axis) by habitat type (x axis) 1904-1953, and 1954-2004 56 Figure 2-11 Average decadal disturbance frequencies (a) and average decadal return intervals (b) by cover type for the two time periods 57 Figure 2-12 Average decadal disturbance return intervals and frequencies by habitat type for the two time periods 58 Figure 2-13 Average annual disturbance return intervals by cover type over time 59 Figure 2-14 Area o f each cover type by disturbance type 59 Figure 2-15 Percent o f the area of each cover type in early serai stage 60 Figure 2-16 Policy recommendations for % area in early serai stage 61 Figure 3-1 Location o f study area 96 Figure 3-2 A n example o f zoning at the a) habitat type and b) landscape unit level 97 Figure 3-3 Disturbance (a) and mean cumulative disturbance (with standard error bars) (b) indicators at 9 different disturbance rates, for Scenario 1 97 Figure 3-4 Mean area (ha) in early (a), young (b), mature (c), and old (d) serai stage indicators (with standard error bars) at 9 different disturbance rates, for 3 runs of Scenario 1 98 viii Figure 3-5 Mean age class distribution, presented as cumulative % of the study area, for Scenario 2, at 4 o f the planning periods in the analysis 98 Figure 3-6 Comparison of forest mean disturbance (a) and serai stage (b) indicator values between scenarios 99 Figure 3-7 Comparison of the forest age class distribution between scenarios 99 Figure 4-1 Location o f study area in B C 122 Figure 4-2 Histogram o f annual area disturbed ( A A D ) 123 Figure 4-3 Lognormal probability plot o f annual area disturbed 124 Figure 4-4 Annual area disturbed (ha) 1910-1999 125 Figure 4-5 Mean P D O index by season, 1910-1999 126 Figure 4-6 Scatterplots o f broad-scale climate variables 127 Figure 4-7 Spring and Summer P D O index and the natural log o f area disturbed (ha).. 128 Figure 4-8 Crosscorrelation of annual area disturbed and a) summer P D O index, and b) spring P D O index 128 Figure 4-9 Autocorrelation of annual area disturbed over time 129 Figure 4-10 Actual disturbance levels vs predicted disturbance levels 130 Figure 4-11 a) Quantile residuals plotted against fitted model values, b) Normal probability plot o f quantile residuals 130 Figure 4-12 Actual disturbance, primarily mountain pine beetle, versus predicted disturbance from the Poisson regression model 131 Figure 5-1 % area disturbed, by decade for each disturbance type 144 rx A C K N O W L E D G E M E N T S I would like to gratefully acknowledge the support of my two primary supervisors, Bruce Larson and Steve Dewhurst. With endless patience and many long conversations they helped me see the path through this dissertation work. By providing editorial comments, they also improved the writing and thoughts expressed here. Thanks in particular to Steve Dewhurst who gave a substantial amount of time and energy to teach me about forest modelling. I appreciate the feedback and discussions from my other committee members. Thanks to Chris Hawkins for guiding me in the world o f forest management. Also , thank-you to Sarah Gergel for sharing her knowledge of landscape ecology. Steve Taylor o f the Canadian Forest Service kindly provided the Natural Disturbance Database used throughout this research. Funding for this research was provided by the National Sciences and Engineering Research Council, the Canada-US Fulbright Program, the Ecological Restoration Institute at Northern Arizona University, the University o f British Columbia, and the Canadian Forest Service. A l l this would not have been possible without the endless support and encouragement from my mentors, family and friends. The list is too long; I am a very lucky person to have you all. x C O - A U T H O R S H I P S T A T E M E N T The co-authors (Steve Dewhurst and Bruce Larson) listed on all three manuscripts (Chapters 2 through 4) contributed by helping to identify the research problem, the design of the study, and the overall management implications. A l l co-authors also provided editorial support. The model described in Chapter 3 was developed by Steve Dewhurst (outside o f this research) and modified for use in this research. Otherwise, I completed the work described in this dissertation. This includes data collection and preparation, performing the analyses, and writing the manuscripts. Chapter 3 has been accepted into Ecological Modelling. This journal requires that this manuscript must not be published elsewhere, with the exception o f being included in a dissertation. xi CHAPTER 1 INTRODUCTION, LITERATURE REVIEW, STUDY AREA & DATA 1.1 I N T R O D U C T I O N Increasingly Canada, and British Columbia specifically, is adopting a policy of sustainable forest management which attempts to reconcile our economic objectives for the forest, with those inherent processes and patterns which, when maintained, provide for the health and full functioning of the forested ecosystem over time and space. Commonly, research into sustainable forest management (SFM) focuses, in part, on developing an understanding of forest dynamics, including the destructive (such as fire and insect attack) (Delong & Kessler, 2000; Wei et al., 2003) and regenerative processes (Frohlich & Quednau, 1995; Fitzsimmons, 2003). Limited work also addresses the effects of environmental factors such as climate on landscape level forest dynamics (Delong, 1998; Zhang et al., 2000). Research demonstrating the variability o f ecosystem patterns and processes historically, even during periods o f relatively uniform climate conditions, influences our knowledge of disturbance and forest management (Bergeron et al., 2002). Over the past 100 years, forest management has altered landscape processes such as fire, pest outbreaks, and species composition. Research shows that fire suppression and harvesting changes forest species composition (Ward, 2001; McGregor, 2002; Rhemtulla et al., 2002), while harvesting also fragments habitat, causing loss of connectivity and interior habitat conditions (Sachs et al., 1998; Tinker et a l , 2003). Quite often, current policies of sustainable forest management seek to rectify the damage by past management activities by tailoring activities more closely to the observed patterns o f historical conditions. For example, many researchers believe that an evaluation of the range of variability in landscape patterns (distribution of age classes, serai stages, species assemblages, etc.) wi l l allow the emulation of similar patterns with silvicultural and other management activities, including restoration (Landres et al., 1999; Haeussler et al., 2002; Wong & Iverson, 2004). This, they state, is important for biodiversity, forest productivity, and wildlife habitat because species are adapted to variability in processes such as fire disturbance, as well as the range o f habitat created by such processes (Delong, 2002b; Kuuluvainen, 2002). In British Columbia (BC), management guidelines specifically state that areas "on which timber harvesting is to be carried out [should] resemble, both spatially and temporally, the patterns of natural disturbance that occur within the landscape" ( B C Ministry o f Forests and Range, 2004). 1 It is also recognized that natural systems are fundamentally stochastic at multiple time and spatial scales however, from the invasion of new species after glaciation, to the regeneration of single trees in gaps created by tree mortality in a dense stand. In particular, global and regional level climate patterns shape landscape patterns and are drivers for broad scale forest processes such as disturbance (Wu & Loucks, 1995; Veblen, 2003). Indeed, climate effects on the landscape appear to be most evident in the regenerative stages and the current mix and distribution o f species reflects the historical influence o f climate on the landscape, rather than current or static climate conditions (Millar & Woolfenden, 1999). Thus, it is also becomes necessary to consider the historic effects of climate variation on landscape patterns (Fule et al., 2002) in sustainable forest management in order to understand the impacts on our future forests, and how this wi l l affect a management policy rooted in the historic conditions o f the past. The purpose of this research is to evaluate and suggest possible refinements to current sustainable management policies within the context o f past, current and future disturbance regimes, and the possibility of a changing climate. The study area is a sub-boreal ecosystem in the central interior of British Columbia which is over 1 million hectares in size. Chapter 2 describes both the historical and current range of variability in disturbances in the study area, and evaluates current S F M policy within the context of this disturbance variability. Chapter 3 models theoretical future forest conditions under our current management policy where harvesting must emulate natural disturbances. Chapter 4 looks at the relationships between historical disturbances and broad and fine-scale climate variables, and the impacts of the results on future management policies. Chapter 5 quantifies the data accuracy for all datasets used in this dissertation. Finally, Chapter 6 outlines the overall findings, recommendations, and conclusions o f this work. The remainder of Chapter 1 is organized as follows: literature review, description o f study area, and data summary. 1.2 L I T E R A T U R E R E V I E W Three broad concepts are reviewed in this next section: 1) range of variability and its application to sustainable forest management, 2) a review of the definition of disturbance, and descriptors o f disturbance regimes, and 3) the effects of climate on range of variability and disturbance regimes. Literature specific to each chapter is also reviewed in the Introduction to individual chapters themselves. 2 1.2.1 Range of variability and application to S F M 1.2.1.1 Definition and discussion o f the concept of range o f variability Researchers continually struggle to characterize landscape dynamics in a manner so as to be realistic, while maintaining the ability to test hypotheses about the underlying processes affecting a system (whether in space or time). In particular, processes, which cause changes on the landscape, are certainly a dominant research focus (see for example Weir & Johnson (2000) and Tinker et al. (2003)) as we try to study ecosystems from multiple perspectives including conservation o f biodiversity, sustainable forest management practices, and socio-economic needs. We now recognize that disturbance at the landscape level is an agent o f change in all ecosystem types, but can vary with time. Traditionally, however, natural disturbances were thought to be separate from the ecological system, having a negligible role in shaping ecosystem structure and function (Chesson & Case, 1986; Grimm & Wissel, 1997; Perry, 2002; W u & Loucks, 1995). Since disturbance was not a fundamental cause of change on the landscape, scientists believed that ecosystems developed a "permanence o f structure" as a self-perpetuating climax condition (Wu & Loucks, 1995). The "climax" was largely viewed as highly stable, increasing with successional progression (Perry, 2002). Once scientists recognized the role o f disturbance in ecosystem function, nature was then viewed as a balance of "destructive and constructive forces" (Wu & Loucks, 1995). Stability remained an inherent part of ecosystems, however, because it had the ability "to return to an equilibrium state after a temporary disturbance," p. 17 (Holling, 1973). Even disturbance patterns were thought to be predictable, despite their obvious stochastic nature. Over a long enough time period, depending on the disturbance type, disturbances were thought to be predictable, at least on average (Perry, 2002). Recognition that ecological systems and processes, such as disturbance, are not deterministic has shifted our understanding of landscape dynamics. However, this recognition is not necessarily new; Drury and Nisbet (1973) were one o f the first to criticize traditional paradigms of succession, demonstrating nonequilibrium in many ecosystem types. A major concept that currently reflects our changing viewpoint today is that of natural or historic range of variability. There are advantages and disadvantages in this concept, but it is a framework for studying ecological systems that has proved useful. 3 The concept o f natural range o f variability (RV) and historical ecology are reviewed in depth by Landres et al. (1999) and Swetnam et al. (1999). These papers are often referenced in any published work on the subject, possibly because they were the first to summarize and clarify many confusing issues in an emerging field. Since this time there have been many publications studying the variability of natural systems and how this variability might have shaped the landscape in the past (Fule et al., 2002; Rhemtulla et al. , 2002; Tinker et al., 2003). Confusion over terminology still seems to be a formidable obstacle to overcome when using the technique o f R V , despite the work of Landres et al. (1999) to provide a common structure for research. Landres et al. (1999) define natural range o f variability as: "the ecological conditions, and the spatial and temporal variation in these conditions, that are relatively unaffected by people, within a period of time and geographical area appropriate to an expressed goal" (p. 1180). In British Columbia on the other hand, Wong and Iverson (2004) have suggested the following: "the temporal and spatial distribution of ecological processes and structures prior to European settlement o f North America" (p.2). Swetnam et al. (1999) quote Aldo Leopold: " A science of land health needs, first of all, a base datum of normality, a picture of how healthy land maintains itself as an organism" (p. 1189). There is only one constant in the three definitions referenced above: R V is the study of spatial and temporal variability of ecosystems within a time period that is considered free from unusual perturbations that would affect ecosystem health and function. Most often, scientists have viewed human activities as the cause for unusual disturbances and changes to the ecosystem (Noss et al., 1995), but severe or rapid climate changes could also have strong impacts (Millar & Woolfenden, 1999). Whether or not to include Aboriginal influences, and how to define the time period o f study can only be determined by the objectives of the research. There are several issues that require clarification about R V . First is the use o f the word "natural" rather than "historical" (as in historical range o f variability), which is often used interchangeably in the literature. In this work, I intentionally avoid the use o f either word to avoid confusion. The term historical implies that the conditions being measured must have occurred in the past. In the case where there are areas unaffected by human interventions (possibly parks if disturbances are not excluded), these may serve as a basis for characterizing the R V of a nearby ecosystem (Landres et a l , 1999). There is ongoing debate over whether humans are in fact part o f the natural world (and thus we should not exclude their activities when looking at R V ) (Purdon, 2003). The purpose o f R V , I believe, is analogous to that o f a control sample in a laboratory experiment. It is the act o f simplifying the problem in order to understand 4 processes which might be masked by influences not of interest in the study (Millar & Woolfenden, 1999). For example, in studies o f the effects of fire regimes on landscape structure, it is preferable to carry out the studies on a landscape without timber harvesting or fire suppression than one with those management activities. The second issue is the use of R V for management prescriptions, whether for restoration or other objectives. In this context, R V is a useful as guide, not as a deterministic prescription for on the ground conditions (Swetnam et al., 1999; Rhemtulla et al., 2002; Tinker et al., 2003). The goal of R V is not to necessarily recreate conditions (historical or otherwise) on the landscape, but rather, to understand and characterize the mechanisms that caused change from previous or natural conditions to those o f today (Landres et al., 1999). This includes the recognition that the landscape is in nonequilibrium and random chance events do occur (Fule et al., 2002). Since this is a coarse filter approach to management (Landres et al., 1999), it cannot be expected to account for all the variability in a system at all temporal or spatial scales. Wong and Iverson (2004) contend that R V is still the "best available model for maintaining conditions to which most species are adapted" (p. 2). Perhaps this statement could be adopted in modified form to read, "it is a good model for understanding the conditions to which most species are adapted, and how these species will respond to changes in these conditions." 1.2.1.2 Examples of R V Several papers have been published by E . A . Johnson from the University of Calgary, and his colleagues, on fire regimes in the boreal forest o f Canada (Johnson, 1992; Johnson et al., 2001; Weir & Johnson, 2000). Weir et al. (2000) created a time since fire map for the Prince Albert National Park in Saskatchewan. This, in combination with fire frequency data was used to characterize long-term fire regimes. Using distribution-fitting techniques, the authors have identified the R V in fire regimes, within 3 distinct climate and land use periods. Prior to 1890, the study area experienced very short fire cycles (time it takes to burn an area the size o f the study area at least once) o f 15-25 years. Even though the cool and moist climate conditions of the Little Ice Age prevailed, there was significant lightning activity in the region, resulting in many fires. In the northern portion of the park between 1890-1945 the fire cycle increased to 75 years, probably due to less frequent lightning activity. Vegetation would have responded accordingly, no longer favouring species more adapted to shorter fire cycles, and shifting the age 5 distribution into more mature classes. After 1945, the fire cycle may have continued to increase, but the variance in the data was too high to make conclusions. Nevertheless, a shift in state from conditions pre-1890 to conditions post-1890 was evident in the north. The shift was attributed to changing climate conditions consistent with the end of the Little Ice Age (and hence was a parameter shift). Fire cycles in the south portion o f the park did not change significantly until 1945. Even though the climate conditions were changing, fire cycles remained between 15-25 years. In the early 1900's settlement began in the south on the fringe of the park boundary. Agriculture land clearing by homesteaders was largely responsible for fire activity. Large debris piles were burned, and these fires often escaped northward into the park. In this part of the park, therefore, even though climate had changed, a state change did not occur until after 1945 when land clearing activities stopped. As another example, the R V o f spruce budworm populations in eastern Canada is described by Holling (1973) and others. This species persists in an ecosystem at either endemic or epidemic levels in the ecosystem, depending on many factors, including predation, availability o f host species (primarily balsam fir; Abies lasiocarpa (Hook.) Nutt.), and climate conditions (Holling, 1973; Peterson et al., 1998). During drier climate periods, the budworm population expands, and, i f mature hosts are available and predation is sufficiently low, can increase to epidemic proportions. Once hosts are no longer available due to mortality the population collapses to a lower population state. Persistence in this case, is maintained by the resilience of the population, even though it appears to be highly unstable (Holling, 1973). 1.2.2 A review of the definition of disturbance, and descriptors of disturbance regimes Increasingly, emulating the natural disturbance regimes o f forested landscapes are being championed as a method to both sustain biodiversity and practice harvesting in an ecological manner, thereby achieving sustainable forest management (Boychuck et al., 1997; Delong, 2002b; Mitchell et al., 2002; Wei et al., 2003). By emulating or mimicking natural disturbances using silviculture activities, it is presumed that the structural variability o f an ecosystem is maintained within the adaptive limits o f the species in that ecosystem (Bergeron et al., 2002). The natural disturbance regimes which are used as reference points are generally characterized from some point in time or space where human influences were either relatively small or absent; this is the concept o f natural range of variability (Landres et al., 1999). 6 There is, however, considerable debate over what a disturbance is, and how a disturbance regime can be quantified (see for example the numerous definitions and uses of the term disturbance as outlined in Laska (2001)). Definitions of disturbance can depend on how nature is perceived, either as a balance of death and renewal processes (equilibrium paradigm), as a continually changing and shifting entity (nonequilibrium paradigm) (Wu & Loucks, 1995), or as a mixture of unstable and stable phenomenon, highly dependent on spatial and temporal scale (Turner et al., 1993). 1.2.2.1 Definition o f disturbance Over the past 30 years, our notion of what constitutes a disturbance has gone through various stages of development. Laska (2001) provides a comprehensive review of the subject, citing 13 "exemplary" definitions o f disturbance since 1977. Today, most foresters and ecologists recognize that disturbance is part of the natural cycle, no longer viewing ecosystems as constant in space and time (Wu and Loucks 1995). Issues of terminology now appear rooted in the notion o f scale, in both space and time, or as Laska (2001) contended, differences between the reductionist and holistic approach. Two common definitions o f disturbance illustrate this point. Grime (1979) describes natural disturbance as a "process" that causes the "partial or total destruction" of plant biomass (as quoted in Laska (2001)). This is the reductionist point of view, which focuses primarily on individuals or the effects on organisms. Disturbance by this definition occurs at fine spatial scales because it is a process that interrupts competition, excluding possible effects on the abiotic environment. The holistic viewpoint is from White and Pickett (1985) for whom disturbance is "a relatively discrete event in time that disrupts ecosystem, community or population structure and changes resources, substrate availability, or the physical environment." (p. 7). Disturbance by this definition occurs at broader spatial scales affecting not only individual organisms but communities, and ecosystems. Disturbance affects not only the biotic but also the abiotic. In addition, disturbances can be destructive events as well as environmental cycles and changes. Furthermore, the scale at which disturbances occur varies with both space and time. Pickett et al. (1989) define disturbance based on scale: "Disturbance is a change in the minimal structure caused by factors external to the level of interest" (p. 132). Minimal structure is the level of organization under study, whether it is the individual, community or landscape (Laska, 2001). A n y change detectable at this scale caused by external factors can be considered a disturbance by this definition. 7 Many scientists have pointed out that disturbances can be both autogenic (internal) and allogenic (external) (Oliver and Larson 1996). In the equilibrium paradigm disturbances were believed to be purely exogenous (such as a fire which burns independently of vegetation type or age), while ecosystem structure and function was created from endogenous processes and interactions (such as competition promoting succession) (White & Pickett 1985). Quite often, however, autogenesis and allogenesis are not easily distinguishable, as in the case where vegetation condition, structure, health, type etc. determines its susceptibility to a particular type o f disturbance (Oliver and Larson 1996). Definitions of disturbance which ignore this potential feedback and dependency between the ecosystem and the disturbance itself, such as that proposed by Pickett et al. (1989), may not be accurate, except in the case where it can be proven to be true. I have chosen, in the remainder o f the dissertation, to adopt the popular definition of disturbance proposed by White and Pickett, now 20 years ago: "a relatively discrete event in time that disrupts ecosystem, community or population structure and changes resources, substrate availability, or the physical environment." Although very broad, such a definition is applicable to events which may shape the landscape at a variety o f spatial and temporal scales, from individual tree mortality, to large crown fires. Climate, by this definition, is itself a disturbance agent, as cycles and long-term changes impact both the biotic (e.g., windthrow in a forest) and abiotic (e.g., lightning strikes causing fires). Consistent with these authors, and with Pickett et al. (1989), any study of disturbance and disturbance regimes must be within a specific context with respect to scale. This will be discussed further below. 1.2.2.2 Definition o f a disturbance regime A disturbance regime can describe both the disturbance and its inevitable impact on the landscape. It is the "temporal and spatial pattern of creation of open or altered patches" (White & Pickett, 1985). A landscape can be described as a "space-time mosaic" of patches, created by changes at the patch level, such as death o f species (disturbance) and subsequent replacement by the same or other species over time (Watt, 1947). The concept o f a disturbance regime thus becomes a framework for describing and, i f possible, quantifying disturbances and associated changes brought about by disturbance through time and space. A patch can be defined as a cohesive spatial unit created by disturbance and subsequent vegetation development after disturbance (Watt, 1947). Because disturbance is a process which creates changes in space, the minimum unit of disturbance has long since been recognized as the 8 patch (Watt, 1947; White & Pickett, 1985; Jentsch et al., 2002). A patch is not spatially independent and is affected by events in neighbouring patches (White & Pickett, 1985). Multiple patches thus define the heterogeneity or patchiness of the landscape (Kuuluvainen, 2002). This patchiness varies spatially depending on the disturbance and possible interactions with vegetation, and also varies with time (White & Pickett, 1985). A patch is therefore analogous to grain as used in landscape ecology (O'Neill & Smith, 2002) or minimum mapping unit, and is one component of spatial scale. Multiple patches define the extent of a particular disturbance under consideration (Pickett et al., 1989; Jentsch et al., 2002). Extent is another component of spatial scale (O'Neill & Smith, 2002) and delineates the boundary of the area under consideration. This is the same as minimal structure proposed by Pickett et al. (1989). 1.2.2.3 Descriptors of disturbance regimes- frequency Frequency is a measure o f how often a disturbance reoccurs within a given area o f interest. Disturbance frequency affects many aspects of the forest ecosystem, including species composition, structure, interactions, and even patch size (Runkle, 1985; Ryan, 2002). Vegetation types can be adapted to a specific disturbance frequency, with species that have a shorter life span and faster, maturity rates more prevalent in areas with high frequencies (e.g., lodgepole pine) and vice versa. The age class distribution o f a forest is also impacted by fire cycles. Higher frequency events may lead to a younger forest composition than less frequent (Weir & Johnson, 2000). In general, more frequent disturbances are smaller than less frequent, although there are exceptions to this (Ryan, 2002). Ponderosa pine (Pinus ponderosa P . & C. Lawson) is an example of a species adapted to a frequent (every 2-10 years), but not intense fire regime (Moore et al., 1999). Ponderosa pine tend to produce abundant seed crops but historically, successful regeneration was only about 1-4 trees per hectare per decade. Frequent surface fires tend to kil l the younger pine, but once established older trees develop thick bark making them much more fire resistant. In the understory, shorter lived herbaceous and shrub vegetation dominates, providing the fuel necessary for frequent fires. Disturbance frequency can be expressed as a value averaged over a certain time period, or as a distribution over time. It can be quantified using a number of methods, including disturbance rate, return interval, and rotation period. Johnson and Gutsell (1994) point out that 9 there can be confusion over how these terms differ, which is problematic not only for the research community, but also for managers applying the results of scientific studies. Survivability or return interval is the time that has elapsed since a disturbance has occurred; it is not the same as a disturbance rate. It can be expressed as the probability that elements in a patch (being the minimum unit o f analysis) survive to the next time period. The inverse of survivability is the disturbance rate which can be expressed as the probability that elements in a patch do not survive to the next time period. The rotation period or disturbance cycle, is how long it takes for an area equal to the area of interest to burn once (Johnson & Gutsell, 1994). Reed (2006) suggests abandoning this latter term however, since in most situations, the disturbance cycle will exceed the return interval. Depending on whether the variable o f interest is survival or mortality as well as the agent o f disturbance, the above 3 measures all represent frequency equally well, but it is how these measures are applied to management that raise concern. Disturbance frequency is often represented by a single value (such as return interval) in the literature when describing specific disturbance regimes, especially for use as guidelines in management (Delong, 1998; Bergeron et al., 2002; Perry, 2002). A s with any statistical summary technique, this presents an estimate (usually average) of the distribution of the actual values in space or time (or both). A mean value can describe the frequency of disturbances for an area, on average, and can aid in the understanding of overall stand or landscape dynamics. For example, i f viewing 3 stands over increasing elevation, the mean fire return interval may also increase, suggesting that lower stands experience disturbances more often than higher elevation stands. Frequency distributions therefore are preferable, i f the goal is to determine management targets, while single values are useful for general comparisons and inferences only. Perry (2002) cautions, however, since the assumption that the distribution itself wi l l remain static in time or even space is erroneous under the nonequilibrium view of ecosystem dynamics. Johnson and Larsen (1991) addressed this problem through the identification of changes in fire frequency patterns in the Canadian Rockies attributable to differences in climate over the past 380 years. Weir et al. (2000) spatially partitioned the landscape, and compared fire frequencies between two different areas, in different temporal periods. In both these studies, the authors fit the data to an exponential distribution model which assisted them in identifying patterns in the data. 10 1.2.2.4 Descriptors o f disturbance regimes- spatial patterns The spatial pattern created by disturbance can be described in a number of ways. Most commonly, extent is used when characterizing a disturbance regime, and is a measure o f the area disturbed, expressed as percent, and/or change in time or space (Runkle, 1985). The spatial distribution of patch sizes created by disturbance is also applied to the study of spatial patterns (Delong, 1998; Bergeron et al., 2002). Less commonly, scientists use metrics to quantify the shape o f the disturbance such as circularity, and compactness (see, for example Weir & Johnson, 2000 who used metrics from McGarigal and Marks, 1994). The heterogeneity of vegetation, local topography, micro environment (e.g., soil moisture), and weather patterns influence disturbance extent (Runkle, 1985; Ryan, 2002). And, in . the case of some disturbances such as insects and fungus which can spread from previous year's infestation, so does the location o f past disturbances control current disturbance extent. Disturbance extent itself affects regeneration, depending on the reproductive strategy o f the species and the severity or intensity of the disturbance (Oliver & Larson, 1996). Species that, have highly motile seed would be at an advantage, for example, to recolonize a large disturbance over those species with limited seed dispersal range, and regeneration by wind blown seed generally decreases towards the interior o f the disturbed patch. Patch size, as created by disturbance, is an important factor for wildlife habitat (Delong, 1998). Wildlife species are adapted to use the range o f patch sizes created by natural disturbances; some are able to utilize the entire range, while others are dependent on a narrower niche (Ries et al., 2004). Depending on the purpose and spatiotemporal scale o f the study, different data, or combinations o f data, typically can be used to determine extent, including dendroecology (Zhang et a l , 1999), remotely sensed data (Millward & Kraft, 2004), forest inventory data (Delong, 1998), and historical maps (Shinneman & Baker, 1997). Data can be continuous (as in the dendroecological study of Zhang et al. (1999)) or discrete (as in the historical study of Shinneman and Baker (1997)). Disturbances generally create patches within a range of sizes within the landscape, from very small, to very large, governed individually by the processes which control spatial extent. This spatial distribution o f patches is an important part of the disturbance regime, and has important implications not only for habitat, but also for development of management strategies (Delong, 1998; Bergeron et a l , 2002). In looking at the spatial distribution of patches created by stand-replacing fires, for example, many small patches are created, while there may only be a few large patches. Bergeron et al. (2002), for example, found that 85% of fires in a Quebec 11 forest were less than 100 ha, but that those fires over 20,000 ha burned 40% of the landscape. Delong (1998) studied historical disturbances in the northern B C , and concluded that forest policy in the region failed to incorporate the full range of patch sizes that result from disturbances. Large patches, especially those 1000 ha or greater, occurred historically as a result of big fires. However, because the public would accept clearcut harvesting in patches of this size, policy recommended a maximum patch size below the 1000 ha. Disturbance shape is affected by several factors including vegetation, topography and weather (Ryan, 2002). Measures of spatial configuration are increasingly being included in studies o f disturbance regimes as spatial analysis techniques and tools are more readily available. Weir et al. (2000) used four shape variables (form ratio, circularity, compactness, and radius ratio) to describe historical fire shape as related to climate and land use in the Saskatchewan boreal forest. They found that shorter fire cycles, as caused by land use in the southern part of the study area, produced fires that were larger, and more oblong than fires in the north, which experienced a longer fire cycle and were more circular and compact. In the south, the patterns reflect more overburning than in the north, which is characterized by the circular remnants of burns in the past. 1.2.2.5 Descriptors o f disturbance regimes- magnitude Disturbance magnitude is commonly expressed as either severity or intensity, or both. Severity is a measure of the impact on the ecosystem, including vegetation and abiotic factors such as soil (Runkle, 1985). Intensity is the actual rate o f the force of the disturbance (White & Pickett, 1985) which, for fire, would be the heat released per unit length o f fire front or area over time (Johnson, 1992; Ryan, 2002). Intensity and severity then, are inevitably linked: a high intensity burn would also result in more severe impacts and vice versa (Johnson et al., 2003). The magnitude o f the disturbance is affected by the susceptibility of the vegetation (fuel for fire, host species for insects), terrain (which can limit the spread of disturbance), weather (local weather events such as lightning storms), and climate (long-term trends in weather which might increase or decrease disturbance initiation) (Ryan, 2002). Disturbance magnitude impacts a multitude of ecological processes, including regeneration. Species such as lodgepole pine (Pinus contorta var. latifolia Dougl. ex Loud.), for example, are adapted to a certain magnitude o f disturbance. Lodgepole pine utilizes the mineral soil substrate for regeneration that is created by a fire severe enough to remove the duff or non-organic layers o f the soil. In addition, it has serotinous cones which require heat to open and release the seed, although the required heat is 12 not all that intense (Lotan 1975). Too severe a fire would kill the seeds, but too light might not remove enough of the soil's organic layer. Most of the literature in forestry focuses on severity, probably because it is much easier to measure post-disturbance than intensity which requires experimentation and observation at the time the disturbance is occurring over very small areas (Ryan, 2002). Fireline intensity, for example, can be determined from the rate o f spread, amount of fuel burned, and the heat o f combustion of the fuel (Ryan, 2002). Insect disturbance could be quantified by counting number of individuals per tree, or number of entrance/exit holes. As a compromise, many studies which are post-disturbance substitute measures of severity to infer intensity. Severity is often determined by a measure o f disturbance impact on the ecosystem, and wil l vary depending on the purpose of the study. For fires, severity was traditionally measured as depth o f burn into the soil, and surface charring, but increasingly is measured as the impact on above ground structures as well (Ryan, 2002)! Fire severity is often used to describe differences between surface fires burning the understory, ground fires, which burn the understory and into the soil, and crown fires which ignite the crowns o f overstory vegetation (Oliver & Larson, 1996). For example, one study classified historical fires as moderate severity i f the canopy scorch was over 50% or just under 100% and high severity i f canopy foliage was 100% scorched (Odion et al., 2004). Severity characterizes the way any one disturbance affects the landscape, but, like frequency, is distributed continuously over a range of values when multiple disturbances are considered. As Shinneman and Baker (1997) point out, a disturbance regime dominated by disturbances of one severity such as surface fires, does not preclude disturbances of higher or lower severity, such as crown fires; a disturbance regime can be in nonequilibrium. For example, in a low severity fire regime e, most of the fires are surface fires, but some become moderate fires and a few high severity crown fires (Agee, 1993). 1.2.3 Climate change, range of variability and disturbance regimes Temperatures in North America are projected to.increase 1-3°C over the next 100 years ( IPCC 2002). Within the past century, climate warming has been linked to the alteration of many forest dynamics processes such as: instability in northern species abundance and distribution (Parmesan & Yohe, 2003), an increase in forest fire size and frequency (Warren, 2004), and an escalation in pest outbreaks (Volney & Fleming, 2000). At the landscape level, global and regional level climate is the driver of broad scale forest dynamic processes such as disturbance 13 regimes, and successional pathways (Wu & Loucks, 1995; Veblen, 2003). Many authors have pointed out that with climate change, ecosystems wi l l not respond as a whole, but will respond as individual species. New ecosystems will form through the interactions o f individual species, including new and/or exotic species, and environmental conditions, along with adaptations (Davis, 1986; Davis & Shaw, 2001; Hansen et al., 2001). This new ecosystem is therefore unpredictable, although a wealth of studies have been carried out to try and forecast possible trajectories (e.g., Thomas et al. 2004). The rapidity of the response ecosystems to climate change is o f great concern to most scientists. Climate changes are projected to occur much more quickly in the future than they did in the past, which means the ability of species to adapt or migrate wi l l be compromised (Davis,. 1986). In northern parts o f the continent, climate warrning may be more severe, increasing at a faster rate and to higher temperatures than in the south (Stewart et a l , 1998). Disturbance regimes wi l l be affected, and wil l continue to be primary agents o f ecosystem reorganization (Dale et al., 2001; Hansen et al., 2001). The entire disturbance regime could be affected, including frequency (Weir & Johnson, 2000), spatial patterns (Logan & Powell, 2001), and magnitude (Millar & Woolfenden, 1999). This wi l l open up areas to migration from outside species, and alter successional pathways of the forest (Dale et al., 2001). Most recently, researchers have predicted massive species extinctions as a result o f habitat loss in one form or another (Thomas et al., 2004). The belief is that there is not enough resilience in populations, communities or ecosystems to accommodate such changes within short time periods. Most concerning, however, may be that a change in the ecosystem may not be readily apparent, especially in forests where established trees continue to grow, but not necessarily reproduce under changing climate conditions. This may occur in areas where disturbance frequencies decrease as occurred over in Saskatchewan (Weir et al., 2000). In areas where disturbance frequencies increase or become more severe, shifts in vegetation patterns may occur within a few years, as occurred in New Mexico (Allen and Breshears, 1998). 1.3 D E S C R I P T I O N OF S T U D Y A R E A The study area in central British Columbia is primarily located in the Morice, Lakes, and Prince George Timber Supply Areas (TSAs) , around the traditional territories o f the Cheslatta Carrier Nation, externally delineated by the largest boundary o f the dry cool sub-boreal spruce (SBSdk) biogeoclimatic subzone (Figure 1-1), with a total area of approximately 1.2 million ha. Only land classified as being in the SBSdk subzone was included in the studies described in Chapters 2 and 14 4. However, other biogeoclimatic variants were included in smaller amounts for the study in Chapter 3; these are also shown in Figure 1-1 and described in Table 1-1. These ecosystem types were included in the analysis in Chapter 3 because disturbances often cross ecosystem boundaries, and although not a spatially explicit model, this accounted for the historical likelihood that disturbance would reach into adjacent areas. A primary factor affecting the choice of this area as the site for this research was the fact that, although a significant portion o f the mature pine has been infested by mountain pine beetle (MPB) (Dendroctonus ponderosae), many opportunities exist to manage younger pine stands, stands of other species, and stands o f mixed species, as well as to restore (in the absence o f timely regeneration to adequate stocking levels) MPB-attacked stands. Past disturbance events (both natural and human in origin) have significantly altered landscape patterns, and there are now concerns about the long-term impacts of these disturbances on timber supply, wildlife habitat, recreation, and non-timber resources. 1.3.1 Environment and vegetation The ecosystems in the study area can be broadly defined into two types - dry, warm ecosystems, and wet, cold ecosystems. The most common ecosystem, the SBSdk, and minor ecosystem, the SBSdw3 (the Stuart variant of the dry, warm sub-boreal spruce subzone), are dry ecosystems, occurring at lower elevations in the area, and, because o f a predominance o f disturbance, are characterized by early serai species such as lodgepole pine and trembling aspen (Populus tremuloides Michx. ) . Other ecosystems that tend to be wetter, and colder are the SBSmc2 (Babine variant o f the moist, cold sub-boreal spruce subzone), ESSFmc (moist, cold Engelmann spruce subalpine fir subzone), and E S S F m v l (the Nechako variant of the moist, very cold Engelmann spruce subalpine fir subzone). These ecosystems occur at higher elevations, and have overall fewer disturbances, lending the species composition to longer-lived hybrid spruce (Picea glauca (Moench) Voss x Picea engelmannii Parry ex Engelm.) and balsam fir. The ESSFmc (parkland), E S S F m v l (parkland), and A T are all higher elevation parkland variant types that currently have little suitability for tree growth, except as krummholz (Meidinger and Pojar 1991). 1.3.2 Natural disturbance There are three broad disturbance regimes in the study area (Table 1-1), as defined by the B C Biodiversity Guidebook (BC Ministry o f Forests, 1995) as natural disturbance types. Most of the 15 study area is dominated by frequent stand-initiating events. At higher elevations, disturbances are much less common. In those ecosystems characterized by a frequent, stand-initiating disturbance regime (SBSdk, SBSmc2, SBSdw3) forest fires and insect outbreaks are common stand-replacing disturbance events (Delong, 2002a). Specifically, stand replacing fires occur at a mean interval of 125 years ( B C Ministry of Forests, 1995), disturbing about 1% of the forested area each year (Delong, 2002a). Fires are often very large; in the absence o f European fires greater than 1000 ha were prevalent (Delong, 2002a). The resulting landscape is characterized by a variety o f different serai stage patches, all which are even-aged, except for the oldest. Although stand-replacing fires are the most common disturbance in the area (Delong, 2002a), mountain pine beetle also has played an important role in shaping the landscape. Little is known about the extent of mountain pine beetle infestation historically. Whether this disturbance occurred at cyclic epidemic levels, as is occurring today, or whether infestation was less severe is uncertain. Other disturbance which occur in these ecosystems include: windthrow, pathogens (e.g., root rot), and other pests (e.g., spruce bark beetle) Since human influence, fire suppression and forest management may have significantly altered disturbance regimes in the study (Delong, 2002a), although some authors disagree that fire suppression has significantly altered such regimes (Johnson et al. 2001); this in an ongoing debate which is touched upon in this research. Although stand-replacing disturbances in those ecosystems with infrequent stand-initiating events (ESSFmc, E S S F m v l ) occur over longer intervals, the dominant disturbance agents are similar to those of other ecosystems above, that is, stand-replacing fire and mountain pine beetle (Delong, 2002a). Return intervals range from 200-900 years. Other disturbances such as spruce beetle (Dendroctonus rufipennis) and western balsam bark beetle (Dryocoetes confusus) also occur, especially in the absence of stand replacing disturbances and these ecosystems are usually multi-aged and more open (Delong, 2002a). Alpine tundra and subalpine parkland ecosystem types (AT, ESSFmc parkland) have very short growing seasons. These ecosystems are very susceptible to disturbance damage by grazing animals such as cows (BC.Ministry of Forests, 1995). 1.4 D E S C R I P T I O N OF D A T A The dataset used in Chapters 2-5 consisted of a series o f spatial data layers, combined using GIS (Table 1-2). Data were classified on both an annual time scale (Chapter 3) and a decadal time 16 scale (Chapters 2 and 4) based on the year of disturbance. For Chapter 5, a non-spatial climate dataset, with an annual time step, was created using climate station data, and data on the Pacific Decadal Oscillation (Table 1-2). A n overview of the datasets is provided in this section as well as in the Methods section of individual chapters. 1.4.1 Disturbance data 1.4.1.1 Fire Spatial data on lightning- and human-caused fires were available from the B C Natural Disturbance Database (NDD). This database was the source for both fire and insect disturbances as far back as 1920, and was compiled by Steve Taylor o f the Canadian Forest Service from a number of historical sources'. The data describe the spatial extent of fires, delineated in polygons and available for each year. Both fire ignition types (lightning and human) were included in the analysis for 3 reasons: 1) within the time period o f analysis anthropogenic wildfires have had an impact on forest structure, 2) regardless of ignition sources, stands which caught on fire had the potential to burn, 3) it is possible that in historical times, when fire was considered an unnatural process, some fires were incorrectly attributed to human causes (E. Johnson, University o f Calgary, pers. comm.). Fires were further identified from the current Vegetation Resources Inventory (VRI) data, using age as-a surrogate for disturbance, a method used by Delong 1998 (1998). V R I also contains history information which lists a limited number o f fires, but these were generally in more recent times than the fire historical timeframe of reference. Both Delong (1998) and Weir et al. (2000) state that caution is necessary when using this method to reconstruct fire history in an area because some o f the stands may have more recent fire events which wil l fragment the appearance o f the fires in the age class data. For this research, this caution is well noted. It was necessary to supplement the historical fire data at the turn of the century with the disturbance data derived from age classes, however, because the Natural Disturbance Database currently underestimates the number and perhaps extent of the fires in the early part of the century (B. Hawkes, Canadian Forest Service, pers. comm.). ' Information on the database is available at: http://www.pfc.forestry.ca/fires/disturbance/index e.html 17 1.4.1.2 M P B Spatial M P B data were available from the N D D , as well as from M O F 2003/2004 survey data. Data were the spatial extent of the attacked polygon and were from aerial pest survey data. The N D D data were created by B C Ministry of Forests by flying over the area and having someone sketch the extent of the infestation onto 1:100,000 or 1:125,000 scale maps. Each polygon also has a level o f infestation associated with it, ranging from 0-11% infestation to gray/dead stands. Only severely infested stands (30% or greater attack levels) were used in the analysis because infestation at this attack level was assumed to be a stand-replacing event. A l l studies in this analysis only considered stand-replacing disturbance events to simplify the analysis. 1.4.1.3 Vegetation data The B C Ministry o f Sustainable Resource Management ( M S R M ) supplied the most recent version (2003) o f the Vegetation Resources Inventory (VRI) spatial dataset for the area. The V R I data were used to stratify the landscape into habitat types (defined by slope and leading vegetation type (tree, shrub, and no vegetation), cover types (leading species), and current age. This stratification was hierarchical, with decreasing levels o f detail from habitat types down to age classes. Specifically, each habitat type contains many cover types, and each cover type can have several age classes. In areas affected by mountain pine beetle, the stand age o f lodgepole pine in the inventory was corrected to the year when attack levels were greater than 30%. 1.4.1.4 Climate data Precipitation and temperature data were obtained for the Ft. St. James climate station in B C , from the Environment Canada's Adjusted Historical Canadian Climate Data collection (Environment Canada 2005). These data have been corrected for research purposes from the original historical climate station data. Monthly maximum and minimum temperatures, and monthly total rainfall and snowfall data for the period 1904-2003 were obtained. Monthly Pacific Decadal Oscillation (PDO) Index data for the same time period were downloaded from the University o f Washington, as compiled by Nate Mantua (Mantua and Hare, 2006). 18 1.5 T A B L E S Table 1-1 Biogeoclimatic subzone and variant environmental and vegetation characteristics (from Meidinger and Pojar 1991, and Delong et al. 1993) Biogeoclimatic Mean Mean Mean Mean Mean Elev. Dominant Tree Species Biodiversity Total area Variant Annual Annual Annual Growing Frost- Range Guidebook (ha) of Temp. Precip. Snow Degree- free (m) Natural variant in (°C) -Rain (cm) days Period Disturbance study area (mm) (>5°C) (days) Type S B S d k 2.1 480.6 188.1 1028 70 700- Lodgepole pine 3 1,079,940 1050 Trembling aspen Douglas-fir (Dry sites) SBSmc2 1.5 574.4 237.1 947 116 850- Hybrid spruce 3 175,837 (Babine 1350 Subalpine fir variant) Lodgepole pine (dry sites) E S S F m d 1.1-1.8 514.1- 246.5- 629-801 32-79 950- Hybrid spruce 2,5 48,746 1995.4 1431.0 1800 Subalpine fir Lodgepole pine (dry sites) E S S F m v l 1:1-1.8 514.1- 246.5- 629-801 32-79 1150- Hybrid spruce 2,5 8,994 (Nechako 1995.4 1431.0 1550 Subalpine fir variant) Lodgepole pine (dry sites) SBSdw3 2.6 494.4 204.2 1089 83 750- Lodgepole pine 3 1,468 (Stuart variant) 1100 Douglas-fir Hybrid spruce Subalpine fir (higher elevation) A T -1.8 755.5 551.4 427 21 1000 + n/a 5 207 Table 1-2 Source and description of spatial and non-spatial data. Data Description Source Spatial? Vegetation data Vegetation Resources Inventory (2003) BC Ministry of Forests & Range Yes Fire Natural Disturbance Database (1914-2000) Canadian Forest Service Yes Other historical disturbances Vegetation Resources Inventory (2003) BC Ministry of Forests & Range Yes Mountain pine beetle Natural Disturbance Database (1914-2000) & Aerial flight data (1999 -2005) Canadian Forest Service & BC Ministry of Forests & Range Yes Ecosystem stratification Biogeoclimatic zones (2005) BC Ministry of Forests & Range Yes Climate station data Ft. St. James climate station (1895-2004) Environment Canada No Pacific Decadal Oscillation PDO (1904-2004) University of Washington No 20 1.6 FIGURES 127»ffO"W 126'0"(TW 125 W W 124WW Figure 1-1 Location of study area in BC. The outside boundary of the study area, shown in white, follows the dry cool Sub-boreal Spruce biogeoclimatic zone (SBSdk). Lighter gray areas are ecosystems outside of the study. 21 1.7 R E F E R E N C E S Agee, J. K . (1993). Fire Ecology of Pacific Northwest Forests. Washington, D C , U S A : Island Press. Allen, C D . and Breshears, D . D . 1998. Drought-induced shift of a forest-woodland ecotone: Rapid landscape response to climate variation. Ecology. 95(25): 14839-14842, B C Ministry of Forests. (1995). Forest Practices code of British Columbia, Biodiversity Guidebook. Victoria, B C : B C Ministry of Forests. B C Ministry of Forests and Range. (2004). Forest Range and Practices Act (FRPA). Retrieved M a y 16, 2006 from http://www.for.gov.bc.ca/tasb/legsregs/frpa/frpa/fipatoc.htm Bergeron, Y . , Leduc, A . , Harvey, B . D. , & Gauthier, S. (2002). Natural fire regime: A guide for sustainable management o f the Canadian boreal forest. Silva Fennica. 36(1): 81-95. Boychuck, D . , Perera, A . H . , Ter-Mikaelian, M . T., Martell, D. L . and L i , C . (1997). Modelling the effect of spatial scale and correlated fire disturbances on forest age distribution. Ecological Modelling. 95: 145-64. Chesson, P. L . , & Case, T. J. (1986). Overview: Nonequilibrium community theories: Chance, variability, history and coexistence; community ecology. In J. Diamond, & T. J. Case (Eds.), Community Ecology. N e w York : Harper and R o w Inc. pp. 665. Dale, V . H . , Joyce, L . A . , McNul ty , S., Neilson, R. P., Ayres, M . P., & Flannigan, M . D . et al. (2001). Climate change and forest disturbances. Bioscience. 51(9): 723-734. Davis, M . B . (1986). Climatic instability, time lags, and community disequilibrium. In J. Diamond, & T. J. Case (Eds.), Community Ecology. New York : Harper and Row Inc. pp.665. Davis, M . B . , & Shaw, R. G . (2001). Range shifts and adaptive responses to quaternary climate change. Science, 292, 673-679. 22 Delong, S. C. (2002a). Natural Disturbance Units of the Prince George Forest Region: Guidance for Sustainable Forest Management. Prince George: B C Ministry of Forests and Range. Delong, S. C . (2002b). Using nature's template to best advantage in the Canadian boreal forest. Silva Fennica. 36(1): 401-408. Delong, S. C. (1998). Natural disturbance rate and patch size distribution o f forests in northern British Columbia: Implications for forest management. Northwest Science. 72: 35-48. Delong, S. C , & Kessler, W. B . (2000). Ecological characteristics of mature forest remnants left by wildfire. Forest Ecology and Management. 131: 93-106. Delong, S.C., Tanner, D. , and Jull, M . J . 1993. A field guide for site identification and interpretation for the southwest portion o f the Prince George Forest Region. Land Management Handbook 24. B C Ministry o f Forests Research Branch. Victoria, B C . Retrieved Dec. 7, 2006 from: http://www.for.gov.bc.ca/hfd/pubs/Docs/Lmh/Lmh24.htm Drury, W . H . , & Nisbet, I. C. T. (1973). Succession. Journal of the Arnold Arboretum. 54(3): 331-368. Environment Canada. 2005. Adjusted Historical Canadian Climate Data. Retrieved Apr i l 2006 from: http://www.cccma.bc.ec.gc.ca/hccd/ Fitzsimmons, M . (2003). Effects o f deforestation and reforestation on landscape spatial structure in boreal Saskatchewan, Canada. Forest Ecology and Management. 174: 577-592. Frohlich, M . , & Quednau, H . D . (1995). Statistical analysis of the distribution pattern of natural regeneration in forests. Forest Ecology and Management. 73: 45-57. Fule, P. Z . , Covington, W. W. , Moore, M . M . , Heinlein, T. A . , & Waltz, A . E . M . (2002). Natural variability in forests o f the Grand Canyon, U S A . Journal of Biogeography. 29: 31-47. Grimm, V . , & Wissel, C. (1997). Babel, or the ecological stability discussions: A n inventory and analysis o f terminology and a guide for avoiding confusion. Oecologia. 109: 323-334. 23 Gunderson, L . H . (2000). Ecological resilience - in theory and application. Annual Review of Ecology and Systematics. 31: 425-439. Haeussler, S., Bedford, L . , Leduc, A . , Bergeron, Y . , & Kranabetter, J. M . (2002). Silvicultural disturbance severity and plant communities o f the southern Canadian boreal forest. Silva Fennica. 36(1): 307-327. Hansen, A . J., Neilson, R. P., Dale, V . H . , Flather, C . H . , Iverson, L . R. , & Currie, D . J. et al. (2001). Global change in forests: Responses of species, communities, and biomes. Bioscience. 51(9): 765-779. Holling, C . S. (1973). Resilience and stability on ecological systems. Annual Review of Ecology and Systematics. 4: 1-23. IPCC. (2002). Climate Change and Biodiversity. Intergovernmental Panel on Climate Change. H Gitay, A Suarez, RT.Watson, DJ Dokken (Eds). Geneva, Switzerland: IPCC. pp 85. Retrieved July 19, 2006 from: http://www.ipcc.ch/pub/techrep.htm Jentsch, A . , Beierkuhnlein, C , & White, P. S. (2002). Scale, the dynamic stability o f forest ecosystems, and the persistence of biodiversity. Silva Fennica. 36(1): 393-400. Johnson, E . A . (1992). Fire and Vegetation dynamics: Studies from the North American Boreal Forest. Cambridge, U K : Cambridge University Press. Johnson, E . A . , & Gutsell, S. L . (1994). Fire frequency models, methods and interpretations. Advances in Ecological Research. 25: 239-287. Johnson, E . A . , & Larsen, C. P. S. (1991). Climatically induced change in fire frequency in the southern Canadian Rockies. Ecology. 72(1): 194-201. Johnson, E . A . , Miyanishi, K . , & Bridge, S. R. J. (2001). Wildfire regime in the boreal forest and the idea of suppression and fuel buildup. Conservation Biology. 15(6): 1554-1557. Johnson, E . A . , Morin, H . , Miyanishi, K . , Gagnon, R. , & Greene, D . F. (2003). A process approach to understanding disturbance and forest dynamics for sustainable forestry. In P.J . Burton, C. Messier, D . W . Smith, and W . L . Adamowicz, (eds.). Towards sustainable management of the boreal forest, (pp. 261-306). Ottawa, Ont.: N R C Research Press. 24 Kuuluvainen, T. (2002). Natural variability of forests as a reference for restoring and managing biological diversity in boreal Fennoscandia. Silva Fennica. 36(1): 97-125. Laska, G . E . (2001). The disturbance and vegetation dynamics: A review and an alternative framework. Plant Ecology. 157(1): 77-99. Landres, P. B . , Morgan, P., & Swanson, F. J. (1999). Overview o f the use of natural variability concepts in managing ecological systems. Ecological Applications. 9(4): 1179-1188. Logan, J. A . , & Powell, J. A . (2001). Ghost forests, global warming, and the mountain pine beetle (coleoptera: Scolytidae). American Entomologist. 47(3): 160-172. Lotan, J. E . 1975. The role o f cone serotiny in lodgepole pine forests. Symposium proceedings, Vol. 1: Management of lodgepole pine ecosystems D . M . Baumgartner ed. Washington State University Cooperative Extension Service. Mantua, N . and Hare, S. 2006. Pacific Decadal Oscillation Data. Retrieved Apri l 2006 from: http ://www.jisao .Washington, edu/pdo/ McGarigal , K . and Marks, B . J . 1994. F R A G S T A T S : Spatial Pattern Analysis Program for Quantifying Landscape Structure (Version 2.0). Retrieved Dec. 6, 2006 from: http://www.umass.edu/landeco/research/fragstats/fragstats.html. McGregor, M . (2002). Developing Reference Conditions for Ecological Restoration of Forests in Sub-boreal British Columbia. Thesis submitted in partial fulfillment of the requirements for the degree of Master o f Science in Natural Resources and Environmental Studies, University o f Northern British Columbia. Prince George, B C : University o f Northern B C . pp. 149. Meidinger, D . and Pojar, J. 1991. Ecosystems o f British Columbia. Special report series no. 6. B C Ministry of Forests Research Branch. Victoria, B C . Millar, C . I., & Woolfenden, W . B . (1999). The role of climate change in interpreting historical variability. Ecological Applications. 9(4): 1207-1216. Mil lward, A . A . , & Kraft, C . E . (2004). Physical influences o f landscape on a large-extent ecological disturbance: The northeastern North American ice storm of 1998. Landscape Ecology. 19:99-111. 25 Mitchell , R. J., Palik, B . J. , & Hunter, M . L . J. (2002). Natural disturbance as a guide to silviculture. Forest Ecology and Management. 155: 315-317. Moore, M . M . , Covington, W . W., & Fule, P. Z . (1999). Reference conditions and ecological restoration: A southwestern ponderosa pine perspective. Ecological Applications, 9(4), 1266-1277. Noss, R .F . , LaRoe, III E .T . and Scott J . M . (1995). Endangered ecosystems of the United States: a preliminary assessment of loss and degradation. National Biological Service Biological Report 28. U.S. Department of the Interior, Washington, D . C . , U S A . Odion, D . C . , Frost, E.J . , Strittholt, J.R., Jiang, H . , Dellasala, D . A . , & Mori tz , M . (2004). Patterns o f fire severity and forest conditions in the Western Klamath Mountains, California. Conservation Biology. 18(4): 927-936. Oliver, C. D . , & Larson,' B . C. (1996). Forest Stand Dynamics. N e w York : John Wiley & Sons, Inc. O'Neill , R. V . , & Smith, M . A . (2002). Scale and hierarchy theory. In S. Gergel, & M . G . Turner (Eds.), Learning Landscape Ecology. New York : Springer-Verlag, Inc. Parmesan, C , & Yohe, G . (2003). A globally coherent fingerprint o f climate change impacts across natural systems. Nature. 421: 37-42. Perry, G . L . W. (2002). Landscapes, space and equilibrium: Shifting viewpoints. Progress in Physical Geography. 26(3): 339-359. Peterson, G . , Allen, C. R., & Holling, C. S. (1998). Ecological resilience, biodiversity, and scale. Ecosystems. 1: 6-18. Pickett, S. T. A . , Kolasa, J. , Armesto, J. J., & Collins, S. L . (1989). The ecological concept of disturbance and its expression at various hierarchical levels. Oikos. 54: 131-139. Purdon, M . (2003). The nature of ecosystem management: Postmodernism and plurality in the sustainable management of the boreal forest. Environmental Science & Policy, 6, 377-388. 26 Reed, W . 2006. A note on fire frequency concepts and definitions. Canadian Journal o f Forest Research. 36: 1884-1888. Rhemtulla, J. M . , Hall, R. J. , & Macdonald, S. E . (2002). Eighty years o f change: Vegetation in the montane ecoregion o f Jasper National Park, Alberta, Canada. Canadian Journal of Forest Research. 32: 2010-2021. Ries, L . , Battin, J., & Sisk, T. (2004). Ecological responses to habitat edges, mechanisms, models and variability. Annual Review of Ecology, Evolution and Systematics. 35: 491-522. Runkle, J. R. (1985). Disturbance regimes in temperate forests. In S. T. A . Pickett, & P. S. White (Eds.), The Ecology of Natural Disturbance and Patch Dynamics. Florida: Academic Press Inc. Ryan, K . C. (2002). Dynamic interactions between forest structure and fire behaviour in boreal ecosystems. Silva Fennica. 36(1): 13-39. Sachs, D . L . , Sollins, P., & Cohen, W. B . (1998). Detecting landscape changes in the interior o f British Columbia from 1975 to 1992 using satellite imagery. Canadian Journal of Forest Research. 28: 23-26. Shinneman, D . J., & Baker, W . L . (1997). Nonequilibrium dynamics between catastrophic disturbances and old-growth forests in ponderosa pine landscapes of the Black Hills. Conservation Biology. 11(6): 1276-1288. Stewart, R . B . , E . Wheaton, and D . L . Spittlehouse. (1998). Climate change: implications for the Boreal forest. In A . H . Legge and L . L . Jones (eds.) Emerging Air Issues for the 21st Century: The Need for Multidisciplinary Management, Proceedings of a Speciality Conference, Sep. 22-24, 1997, Calgary, AB.. A i r and Waste Management Assoc., Pittsburgh, P A , pp 86-101. Swetnam, T. W. , Allen, C. D . , & Betancourt, J. L . (1999). Applied historical ecology using the past to manage for the future. Ecological Applications. 9(4): 1189-1206. Thomas, C , Cameron, A . , Green, R. E. , Bakkenes, M . , Beaumont, L . J. , & Colligham, Y . et al. (2004). Extinction risk from climate change. Nature. All: 145-148. 27 Tinker, D . B . , Romme, W . H . , & Despain, D . G . (2003). Historic range of variability in landscape structure in subalpine forests o f the greater Yellowstone area, U S A . Landscape Ecology. 18:429-439. Turner, M . G. P., Romme, W . H . P., Gardner, R. H . P., O'Neill , R. V . P., & Kratz, T. K . P. (1993). A revised concept of landscape equilibrium: Disturbance and stability on scaled landscapes. Landscape Ecology. 8(3): 213-227. Veblen, T. T. (2003). Historic range of variability of mountain forest ecosystems: Concepts and applications. Forestry Chronicle. 79(2): 223-226. Volney, W . J. A . , & Fleming, R. A . (2000). Climate change and impacts of boreal forest insects. Agricultural Economics and Environment. 82: 283-294. Ward, T. (2001). Sustainability for British Columbia forestry. Journal of Sustainable Forestry. 12: 153-169. Warren, F. J. (2004). Climate change impacts and adaptation: A Canadian perspective. Ottawa, Ontario: Climate Change and Adaptation Directorate, Natural Resources Canada. Watt, A . S. (1947). Pattern and process in the plant community. The Journal of Ecology. 35(1/2): 1-22. Wei , X . , Kimmins, J. P., & Zhou, G. (2003). Disturbances and the sustainability o f long-term site productivity in lodgepole pine forests in the central interior of British Columbia-an ecosystem modeling approach. Ecological Modelling. 164: 239-256. Weir, J. M . H . , Johnson, E . A . , & Miyanishi, K . (2000). Fire frequency and the spatial age mosaic of the mixed-wood boreal forest in western Canada. Ecological Applications. 10(4): 1162-1177. White, P. S., & Pickett, S. T. A . (1985). Natural disturbance and patch dynamics: A n introduction. The Ecology of Natural Disturbance and Patch Dynamics. S .T.A. Pickett & P.S. White, eds. 27-42. 28 Wong, C , & Iverson, K . (2004). Range o f natural variability: Applying the concept to forest management in central British Columbia. BC Journal of Ecosystems and Management. 4(1): 1-14. Wu, J., & Loucks, O. L . (1995). From balance of nature to hierarchical patch dynamics: A paradigm shift in ecology. The Quarterly Review of Biology. 70(4), 439-466. Zhang, Q., Alfaro, R. I., & Hebda, R. J. (1999). Dendroecological studies of tree growth, climate and spruce beetle outbreaks in central British Columbia, Canada. Forest Ecology and Management. 121: 215-225. Zhang, Q.-B. , Hebda, R .J . , Zhang, Q.J. and Alfaro, R.I. (2000). Modelling tree-ring growth responses to climatic variables using artificial neural networks. Forest Science. 46(2): 229-239. 29 C H A P T E R 2 R A N G E O F VARIABILITY IN DISTURBANCES IN A S U B - B O R E A L E C O S Y S T E M IN BRITISH C O L U M B I A , W I T H IMPLICATIONS F O R P O L I C Y 2 2.1 I N T R O D U C T I O N Detection and characterization of the effects o f disturbance on the landscape is essential to ensure that in managed ecosystems, management practices such as timber harvesting are compatible with the ecosystem's capacity to recover from disturbance (Bergeron et al., 2002). A n evaluation of the range of variability ( R V ) in both disturbance processes (e.g., type, frequency, severity, and extent) and landscape conditions (e.g. distribution of age classes, serai stages, and species assemblages) provides a framework for the achievement and perpetuation of similar conditions through well-planned management activities (Landres et al., 1999; Haeussler et al., 2002; Wong & Iverson, 2004). Species are adapted to variability in processes such as fire disturbance, as well as the range of habitat created by such processes, which is an important consideration in the sustainability of biodiversity, forest productivity, and wildlife habitat (Delong, 2002b; Kuuluvainen, 2002). Quantifying the R V requires studying the spatial and temporal variability of ecosystems within a specific time period (Landres et al., 1999; Swetnam et al., 1999). Most often, scientists have viewed human activities as a dominant source of unusual disturbances and changes to ecosystems (Noss et al., 1995), but severe or rapid climate changes could also have strong impacts (Millar & Woolfenden, 1999). Whether or not to include Aboriginal human influences, and how to define the time period of study must be determined by the objectives of the research. Nevertheless, there are two overall assumptions usually made when using the concept o f R V : 1) variability can be described by historical or natural conditions, and 2) the spatial and temporal variation in these conditions is integral to ecosystem health and function (Landres et al., 1999; Delong, 2002b). The goal of using R V as a guide is not to necessarily to re-create conditions (historical, natural or otherwise) on the landscape, but to understand the mechanisms that caused change from previous or natural conditions to those o f today (Landres et al., 1999). Sustainable forest management (SFM) attempts to integrate our understanding of the R V in the natural landscape with economic and social objectives for the forest to protect the full range of functioning in 2 A version of this chapter will be submitted for publication to Canadian Journal of Forest Research. Campbell, K .A . , and Dewhurst, S.M. Range of variability in disturbances in a sub-boreal ecosystem in British Columbia with 30 forested ecosystems over time and space. In British Columbia, one goal of S F M is to manage our lands in a similar manner as natural disturbances; that is to harvest at similar rates (temporal patterns) and to create similar landscape conditions (BC Ministry o f Forests and Range, 2004). Progress toward this goal o f S F M can be determined using indicators, supported by scientific research, which measure specific elements o f the landscape (Karjala & Dewhurst, 2003). Disturbance indicators, for example, describe the disturbance regime in terms of disturbance frequency, magnitude, and spatial configuration. Specific indicators might include measures of disturbance return interval, severity of disturbance, and disturbance size. Landscape condition indicators, on the other hand, describe the type and amount o f land cover, structural characteristics, and spatial configuration. Specific indicators might include quantification o f the landscape in terms o f species composition, serai stage distribution (age), and patch size. Indicators are useful in that they allow: 1) the comparison of a consistent set of indicator values over time to measure successes and failures towards a certain goal such as S F M , and 2) the comparison of indicator values to a specific target, which may have been set as policy for management. The purpose o f this research is to investigate the historical and current R V in natural disturbances in the central interior of B C in order to assess and refine current S F M policies of management. The objectives were to: 1. Characterize the R V of disturbances using disturbance and landscape indicators within the past 100 years. 2. Evaluate the how the R V in disturbance and landscape indicators changed over time. 3. Compare disturbance and landscape indicator values with current S F M policy. 2.2 M E T H O D S 2.2.1 Study area description The study area encompasses 1 million ha in the central interior o f B C (Figure 2-1). The area comprises the dry cool subzone of the sub-boreal spruce ecosystem (SBSdk) within the administrative boundaries o f the Lakes Timber Supply Area (TSA) , with a smaller portion in the Morice and Prince George TSAs , all in the Northern Forest Region. Mean annual temperatures in the study area are cool, ranging from 1.1°-2.1° C (Delong et al., 1993). In the winter, snow implications for policy .31 depths can reach up to 250 cm, while the rest o f the year, an average of 500 mm falls as rain. The growing season is relatively short, with only about 70 frost-free days a year. The dominant leading species on the landscape is lodgepole pine {Pinus contorta var. latifolid). Interior spruce (Picea glauca (Moench) Voss x engelmannii Parry ex Engelm.), aspen (Populus tremuloides Michx.) , Cottonwood (Populus trichocarpa Torr. & A . Grey), birch (Betula papyri/era), and black spruce (Picea mariana (Mill.) Britton, et al.) are also abundant, with minor amounts o f Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), hemlock (Tsuga heterophylla (Raf.) Sarg.), and Engelmann Spruce (Picea engelmannii Parry ex Engelm.). The age of treed stands range from 1 to about 400 years. Both natural and anthropogenic forces contribute to a disturbance regime that is characterized by frequent, stand-initiating disturbance events. Historically, stand-replacing fires were common (Delong, 2002a), with an expected mean fire return interval of 125 years in drier areas and 200-900 years in wetter areas. Historic fires burned areas as large as 15,000 ha in size (Bunnell, 1995). The resulting landscape is primarily composed o f even-aged patches, with some remnant stands (Delong, 1998). Burning by Aboriginal people, who have occupied lands in and around the study area traditionally, is also a disturbance agent (Heitzmann, 2001). The extent to which First Nations in the study area used fire is only beginning to be studied, although there are suggestions that it was part o f traditional culture (Gottesfeld, 1994). Early in 20 t h century, railway building and logging activities began to shape the landscape, moving into the interior of the province by the 1940's (Fulton, 1910; Sloan, 1945; Sloan, 1956). Fire suppression activities were mandated, and between 1944 and 1955, over 1300 fires had been extinguished in the province (Sloan, 1956). A s of 2006, timber harvesting activities yielded 20 million m 3 of timber per year in and around the study area (Lakes, Morice, and Prince George T S A s combined), and today, fire suppression efforts are nearly 92% effective . Broad-scale mountain pine beetle outbreaks are not historically uncommon in B C , occurring in the last century in the Kootenays, north-central B C (Babine and Stikine) and the Chilcotin (Taylor & Carroll , 2004). Throughout B C , Taylor and Carroll (2004) found that 17% of the landscape would have been susceptible to the mountain pine beetle early in the 20 t h Century, as compared to 55% in 2003. A s o f 2005, 283 million m of wood has been killed in http://www.for.gov.bc.ca/protect/suppression/ 32 B C 4 by a recent epidemic, which began as early as 1993. The epidemic has been projected to reach its peak by 2006, although the infestation will still continue after that time (Eng, 2005). In the study area, the recent outbreak appears to have impacted the landscape more severely than historical outbreaks, but future research may repute this. In the past 50 years there has been minor mountain pine beetle activity in the area. However, this information may not have occurred with the same severity over as large an area as currently. 2.2.2 Characterize the R V of disturbances using indicators wi th in the past 100 years Disturbance and landscape indicators were calculated using historical disturbance datasets and vegetation data, described in detail below. The 100-year time period (1904-2004) was divided into two 50-year periods: historical (1904-1953) and current (1954-2004). The choice of 100 year time period was constrained by data availability, but we explored the implications of our choice using some simple graphical methods. Both disturbance and landscape indicators were chosen based on data availability, but also, more importantly, based on S F M planning and policy documents. 2.2.2.1 Description o f vegetation data The former B C Ministry o f Sustainable Resource Management supplied the Vegetation Resource Inventory (VRI) data for this study. This dataset contains detailed information on vegetation for the study area, including species composition, age, topography, and broad vegetation classes at a scale of 1:20,000. These data were used to stratify the landscape both for historical and current conditions. There were 4 strata: forest, cover type, habitat type, and age class (Table 2-1). Strata were divisions of the original spatial data, created by classifying stands (polygons) based on common characteristics such as stand or cover types, habitat types, and age. Disturbance indicators and landscape indicators could then be reported by strata, at least for forest, cover type, habitat type. The age stratum was used only to calculate serai stage indicators. Strata are defined as unique combinations of characteristics of the landscape. The entire vegetated portion o f the study area is the forest level stratum, which includes both treed and non-tree (shrub) areas. Cover type strata were defined based on leading species comprising 50% or 4 British Columbia's Mountain Pine Beetle Action Plan 2005-2010, online at: http://www.for.gov.bc.ca/hfp/mountain pine beetle/actionplan/2005/ 33 greater basal area in the stand. Habitat type strata were defined based on broad scale topography (upland vs. wetland) and vegetation characteristics (treed, > 10% crown cover, or not treed, < 10% crown cover). For lack of alternate information, it was assumed that the cover type and habitat type stratifications were static in time. Age class strata were based on projected age classed into 10-year intervals. In areas affected by mountain pine beetle, the stand age of lodgepole pine in the inventory was corrected to the year when attack levels were greater than 30%. For example, i f the stand was reported as being 100 years old, but had greater than 30% mountain pine beetle attack in 2000, the stand age was corrected to be 4 years old. 2.2.2.2 Description o f disturbance datasets Fire and mountain pine beetle disturbance data were obtained, in part, from the Canadian Forest Service as part o f the Natural Disturbance Database ( N D D ) . The N D D is a spatial, polygon based (vector) dataset, which contains information on the spatial extent of the disturbance, year of disturbance and, for mountain pine beetle only, disturbance severity. Data were available for 1920-2000 and were at a scale of 1:250,000. Mountain pine beetle data 2001-2004 were obtained from the B C Ministry o f Forests and Range. This dataset was in the same format as the N D D . Spatial fire data were not available for this time period because the Ministry o f Forests and Range no longer tracks the spatial extent of individual fires, but rather simply tracks the point location o f fire origin. Additional disturbance data were derived from the V R I , using age as a surrogate for time since disturbance (Delong 1998) for the period 1904-1953. N o assumptions were made about the cause of the disturbance; these disturbances were classified as being "unknown." Based on conversations with the Canadian Forest Service (B. Hawkes, Canadian Forest Service, pers. comm.) the fire and mountain pine beetle contained in the N D D are less reliable than more recent times. B y identifying additional data based on age, this may present a more accurate picture of disturbance activity in the early part of the 20 t h Century. In addition, inclusion of this data allowed us to characterize the R V in our indicators over an entire 100-year period. Harvesting disturbance was identified based on age as follows: i f a stand was less than 50 years of age, and had not been disturbed by mountain pine beetle or fire, it was assumed to have been harvested. 34 2.2.2.3 Time period of analysis Choosing a time period to quantify R V can be difficult when human influences (Russell, 1997) and long-term climate trends affect the system (Millar & Woolfenden, 1999). Ideally this decision would be based on a clear understanding of the disturbance processes and the resulting temporal or spatial patterns, during a time period that is considered free from unusual perturbations (Landres et al. 1999, Swetnam et al. 1999). Most often, scientists have viewed human activities as the cause for unusual disturbances and changes to the ecosystem (Noss et a l , 1995), but severe or rapid climate changes could also have strong impacts (Millar and Woolfenden 1999). Practically, limited historical data availability, and the prevalence of human activities in forested areas in particular, makes this a difficult task. Whether or not to include Native American influences, and how to define the time period of study can only be determined by the objectives o f the research. The time period used in this study (1904-2004) was constrained largely by data availability. Data were split into two time periods: historical (1904-1953) and current (1954-2004). The purpose o f grouping the data into two distinct time periods was twofold: 1) to have an similar number of years in the two analysis time periods, and 2) to facilitate indicator comparisons between a time with very little human influence, 1904-1953, and a time with more human influence, 1954-2004. Based on a review of historical documents which indicated that fire suppression began in the 1950's in this area, disturbances in the historical time period, 1904-1953 were considered to have been less affected by humans than those occurring 1954-2004. In essence, fire, mountain pine beetle, and unknown disturbances in the historical time period were "natural" while those in the current time period were less so. The distinction between the two time periods admittedly ends up being somewhat arbitrary because there would be no clear differentiation in activities between the time periods. Harvesting and fire suppression started sometime in the early part of the century, but exact dates are not known, nor are they expected to coincide. To determine whether or not the historical time period was representative of natural conditions, we carried out a simple graphical exercise. This method, was suggested by Weir etal . (2000) for the boreal forest, which has similar disturbance regimes as the sub-boreal. Under a stand-replacing fire regime which burns independently o f age, the theoretical age class distribution follows a negative exponential, or "reverse ' J ' " distribution (Van Wagner, 1978; Johnson & Gutsell, 1994; Weir & Johnson, 2000). The age class distribution o f stands in the study area can therefore be plotted over time as cumulative percent of the study area on a 35 logarithmic axis. A falling straight line on such a graph, according to Weir et al. (2000), is indicative of a cohesive disturbance regime, represented by specific a disturbance frequency. Over time, a disturbance frequency can change, and these changes are identified as changes in the slope of lines on the graph. The time period when the change occurs is termed the change point. It should be noted that statistical analysis was not carried out (for example, regression) to test the different slopes between time periods. A visual detennination of the location of the lines was relied upon instead. In hindsight, statistical testing might provide a robustness to this method that is currently lacking. Perry (2002) cautions researchers using this method, however, because assuming that the distribution itself wi l l remain static in time or even space is erroneous under the nonequilibrium view of ecosystem dynamics. In stands with management activities such as harvesting, research has demonstrated that the age class distribution does not follow a negative exponential distribution, nor is it desirable for management outcomes (Fall et al. 2004). Nevertheless, as a tool for observing coarse temporal patterns, researchers have shown its applicability in forests with stand-replacing fire regimes (see for example, Johnson and Larsen 1991, Weir et al., 2000). The assumption in this study was that such as change would be present around 1950 or 1960 in the study area, which corresponds to the approximate time when harvesting in the area began to dominate, and fire suppression started. Prior to this change point is the time period used in this study to calculate the R V in historical indicator values. 2.2.2.4 Choosing disturbance and landscape indicators Indicators o f R V must be defined which are measurable given the existing data, and are representative o f the variability o f the disturbance processes and landscape condition themselves. Indicators were also chosen by reviewing available policy to ensure consistency of this study with previous documents whenever possible. The study area covered three separate management units (Lakes Timber Supply Area (TSA) , Morice T S A , and Vanderhoof Forest District in the Prince George Forest Region) each with its own policy. Only one comprehensive document was available, the B C Biodiversity Guidebook ( B C Ministry o f Forests, 1995). This document is over 10 years old now, but some of the management units, such as the Vanderhoof District, are still following these guidelines, although in 2006 a review o f the policy is underway. The policy documents consulted span a period o f about 10 years and were: 1) B C Biodiversity Guidebook ( B C B G B ) ( B C Ministry of Forests, 1995), 2) Vanderhoof Land and Resource Management Plan ( L R M P ) (Province of British Columbia, 1997), 3) Lakes T S A L R M P ( B C Ministry of 36 Environment, Lands and Parks, 2000), and Higher-level Plan Order (Province of British Columbia, July 26, 2000), 4) Natural Disturbance Units of the Prince George Forest Region: Guidance for Sustainable Forest Management (Delong, 2002a), 5) Morice T S A L R M P ( B C Ministry of Sustainable Resource Management, 2004), and 6) Morice T S A Sustainable Forest Management Plan (Morice and Lakes 1FPA, 2005). Not all documents made specific recommendations on indicator targets, but all documents were consulted. Final selection of criteria and indicators is presented in Table 2-2. Landscape condition indicators were chosen to describe the current distribution of cover types, and habitat types in the study area (Indicators 1-3). These indicators were the only indicators which portrayed current conditions only, the remaining describe current and historical conditions for comparison between the two time periods. The fourth landscape condition indicator measured the amount of each cover type in early serai stage currently and historically. Disturbance indicators were chosen to encapsulate measures of the overall disturbance regime. Specifically, indicators were calculated to describe: . Disturbance type (fire, mountain pine beetle, unknown, and harvesting)- Indicators 6a, 7, 11 • Area disturbed overall, and in each cover type and habitat type- Indicators 5, 6b, 6c • Disturbance frequency in each cover type and habitat type- Indicators 10a, 10b Disturbance return intervals over time, and for each cover type- Indicators 8, 9 Indicators o f disturbance type were chosen to illustrate potential differences in the mechanism o f disturbance over the past 100 years. Indicators of disturbance frequency, and disturbance return interval depict the temporal variability of disturbance using terminology which is reflected in policy documents. Finally, although this analysis was not spatial, indicators of overall disturbance act as approximation o f the spatial impact o f disturbance. 2.2.2.5 Calculating indicators A l l spatial data were combined into one dataset using G1S. Data were then summarized aspatially. The final dataset contained area, in hectares, by: • Disturbance year (1904-2004) • Disturbance type (fire, unknown, mountain pine beetle, harvesting) • Cover type (leading species) • Habitat type (slope position and broad vegetation class) • Current serai stage (early, young, mature or old) 37 Calculation o f the indicators follows the description o f the methods in Table 2-2. Historical serai stage indicators were calculated using disturbance year as a time since disturbance surrogate. Thus all area disturbed from 1904 to 1953 was considered to be in early serai stage. 2.2.3 Evaluate indicators changes in time, and compare with SFM policy A l l indicators (with the exception o f indicators 1 & 2) were compared in time using graphical or tabular methods. The indicators were compared between historical (1904-1953) and current (1954-2004 time periods. Statistical comparisons were not possible for some of the indicators, and not desired for others. Indicators 1-6 quantify population, rather than sample, values, and thus statistical analysis was not possible. Indicators 7-9 are averages over time, either 49/50 years or by decade. Similar averages are used in S F M policy documents, and the purpose o f these indicators was to look at the population trends at finer levels o f detail using indicators comparable to policy documents. They are, as such, averages over time o f absolute indicator 1-6 values presented, and as such no statistical testing was required. 2.3 R E S U L T S 2.3.1 Historical time period It was evident from the graph o f cumulative area in each age class (Figure 2-2) that the disturbance regime during the historical time period, 1904-1953, was reasonably cohesive, meaning that there was a relatively straight line during this period. This indicates that the disturbance regime was fairly constant during this period. A change point appears to have occurred around 1874, suggesting that in the future, analysis of historical range of variability could extend another 30 years back in time. Another change point occurs in either 1944, or 1954. This may occur because o f fire suppression, although climate changes, decrease in Aboriginal burning or European agriculture and settlement could have caused this change. Between 1954 and 1993, the increase in cumulative area each decade was less than the previous time periods. A final change point could possibly have occurred in 1994, with again a higher disturbance frequency than 1954-1993. Note that the current disturbance frequency however, is of similar magnitude o f the time period between 1874 and 1953, which suggests that current levels o f disturbance from mountain pine beetle are within the range of variability o f the first part o f the century. 38 2.3.2 Compar i son of indicators over time Indicators 1-3 described the current condition o f the landscape. The dominant cover type (Indicator 1) was lodgepole pine (51%) as expected, followed by shrub types (17%), deciduous species (16%) and interior spruce (15%) (Figure 2-3). Seventy-seven percent of the study was upland and treed habitat (Indicator 2), 18% was wetland and treed habitat, while the remaining habitat types combined make up the remaining 5% o f the landscape (Figure 2-4). Cover types were distributed amongst the different habitat types (Indicator 3) (Figure 2-5). Lodgepole pine was the dominant species in upland, treed habitats, while black spruce was dominant on wetland, treed sites. Some tree species were on the shrub upland and wetland types. This is likely an artefact o f the inventory. When being classified in the inventory, a stand is classified as non-treed i f there is less than 10% tree crown cover. I f younger stands (recently regenerated) were not identifiable as being treed, or i f stands were not sufficiently restocked, it would have been classified as non-treed. This could possibly bias the results of this analysis since there may be greater early serai stage currently than has been reported below. Unless otherwise stated, all the remaining indicators are reported for the 49 year historical time period, or the 50 year current time period. Thirty-three percent of the forested landscape was disturbed between 1904-1953 and between 1954-2004 (Indicator 4; Figure 2-6). Historically, only 2% of the disturbances were fires caused by lightning (Figure 5 a); the majority of disturbances which occurred in this time period were anthropogenic fires (57%>) or o f unknown disturbance origin (41%) (Figure 2-7). In the past 50 years, the majority o f disturbances in this study area were from M P B (55%), and harvesting (44%>), with a minor amount o f fires (1%). Over time, the disturbance frequency and disturbance type has changed (Indicator 6; Figure 2-8). Between 1904 and 1953, disturbances were primarily fires. After 1954, harvesting slowly increased, until the most recent decade, when the mountain pine beetle epidemic began. The last decade had the highest disturbance frequency, followed by the period between 1924 and 1933. Historically, disturbances ranged from 20-60% of the area in each cover type while they ranged from 0-50% of the area in each over type from 1954-2004, depending on cover type (Indicator 5b; Figure 2-9). Deciduous cover types had the most disturbance historically at 60%, but in the past 50 years has had only 14%. Lodgepole pine stands had the highest disturbance rate between 1954 and 2004, but less historically. Disturbance frequency in interior and black spruce cover types declined slightly between historic and current time periods. Historically disturbance in upland treed and wetland treed habitat types was fairly similar between the two time periods, and the same for upland and wetland shrub types (Indicator 5c; 39 Figure 2-10). In the current time period, however, disturbance is greater in wetland habitat types, while there has been minimal change in upland habitat types. Average decadal disturbance frequencies ranged from 4.0 to 12.2% in the historic time period, and 0 to 8.2% in the current time period (Figure 11a). Except for pine, all disturbance frequencies were lower in the past 50 years as compared to frequencies 1904-1953. Historically, average disturbance return intervals each decade ranged from 82 to 250 years, depending on cover type (Figure 2-11). In the past 50 years, disturbance return intervals were, in general, longer, with the exception o f lodgepole pine stand types. Deciduous cover types had much longer return intervals, increasing from 82 years historically to 352 currently. Disturbance return intervals were much longer in the wetland shrub habitat currently, but otherwise there has been overall, little change between the two time periods (Figure 2-12). When viewed over time by cover types, it appears that disturbance return intervals followed a very similar pattern in the first half o f the 20 t h century. After the mid 1940's, when fire suppression began and harvesting became the dominant disturbance agent, disturbance regimes were much more variable (Figure 2-13). That is, prior to this time period individual cover type had different return intervals, but this difference is not as variable in the historic time period as currently. For example, lodgepole pine and spruce have much longer return intervals within this time period than historically, but these return intervals shorten 1984-2004. In the most recent decade, return intervals are the shortest they have been in 80 years. It is clear from Figure 2-8 however, that area disturbed began to decline before 1954, suggesting that some fire suppression activities may have been active during this time period. Also, since disturbances are variable, it may have been that climate during this period was not as conducive to disturbance as previous decades. Historically, fires burned cover types in similar proportion as they were represented on the landscape (Figure 2-14). Over 50% of the area burned during this time period was lodgepole pine, followed by deciduous (19%), shrub (17%), and spruce (11%). Unknown disturbances affected the cover types differently, with deciduous stands making up the most area affected at 44%, lodgepole pine at 38%, and spruce at 16%. In the current time period, the minimal fires that did burn also affected many o f the cover types, although primarily lodgepole pine (63%), deciduous (23%), and spruce (10%). Harvesting in this time period targeted more spruce than other disturbance types (29% of the area), but similarly affected pine (51%) and deciduous (23%). The area in early serai stage ranged from 27% to 52% in the historical time period, and from 12% to 50% in the current time period (Figure 2-15). Lodgepole pine had the greatest 40 increase in early serai stage (from 28% to 52%), while the deciduous cover type had the greatest decrease (from 52% to 12%). The effects of changes in early serai stage demonstrated changes in stand dynamics. For example, overall, the amount o f early serai stage on the landscape was similar historically and currently. However, lodgepole pine stands were older historically than now, while deciduous stands were younger, suggesting that stand dynamics and resulting landscape conditions have been altered between the two time periods. 2.3.3 Compar ison with policy Policy targets are outlined in detail in Table 2-2. The targets for cover types outlined in the Morice S F M P were met in the SBSdk, except for spruce, which, at 15% o f the study area, is below the target of greater than or equal to 28%. Policy recommendation for the amount o f area in early serai stage varied considerably depending on the management unit, and the policy document. In addition, some of the policy documents divided up recommendations by biodiversity emphasis units and resource management zones.5 In the Morice T S A , the S F M P recommended that less than 64% o f the general forested area be in early serai stage, but that less than 50% of the area should be in early serai stage in the high biodiversity emphasis units (Figure 2-16a). These targets were met in the historical and current time period, and in fact only the historical deciduous cover type had greater than 50%> early serai stage. For the Lakes T S A , the Higher Level Plan laid out a range o f targets by B E U and by resource management zone (Figure 2-16b). Resource management zone ( R M Z ) B had the most conservative targets, depending on B E U , from 25%, 32%, and 54% for low, intermediate, and high respectively. Historical amount of early serai stage exceeded the targets in the high B E U in R M Z B in all species, but only black spruce and deciduous cover types exceeded the targets in the intermediate B E U . Lodgepole pine was the only cover type currently that exceeded both the high and intermediate B E U targets in this R M Z , while deciduous and subalpine fir cover types were well below even the high B E U target. None o f the cover types exceeded the highest target of 54% in either of the time periods, but the lodgepole pine cover type was nearing this level. The B C Biodiversity Guidebook and the Natural Disturbance Units by Delong (2002a) have a similar range in target recommendations as the Lakes Higher Level Plan (40% or 54%> by 5 The B C B G B defines biodiversity emphasis units as the following: "Each option is designed to provide a different level of natural biodiversity and a different risk of losing elements of natural biodiversity." Higher biodiversity emphasis units, for example, have less risk to biodiversity, than lower biodiversity emphasis units. 41 B C B G B , and 25-50% by Delong (2002a)) (Figure 2-16c). Most cover types are well within this range or below the targets, with the exception o f deciduous and subalpine fir cover types in the current time period, which are well below the 25% target. Disturbance return interval targets were 93 years (Morice L R M P ) , 100 years (Delong 2002a), 125 years ( B C B G B ) , or, for black spruce stands only, 250 years (Morice L R M P ) . Historically, except for deciduous stand types, all cover types had longer return intervals than the longest target of 125 years (Figure 2-11). Lodgepole pine in the current time period had a return interval o f 125 years however. The black spruce in both historical and current time periods had much shorter return intervals than 250 years. Finally, the disturbance frequencies historically were within the targets recommended by Delong (2002a) of being between 7.5-12.2%, except for the shrub cover type. Currently, however only the lodgepole pine and black spruce cover types are within this range, the other cover types are below this range. 2.4 D I S C U S S I O N 2.4.1 Historical time period Disturbances occurring in the historical time period (1904-1953) were not independent o f human influences (Figure 2-7), but appeared to be representative of the disturbance regime which was dominant up until the time o f fire suppression (Figure 2-2). Prior to about 1874, another disturbance regime with different disturbance frequencies, in fact higher disturbance frequencies, shaped the landscape. Several researchers have found a similar change point in the late 1800's in Canada, and attribute this change to the end o f the Little Ice Age (Weir et al. 2000; Bergeron et a l , 2001; Bridge et al., 2005; Sanborn et al. 2006). Despite cooler and wetter temperatures during this time, lightning activity was elevated causing more area to burn. 2.4.2 Changes in disturbance over 100 years At the broadest scale in this study, the forest, disturbance frequency has not changed in the past 100 years (Figure 2-6), even though the current mountain pine beetle outbreak appears so widespread. However, the mechanism of disturbance has changed (Figure 2-7), which affects both the current and future distribution of forest species and age classes on the landscape. The historic regime was one o f stand-replacing fires, coupled with other disturbance events. Fires, which were prevalent during the beginning o f the 20 t h century, burned all cover types in the study area (Figure 2-14). What was classified as "unknown" disturbance in this study could be a combination o f unrecorded fires (including Aboriginal burning), insect activity (mountain pine 42 beetle, spruce beetle, etc.), conversion to agriculture with increasing settlement, and a miniirium of harvesting. It is unfortunate the causal agent o f these disturbances remains unknown since they comprise such a large portion (41%) of disturbance events in the historical time period; it is difficult to build a comprehensive characterization o f a historical disturbance regime without this information. Future, on the ground research is required to further tease out the various disturbance types which were shaping the landscape during this time period. In the past 50 years, management of the landscape in the study area has included fire suppression efforts, coupled with harvesting as the dominant disturbance. There was minor fire activity, but more recently the mountain pine beetle has become a significant disturbance agent. Although harvesting is now intended to mimic natural disturbance processes this study shows that it does not necessarily create similar conditions on the landscape. With the exclusion of pine, which was affected by mountain pine beetle, there was discrepancy in disturbance rates, return intervals, and percentage of early serai species in cover types such as deciduous and, to a lesser extent, subalpine fir, in addition to pine. This discrepancy seemingly can be explained by the rate at which harvesting is carried out. Bergeron et al. (2002) advise that it is only possible to substitute harvesting for fire in forests which have a low enough disturbance frequency that harvesting can match the rate of disturbance. Harvesting in the study area over the past 50 years matched the rate in the species preferred for commercial purpose, such as pine and interior spruce stands, but less so in deciduous and black spruce (Figure 2-8). It should be noted however that the harvest profile, which targets specific species such as pine over less commercial deciduous species, did not appear much different than the species profile of historical burns alone (Figure 2-14). However, for deciduous species, even though some harvest activity has occurred in these stands, the cumulative disturbance rate of fire and unknown disturbances historically was much greater than currently. Again, this is likely because deciduous stands are not valued commercially in B C at this time. Only recently has the landscape had significant enough disturbance from M P B to return to historic levels of disturbance in the area, but o f course, this is at a much accelerated rate than would occur with fires burning with historical disturbance frequency, since it has occurred within one or two decades. Large fires during the period 1924-1933, combined with fire suppression after 1950, created a landscape o f mature, contiguous lodgepole pine stands that was highly susceptible to mountain pine beetle. In combination with climate changes the mountain pine beetle now is an epidemic which could last another decade or so (Eng, 2005). Because lodgepole pine regenerates easily after fire, however, large fires in the decade 1924-33 likely 43 created the conditions which lead to the current mountain pine beetle epidemic. These forests were all the same age and because of the large area of the fire, vast and contiguous. Regardless, many studies have pointed out the difference between harvesting and natural disturbances such as fire and mountain pine beetle; they are in fact not equivalent in their impacts on the landscape. Both fire suppression and harvesting changes forest species composition (Ward 2001, McGregor 2002, Rhemtulla et al. 2002), while harvesting also fragments habitat, causing loss of connectivity and interior habitat conditions (Sachs et al. 1998, Tinker et al. 2003). In the sub-boreal forest, the spatial patchiness of fires is lost by clearcutting activities, and important forest structure (coarse woody debris, snags, wildlife trees, and remnant living trees/stands) is removed. Similarly, in mountain pine beetle killed stands there can be significant regeneration or mixtures of other non-pine species within the stand, again unlike that which is left by certain harvesting activities such as clearcutting. Thus, while this study demonstrates that in stands with leading species such as spruce we may be close to emulating disturbance with harvesting in terms of the historical disturbance frequency, it does not suggest that harvesting is equivalent to fire or mountain pine beetle in terms of landscape-level impacts. 2.4.3 Range of variabil i ty and management Using the range of variability as a baseline condition for management is a coarse filter approach to management (Landres et al., 1999), and thus cannot account for all the variability in a system at all temporal or spatial scales. Most foresters and ecologists recognize that ecosystems are not stable, but rather dynamic, systems (Wu & Loucks, 1995), and therefore are not always maintained by the repetition of events on a predictable cycle. Our policy in BC manages as if the system is static by quite often recommending only one policy target value rather than a distribution of targets. This study shows that if these targets are used in combination with other measures, such as the variability of the indicator over time, a manager may gain further insight into the dynamics of the system, without sacrificing the simplicity of having management targets. If these targets are not revisited from time to time as new information becomes available, however, they may not achieve the desired forest condition outcomes. In this study, only the most up to date targets (the Morice SFMP) were close to reflecting the historical conditions on the landscape indicated by this research. 44 2.4.4 Fire suppression Recent scientific literature suggests that, in the boreal forest at least, changes in climate patterns, rather than fire suppression, were responsible for the decreased disturbance frequencies observed in many parts of Canada (Bergeron et al. 2001, Bridge et al. 2005). Although climate was not directly addressed in this study, the data here supports this contention, up to a point. Certainly, disturbance intervals, became longer around 1874, causing a flattening o f the line in Figure 2-2. Between 1944 and 1994 however, fire suppression was also apparent as the return intervals became even longer until the mountain pine beetle epidemic. Despite the large area burned 1914-1954 caused by humans (Figure 2-7), the disturbance frequency was relatively stable, and i f fire suppression occurred, counterbalanced the effects any increase in fire activity. First Nations burning likely also contributed to the overall- disturbance regime, but intentional fires by Aboriginal people were discouraged as European settlers moved in (Gottesfeld 1994). A s already pointed out, both the effects o f fire suppression which are evident in this study area, in combination with a harvest rate well below the disturbance rate o f forest fires likely led to the mountain pine beetle epidemic. Barclay et al. (2005) came to a similar conclusion using a Monte Carlo simulation o f fire and a mountain pine beetle susceptibility model. These authors studied the "traversability" o f the landscape in part in the Lakes T S A (as well as the Merritt T S A ) by the M P B and found that longer return intervals (because o f fire suppression) increased the traversability of the landscape. Species composition may well be affected in the study area (Rakochy, 2005), unless a fire occurs (McCullough et al., 1998) thus once again favouring the pine regeneration. In addition, unless a sudden cold snap kills beetles widespread though, our forests wi l l soon be beyond the range o f variability in disturbance as would have occurred historically. 2.4.5 Assumptions of this study and future improvements In this study, it was assumed that the landscape-level species composition and habitat types did not change with time. This was necessary, since historical information was not yet available, but there are limitations to this assumption. In the most recent few decades some areas with interior spruce or deciduous species have been cut and replanted with lodgepole pine, the preferred commercial species. This study would therefore overestimate the amount of disturbance in lodgepole pine stands, and underestimate disturbance in other cover types i f this were the case and the above assumptions were incorrect. 45 If cover types were not static in time, and all eligible pine stands (those 30 years or less currently, that had a historical disturbance) were converted, 13% o f the disturbances in pine stands would have been o f other cover types (data not shown). This number is very high and would reduce the historical disturbance level to 17% of pine stands. Considering that pine regeneration is highly favourable following fire (Lotan & Critchfield, 1990) and that most o f the pine indicator targets calculated in this research were more conservative than previous estimates, the necessary assumptions o f this study do not seem unreasonable. Presuming that burn severity, which affects regeneration (Ryan, 2002; Epting & Verbyla, 2005; Mclntire et al., 2005; Johnstone & Chapin III, 2006), was consistent during this time, pine would be favoured. In Alaska, Epting and Verbyla (2005) found that most stands regenerating after one large fire in 1986 were the same species as pre-fire, for both coniferous and broadleaf species. Further research is required to test whether this holds true for this study area. In this study, it was also assumed that age could substitute as time since disturbance in order to supplement the historical disturbance data. The first problem with this assumption is that the age data may not be accurate. This issue is discussed in further detail in Chapter 5, but indeed, an accuracy assessment of the data showed a consistent upward bias in the inventory. . This means that there may have been fewer disturbances in the historical time period than reported here. Another problem is linked to our assumption above that the current cover types were present in the same place on the historical landscape. I f there has been a transition in species composition in the stand, from, for example, deciduous to pine, or pine to spruce, the stand may have originated at a disturbance point which is older than reported in this study. We recommend therefore that future research consider creating a time since disturbance map for the area, using other methods such as dendrochronology, in combination with a comprehensive archive search for other historical records o f fire. This will help overcome many o f the data issues currently faced by landscape level researchers throughout the province. 2.4.6 Implications for management The implications of this study are that targets must be calculated at finer scales, especially when determining policy recommendations. Targets are generally specific to ecosystems rather than species. For example, the target disturbance return interval for the entire study area was 93, 100, or 125 years, depending on the policy document consulted. Figure 2-11 (b) demonstrates that historic disturbance return intervals for treed stands within the study area ranged from 82 years for deciduous stands, to 167 years for interior spruce stands. B y having an average as a policy 46 target, rather than species-specific targets, it allows managers to harvest (the dominant mode of disturbance currently) more in one species than another, but still meet such an average target. Studies have shown that stands outside their range of variability can become more susceptible to disturbances such as fires or pests (Moore et al., 1999). Policies that are not species-specific therefore, ignore the potential imbalance which can occur when harvesting takes place in only commercially preferred stands. This is an issue in the study area. Deciduous stands, which are not commercially preferred at this time, are older today than historically (Figure 2-15). In addition, criticism of the forest industry has also pointed out that this problem may be one cause o f the M P B epidemic. Lodgepole pine was not a commercial species until the past 20 years, meaning very little was harvested in the 60's and 70's. In part this lead to an abundance o f mature pine on the landscape, despite the increased harvest levels in the 80's and 90's. 2.5 C O N C L U S I O N Several management implications arise from this research. First, managers and policy makers alike must recognize that, in forests with short disturbance return intervals and high disturbance frequency, harvesting alone cannot mimic the effects of fire. Even when targeted to all species within an area, harvesting activities do not occur at the same rate (on average) or with the same temporal variability as fire activity. Allowing some fires to burn is required, perhaps by using zoning techniques and the triad approach are increasingly being recommended (see, for example, Le Gof f et al. (2005)). This approach is to use fire risk zones to plan fire suppression in high risk zones or fire monitoring activities in low risk zones. A significant obstacle to using prescribed burning, however, will continue to be resistance from industry concerned about losses of merchantable timber, as well as local communities concerned about safety. Second, policy targets used for landscape level planning must be reviewed as new information becomes available. These targets should be set at the scale of cover type so that management activities are not concentrated in the less commercial species, which could lead to these stands, at a landscape-level, being older and less healthy than historically. A n abundance of older stands can lead to insect and disease epidemics, such as the M P B , and a change in policy may mitigate such problems. Finally, policy targets should also be revisited given a changing climate, which will affect the disturbance regime o f a particular landscape. 47 2.6 TABLES Table 2-1 Description of the strata used in this study Data layer Description Function in model Cover type Vegetation description. For example, leading species. Describes landscape condition Age class Age classes on the landscape. Describes landscape condition Landscape unit Administrative boundaries, planning units, etc. Ecological boundaries, such as ecosystems or wildlife habitat. Describes zoning Habitat types Describes zoning Special management areas Discrete planning units. For example, could be old-growth management areas, caribou corridors, etc. Describes zoning Table 2-2 Description of disturbance and landscape indicators used to characterize the RV in disturbance, including the methods used to calculate the indicator, and the associated indicator target from policy documents with which to compare. If the policy target column is blank, then no policy target was available. Indicator type Indicator Name Calculation Indicator Targets from Policy Landscape 1 . % a rea by cove r type a r ea occup ied by cover type / total vegetated Cot tonwood : > 1 % , A s p e n : > 4 - 1 0 % , Lodgepo le p ine: > 37-Disturbance 2. % study a rea by habitat type 3. % habitat type by cove r type 4. % early" serai forest by cove r types 5. % a rea disturbed 6. % a r e a d isturbed by : a) d is turbance cause b) c o v e r type c) habitat types 7. % area disturbed ove r t ime, by dis turbance type 8. D is turbance return interva ls by cove r type 9. - Distr ibution of return interva ls over t ime 10. Decada l d is turbance f requency by: a) cove r type b) habitat type 11 . D is turbance type by cove r type a r ea * 100 5 5 % , interior & Enge lmann spruce : > 2 3 - 2 8 % ' a rea of each habitat type / total of the study a r ea *100 a rea of each cove r type / total of each habitat type a rea * 100 a rea of cove r type 1-40 years of age / total High biodiversity emphas i s units ( BEU ) : < 2 5 % 2 , < 4 0 % 2 , < 5 0 % 1 or < 5 4 % 3 , Intermediate B E U : < 3 2 % 2 , < 5 4 % 2 , or < 4 0 % 3 , Low B E U or the genera l forested a rea : 2 5 - 5 0 % 4 , < 5 4 % 2 , or < 6 4 % 1 area of that c o v e r type *100 a rea disturbed / total vegetated area *100 a rea disturbed by disturbance c ause / total a rea d isturbed *100 a rea disturbed by cover type / total a rea of cove r type *100 a rea disturbed by habitat type / total a rea of habitat type *100 a rea disturbed each decade by disturbance type / total vegeta ted area disturbed *100 1 / (area d isturbed by cove r type / total a rea of B lack spruce 250 yea r s 1 Al l other spec i es 93 years A l l spec i es ( including black spruce) 125 y e a r s 3 or 100 y e a r s 4 cove r type / # of years) 0 1 1 / (area d is turbed by d e c a d e / total forested a rea / # of y e a r s ) & a rea disturbed by cover type / total a rea of A l l spec ies , 0 .75-1 .25% per year , or 7 .5-12 .5% per d e c a d e 4 cove r type / # decades* a r ea disturbed by habitat type / total a rea of c o v e r type / # decades* A r e a by cove r type / area of each d is turbance type * 100 1 Mor ice T S A S F M P , or L R M P 2 L akes T S A H L P 3 B C Biodivers i ty G u i d e b o o k 4 De long (2002) 4^  2.7 FIGURES 127WW 126 W W 125WW 124°ffO"W Figure 2-1 Location of study area in BC. The outside boundary of the study area, shown in white, follows the dry cool Sub-boreal Spruce biogeoclimatic zone (SBSdk). Other ecosystems within this boundary were excluded. Dark gray areas are lakes and streams. Lighter gray areas are ecosystems outside of the study. 50 0.000001 1994 1974 1954 1934 1914 1894 1874 1854 1834 1814 1794 1774 1754 1734 1714 1694 1674 Decade Figure 2-2 Cumulative % of study area by year of last disturbance. Year of last disturbance (x axis) was determined from stand age data in the Vegetation Resources Inventory. Cumulative % of the study area (y axis) was calculated by dividing the area in each year of disturbance by the total study area, which is why the values in 2004 are 1. The y axis is plotted in logarithmic scale. The solid black dots are located at the beginning of the historical time period (1904) and the current time period (1954). 51 / % Study Area by Cover Type Deciduous Subalpine fir 51% Figure 2-3 Percent of the study area by cover type for. the current landscape (Indicator 1). Calculated from the 2003 Vegetation Resources Inventory data. % Study Area by Habitat Types Wetland, shrub Figure 2-4 Percent of the study area by habitat types for the current landscape (Indicator 2). Calculated from the 2003 Vegetation Resources Inventory data. 52 Figure 2-5 Distribution of cover types amongst the 4 vegetated habitat types for the current landscape (Indicator 3). 1904-1953 T3 O ' J -a 01 E 1954-2004 % F o r e s t e d A r e a D is turbed 0 1904-1953 • 1954-2004 33% 33% 0% 10% 20% 3 0 % 40% 5 0 % % of the study area disturbed 60% 70% Figure 2-6 Percent area disturbed historically (1904-1953) and currently (1954-2004) (Indicator 4). The x axis is the % of area disturbed in the vegetated portion of the landbase divided by the total area of the vegetated landbase. Cause of disturbances 1904-1953 Cause of disturbances 1954-2004 Figure 2-7 Cause of disturbances historically (1904-1953) and currently (1954-2004) (Indicator 5a). Values are expressed as a % of the total area disturbed each time period. 54 "2 28% -, Decade • Fire • Har\«st • MPB o Unknown Figure 2-8 Distribution of % area disturbed over time, by disturbance type (Indicator 6). Area disturbed each decade (x axis), is expressed as a % of the total vegetated area. Four different disturbances are shown: fire (black bars), harvests (dark gray bars), MPB (white bars), and unknown (light gray bars). m 1904-1953 • 1954-2004 0% 10% 20% 30% 40% 50% 60% 70% %area of cover type disturbed Figure 2-9 Percent area disturbed (y axis) by cover type (x axis) 1904-1953, and 1954-2004 (Indicator 5b). Percent area disturbed was calculated by dividing the total area disturbed in each cover type by the total area of that cover type. 55 Ei 1904-1953 • 1954-2004 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% % of area disturbed in each habitat type Figure 2-10 Percent area disturbed (y axis) by habitat type (x axis) 1904-1953, and 1954-2004 (Indicator 5c). Percent area disturbed was calculated by dividing the total area disturbed in each habitat type by the total area of that cover type. 56 Figure 2-11 Average decadal disturbance frequencies (a) and average decadal return intervals (b) by cover type for the two time periods, historical (1904-1953) and current (1954-2004) (Indicators 7 and 9a). Average decadal disturbance frequencies are the % area disturbed in each cover type divided by 4.9 decades for the historical time period, and 5.0 decades for the current time period. Average disturbance return interval is the inverse of the annual disturbance frequency (% area disturbed divided by 49 or 50 years respectively). 5 7 800 700 600 ra 500 i t 400 3 300 200 100 j 1 Historic Return Intervals I Current Return Intervals 233 1 3 4 140 142 1 2 g I I | | Upland, treed Wetland, treed Upland, shrub Habitat type Wetland, shrub Figure 2-12 Average decadal disturbance return intervals and frequencies by habitat type for the two time periods, historical (1904-1953) and current (1954-2004) (Indicator 9b). Average disturbance return interval is the inverse of the annual disturbance frequency (% area disturbed divided by 49 years for the historical time period or 50 years for the current time period). 58 2500 j ! , 2000 w £ 1500 c c E 1000 500 0 —•— Pine —•— Spruce Deciduous Subalpine fir \ \ % \ \ \ w \ % Decade Figure 2-13 Average annual disturbance return intervals by cover type over time (Indicator 5). Not all cover types were graphed; black spruce and shrub cover types had sporadic disturbance patterns and thus in some decades very high return intervals. Figure 2-14 Area of each cover type by disturbance type (Indicator 11). Values are expressed as a % of the total area of each disturbance type. 59 70% 60% \ n> 5 50% in ro in ^  « >. » 30% H ro ro 20% 10% 0% 50% 28% 27% 27% 35% Cover type 5 2 % 1914-1953 1954-2004 12% Of 3 0 % 2^ Figure 2-15 Percent of the area of each cover type in early serai stage (Indicator 3). Shrubs were excluded. For the historical time period, early serai stands were identified as those which were disturbed during the 49 year time period. 60 b) Cover type Lakes Higher Level Plan 1 1 9 1 4 - 1 9 6 3 I 1 9 5 4 - 2 0 0 4 Intermediate BEU RMZ A & Low B E U RMZ B <High B EU RMZ A Cover types BC Biodiversity Guidebook (BCBG) & Natural Disturbance Units (Delong 2002) I 1 9 1 4 - 1 9 5 3 I 1 9 5 4 - 2 0 0 4 BCBG intermediate B EU | B C B G high BF.tl \ f / Delong (20021 c) X C o v e r types Figure 2-16 Policy recommendations for % area in early serai stage by: a) Morice SFMP, b) Lakes Higher Level Plan, and c) Biodiversity Guidebook and Natural Disturbance Units (Delong 2002). All recommendations are maximum thresholds, with the exception of Delong (2002) who recommends a minimum and a maximum %. 61 2.8 R E F E R E N C E S Barclay, H . J., L i , C , Benson, L . , Taylor, S., & Shore, T. (2005). Effects o f return rates on traversability o f lodgepole pine forests for mountain pine beetle (coleoptera: Scolytidae) and the use o f patch metrics to estimate traversability. Canadian Entomologist. 137(5): 566-583. B C Ministry o f Environment, Lands and Parks. (2000). Lakes District Land and Resource Management Plan. Victoria, B C : B C Ministry of Environment, Lands and Parks. B C Ministry o f Forests. (1995). Forest Practices Code of British Columbia, Biodiversity Guidebook. Victoria, B C : B C Ministry of Forests. B C Ministry o f Forests and Range. (2004). Forest Range and Practices Act (FRPA). Retrieved May 16, 2006 from http://www.for.gov.bc.ca/tasb/legsregs/irpa/frpa/ffpatoc.htm B C Ministry o f Sustainable Resource Management. (2004). Morice Land & Resource Management Plan. Victoria, B C : B C Ministry of Sustainable Resource Management. Bergeron, Y . , Gauthier, S., Kafka, V . , Lefort, P., & Lesieur, D . (2001). Natural fire frequency for the eastern Canadian boreal forest: Consequences for sustainable forestry. Canadian Journal of Forest Research. 31: 384-391. Bergeron, Y . , Leduc, A . , Harvey, B . D . , & Gauthier, S. (2002). Natural fire regime: A guide for sustainable management of the Canadian boreal forest. Silva Fennica. 36(1): 81-95. Bridge, S., Miyanishi, K , & Johnson, E . (2005). A critical evaluation of fire suppression effects in the boreal forest of Ontario. Forest Science. 51(1): 41-50. Bunnell, F. L . (1995). Forest-dwelling vertebrate faunas and natural fire regimes in British Columbia: Patterns and implications for conservation. Conservation Biology. 9(3): 636-644. Delong, S. C. (2002a). Natural Disturbance Units of the Prince George Forest Region: Guidance for Sustainable Forest Management. Prince George: B C Ministry of Forests and Range. Delong, S. C. (2002b). Using nature's template to best advantage in the Canadian boreal forest. Silva Fennica. 36(1): 401-408. 62 Delong, S. C. (1998). Natural disturbance rate and patch size distribution o f forests in northern British Columbia: Implications for forest management. Northwest Science. 72: 35-48. Eng, M . A . (2005). Provincial-level projection of the current mountain pine beetle outbreak an overview of the model (BCMPB V2) and results of year 2 of the project. Victoria, B C : B C Ministry o f Forests. Epting, J., & Verbyla, D. (2005). Landscape-level interactions of the prefrre vegetation, burn severity, and postfire vegetation over a 16-year period in interior Alaska. Canadian Journal of Forest Research. 35: 1367-1377. Fall, A . and Fall , J. (2001). A domain-specific language for models o f landscape dynamics. Ecological Modelling. 141: 1-18. Fulton, F. J. (1910). Royal Commission of Inquiry on Timber and Forestry. Victoria, B C : Province o f British Columbia. Gottesfeld, J. (1994). Aboriginal burning for vegetation management in northwest British Columbia. Human Ecology. 22(2): 171-188. Haeussler, S., Bedford, L . , Leduc, A . , Bergeron, Y . , & Kranabetter, J. M . (2002). Silvicultural disturbance severity and plant communities of the southern Canadian boreal forest. Silva Fennica. 36(1), 307-327. Heitzmann, R . J . Identifying human ignited fires in the central Canadian Rockies over the last millennium. A Collection of Papers Presented at the 33rd Annual Meeting of The Canadian Archaeological Association. J . -L. Pilon, M . W . Kirby & C. Theriault. The Ontario Archaeological Society. Retrieved Aug . 6, 2006 from http://epe.lac-bac.gc.ca/100/200/300/ont_archaeol_soc/annual_meeting_caa/33rd/heitzmann.pdf Johnson, E . A . , & Larsen, C. P. S. (1991). Climatically induced change in fire frequency in the southern Canadian Rockies. Ecology. 72( 1): 194-201. Johnstone, J . , & Chapin III, F. S. (2006). Effects o f soil burn severity on posit-fire tree recruitment in boreal forest. Ecosystems. 9: 14-31. 63 Karjala, M . K . & Dewhurst, S. M . (2003). Including aboriginal issues in forest planning: A case study in central interior British Columbia, Canada: Landscape and Urban Planning. 64: 1-17. Kuuluvainen, T. (2002). Natural variability of forests as a reference for restoring and managing t biological diversity in boreal Fennoscandia. Silva Fennica. 36(1): 97-125. Landres, P. B . , Morgan, P., & Swanson, F. J . (1999). Overview of the use o f natural variability concepts in managing ecological systems. Ecological Applications. 9(4): 1179-1188. Le Goff, FL, Leduc, A . , Bergeron, Y . , & Flannigan, M . (2005). The adaptive capacity of forest management to changing fire regimes in the boreal forest o f Quebec. The Forestry Chronicle. 81(4): 582-592. Lotan, J . , & Critchfield, W. (1990). Pinus contorta dougl. ex. loud., Lodgepole Pine. Silvics of North America. 1:302-315. McCullough, D . G. , Werner, R. A . , & Neumann, D . (1998). Fire and insects in northern and boreal forest ecosystems o f North America. Annual Review of Entomology, 43, 107-127. McGregor, M . (2002). Developing Reference Conditions for Ecological Restoration of Forests in Sub-boreal British Columbia. Thesis submitted in partial fulfillment o f the requirements for the degree o f Master of Science in Natural Resources and Environmental Studies, University of Northern British Columbia. Prince George, B C : University of Northern B C . pp. 149. Mclntire, E . J . B . , Duchesneau, R. , & Kimmins, J . P. H . (2005). Seed and bud legacies interact with varying fire regimes to drive long-term dynamics o f boreal forest communities. Canadian Journal of Forest Research. 35 (11): 2765-2773. Millar, C. I., & Woolfenden, W . B . (1999). The role of climate change in interpreting historical variability. Ecological Applications. 9(4), 1207-1216. Moore, M . M . , Covington, W . W. , & Fule, P. Z . (1999). Reference conditions and ecological restoration: A southwestern ponderosa pine perspective. Ecological Applications. 9(4): 1266-1277. 64 Morice and Lakes IFPA. (2005). Morice TSA Sustainable Forest Management Plan. Prince George, B C : Morice and Lakes IFPA. Noss, R .F . , LaRoe III E .T . and Scott J . M . (1995). Endangered ecosystems of the United States: a preliminary assessment o f loss and degradation. National Biological Service Biological Report 28. U.S. Department of the Interior, Washington, D .C . , U S A . Perry, G . L . W. (2002). Landscapes, space and equilibrium: Shifting viewpoints. Progress in Physical Geography. 26(3): 339-359. Province of British Columbia. (2000) Order Establishing Resource Management Zones and Resource Management Zone Objectives for the Lakes District Pursuant to Sections 3(1) and 3(2) of the Forest Practices Code of British Columbia Act (the Act), 12434-25/LAKES (July 26, 2000). Retrieved Aug. 6, 2006 http://ilmbwmv.gov.bc.ca/ihrib/lup/lnTip/northern/lakes/index.html Province of British Columbia. (1997). Vanderhoof Land and Resource Management Plan. Retrieved May 16: 2006: http ://ilmbwww. go v.bc. ca/ihnb/lup/lrmp/northern/ vanderhf/toc .htm Rakochy, P. I. (2005). Lodgepole stand dynamics a result of mountain pine beetle in the central interior of British Columbia. Thesis submitted in partial fulfillment o f the requirements for the degree of Master o f Science in Natural Resources and Environmental Studies, University o f Northern British Columbia. Prince George, B C : University o f Northern B C . Rhemtulla, J. M . , Hall, R. J. , & Macdonald, S. E . (2002). Eighty years o f change: Vegetation in the montane ecoregion o f Jasper National Park, Alberta, Canada. Canadian Journal of Forest Research. 32:2010-2021. Russell, E . W . B . (1997). People and the land through time. New Haven, CT: Yale University Press. Ryan, K . C. (2002). Dynamic interactions between forest structure and fife behaviour in boreal ecosystems. Silva Fennica, 3(5(1), 13-39. 65 Sachs, D . L . , Sollins, P., & Cohen, W. B . (1998). Detecting landscape changes in the interior of British Columbia from 1975 to 1992 using satellite imagery. Canadian Journal of Forest Research. 28: 23-26. Sanborn, P., Geertsma, M . , Jull, A.J .T. , and Hawkes, B . 2006. Soil and sedimentary charcoal evidence for Holocene forest fires in an inland temperate rainforest, east-central British Columbia, Canada. The Holocene. 16(3): 415-427. Sloan, G . M . (1945). Report of the Commissioner Relating to the Forest Resources of British Columbia. Victoria, B C : Charles F. Banfield. Sloan, G . M . (1956). Report of the Commissioner Relating to the Forest Resources of British Columbia. Victoria, B C : Don McDiarmid. Swetnam, T. W., Allen, C . D . , & Betancourt, J. L . (1999). Applied historical ecology using the past to manage for the future. Ecological Applications. 9(4): 1189-1206. Taylor, S. W., & Carroll , A . L . (2004). Disturbance, forest age, and mountain pine beetle outbreak dynamics in B C : A historical perspective. Proceedings: Mountain Pine Beetle Symposium: Challenges and Solutions. October, 30-31, 2003. 11 pp. Tinker, D . B. , Romme, W . H . , & Despain, D . G. (2003). Historic range o f variability in landscape structure in subalpine forests o f the greater Yellowstone area, U S A . Landscape Ecology. 18:429-439. V a n Wagner, C. (1978). Age-class distribution and the forest fire cycle. Canadian Journal of Forest Research. 8: 220-227. Ward, T. (2001). Sustainability for British Columbia forestry. Journal of Sustainable Forestry. 12: 153-169. Weir, J. M . H . , & Johnson, E . A . , & Miyanishi, K . (2000). Fire frequency and the spatial age mosaic of the mixed-wood boreal forest in western Canada. Ecological Applications. 10(4): 1162-1177. 66 Wong, C , & Iverson, K . (2004). Range o f natural variability: Applying the concept to forest management in central British Columbia. BC Journal of Ecosystems and Management. 4(1): 1-14. W u , J., & Loucks, O. L . (1995). From balance of nature to hierarchical patch dynamics: A paradigm shift in ecology. The Quarterly Review of Biology. 70(4), 439-466. 67 C H A P T E R 3 A H I E R A R C H I C A L S I M U L A T I O N - T H R O U G H - O P T I M I Z A T I O N A P P R O A C H T O F O R E S T D I S T U R B A N C E M O D E L L I N G 6 3.1 INTRODUCTION Over the past decade or so, multiple authors have suggested that managing forests by using harvesting to emulate natural disturbance regimes is one possible method of sustaining biodiversity while also harvesting in an ecological manner, thereby achieving sustainable forest management (Boychuck et al. 1997, Mitchell et al. 2002, Wei et al. 2003). The evidence suggests that by emulating or mimicking natural disturbances using silviculture activities, the inherent variability of ecosystem conditions is maintained within the adaptive limits of all species in that ecosystem (Bergeron et al. 2002). In British Columbia, the forest industry is mandated by legislation under the Forest Range and Practices Act to manage timber harvesting areas under such a paradigm of natural disturbance. Our current view of disturbance is one of nonequilibrium; considering disturbance processes as stochastic, changing over space and time. Contrary to equihbrium concepts common 30 years ago, disturbances are no longer thought of as a balance of growth, death and renewal processes (Wu and Loucks 1995). The concept of natural range of variability (Landres et al. 1999) is one method of characterizing natural disturbance regimes which incorporate the inherent variability of the processes involved. Disturbances are quantified over a predefined time period, often where human influences were either relatively small or absent. This variability of disturbances can be integrated into planning tools such as forest management models, which allow stochastic representation of events using probability theory and other statistical methods (Klenner et al. 2000, Fall and Fall 2001). The application of stochastic landscape-level disturbance models have been demonstrated for a number of research problems including timber supply (Bonar et al. 2003, Armstrong 2004, Peter and Nelson 2005), wildlife habitat (Klenner et al. 2000), climate change (He et al. 2002), natural disturbance dynamics (Boychuck and Perera 1997, Boychuk et al. 1997, Fall et al. 2004, Yang et al. 2004) and planning for fife suppression resources (Wotton and Martell 2005). These models simulate disturbances on the landscape using rules based on mathematical formulations and algorithms which describe the initiation and propagation of the disturbance through time 6 A version of this chapter has been accepted for publication . Campbell, K . A . , and Dewhurst, S.M. A hierarchical simulation through optimization approach to forest disturbance modelling. Ecological Modelling. .68 and/or space. The element of chance arises in these models when disturbance is initiated and spreads randomly or probabilistically. In the previously referenced research, study area extent and grain ranged from about 60,000 ha with 2 ha resolution (Klenner et al. 2000) to 3.5 million ha with 14 ha resolution (Fall et al. 2004). Because disturbances are highly dependent on spatial and temporal scale (Turner et al. 1993), it is necessary to evaluate how models are also affected by scale. There is considerable debate on how to determine the appropriate grain and extent when studying ecological processes such as disturbance. For example, Lertzman and Fall (1998) point out that these two components are positively correlated- as the extent is increased, so often is the grain. In landscape level studies we often view trends over a large extent and a coarse grain, but may be missing finer grain processes that create landscape level patterns. This tradeoff is often necessitated by the finite speed of computers, even with modern advances. The more complex the problem is, the longer it will take to simulate the entire possible range o f events and state changes on the landscape. Heuristic optimization techniques (also known as metaheuristics), such as simulated annealing, are recognized as useful methods for overcoming unwieldy processing times that are a result of exact optimization models, such as integer programming, in timber supply (Lockwood and Moore 1993, Nelson 2003, Bettinger and Chung 2004). Such techniques are common in forest management modelling, especially for timber supply projections (Nelson 2003, Peter and Nelson 2005), but have not before, to our knowledge, been used to model stochastic disturbances at the landscape level. We speculate that heuristic optimization wi l l have similar advantages over simulation approaches for disturbance modelling. We expect that there wil l be a decrease in processing time, despite increased complexity in the problem (similar to the advantage held by heuristic optimization over exact optimization techniques), and believe such techniques will facilitate the direct integration o f stochastic disturbance events into an existing forest management modelling framework. Our approach is not to directly model the process o f disturbance itself- i.e. initiation, spread, termination- but rather to model the temporal patterns o f the disturbance. We assume that we already have a general understanding o f the disturbance process, and the effects o f allogenic and autogenic environmental interactions. The model then replicates the observed temporal patterns using Markov chains. The objective o f this research was to use an existing stochastic-heuristic forest management model and test its ability to model complex forest systems at multiple spatial scales 69 using heuristic optimization techniques to simulate stochastic disturbance processes. This is accomplished through the use of a goal-programming type formulation combined with simulated annealing. We tested the model using data from a larger project in the central interior of British Columbia, studying the long-term spatiotemporal impacts o f both stand-replacing natural disturbances and management guided by current sustainable forest management policies. This paper describes: 1) parameterization of the model using historical disturbance probabilities for the study area and a Markov Chain approach, 2) testing the sensitivity of the model to changing these disturbance frequencies in the optimization framework, 3) exploration of the effects of scale on serai stage and disturbance by setting disturbance targets at three spatial scales in the model, and 4) the preliminary assessment of current S F M policy in the study area using historical disturbance data and the disturbance model. 3.2 M A T E R I A L S A N D M E T H O D S 3.2.1 Study area The study area is located in the central interior of British Columbia, covering approximately 1.2 million ha (Figure 3-1) primarily in the Morice, Lakes, and Prince George Timber Supply Areas. The most common ecosystem in the study area, the Sub-boreal Spruce dry cool ecosystem (SBSdk), occurs at lower elevations and, because o f a predominance o f disturbance, is characterized by early serai species such as lodgepole pine and trembling aspen. At higher elevations, wetter and cooler ecosystems exist which have overall fewer disturbances, lending the species composition to longer-lived hybrid spruce and subalpine fir. Most o f the study area is dominated by frequent stand-initiating disturbance events, but at the higher elevations, disturbances are much less prevalent. Stand-replacing forest fires and insect outbreaks are common, with a mean fire return interval o f 125 years (BC Ministry of Forests 1995). 3.2.2 The model We used a forest management model to test the feasibility o f modelling disturbances using optimization methods. The model is designed for a criteria and indicators framework, and has been used in a number of studies in both B C (Dewhurst and Karjala 2003, Karjala and Dewhurst 2003, Karjala et al. 2004) and Arizona (Dewhurst and Campbell, unpublished data). In addition, the model has been calibrated against provincial standards for timber supply planning (Dewhurst and Karjala 2003). 70 The model is implemented in a vector, rather than a raster environment, and operates at the level of strata. Individual spatial instances of the strata, based on vegetation and site characteristics create individual mapping and analysis units. The spatial instances o f the strata are equivalent to patches, a term often employed in landscape ecology to refer to cohesive spatial units. The model uses an object-oriented approach, and each strata or patch is an object within the model. Each object has a current state (e.g., undisturbed or disturbed), and a set of rules or procedures governing changes to that state. The model uses a message-passing-between objects design where procedures are called by passing a message to an object. Optimization in the model is conceptually similar to standard goal progranmiing, whereby the model attempts to find an optimal solution by inmiinizing the deviations of the landscape conditions from the target landscape objectives (multiple objectives in this case). The following objective function is mirnmized over all goals (g), zones (z), and time periods (p): tttH^h^D-J ( i ) h=\ 1=1 j=\ where: Phtj ~ m e preference weight associated with a deviation above goal h in zone / in time period j D+hij = the deviation above goal h in zone i in time period j Pfoj = the preference weight associated with a deviation below goal h in zone /' in time period j D~MJ = the deviation below goal h in zone i in time period j Goals and zones are set by the user. Goals are targets for indicators which are set in the model's user interface. Zones on the other hand, are input GIS layers that define static boundaries in the spatial data environment. Preference weights are also set by the user, but default to 1. The weights can be increased or decreased incrementally and interactively, while the model is running, for any indicator for any time period. This is usually done when the model is not converging on a particular target. Interactive setting of preference weights in this fashion allows for users to guide the model towards a solution which achieves the disturbance targets. Diaz-Baltiero and Romero (2004) discuss this concept of using preference weights to tell a goal programming-type model that "more is better" or "less is better", and steer it towards a desirable solution The model uses the stochastic heuristic optimization algorithm of simulated annealing (Lockwood and Moore 1993) to systematically find a near optimal global solution in the model. 71. Unlike exact optimization methods often used in goal programming such as the Simplex method, simulated annealing does not necessarily find a truly optimal solution to the problem, but it is more efficient in finding a near optimal solution for larger complex combinatorial problems (Lockwood and Moore 1993). The goal of simulated annealing is to search for improved solutions which systematically decrease the value of the objective function. A n inferior solution (increase in objective function) can be accepted, however, depending on the "temperature" of the model. The temperature of the model is essentially the probability that an inferior solution wil l be accepted, and this temperature "cools" as the optimization progresses. The role of temperature, and the rate at which the temperature cools, differentiates simulated annealing from simulated quenching. The fast cooling rates in simulated quenching have very low probability of inferior solution selection, and vice versa for simulated annealing. With a slower cooling rate, the likelihood o f finding an optimal or near optimal solution is much greater, as is the processing time required. Processing time versus the likelihood o f finding an optimal or near optimal solution is a fundamental trade-off for simulated annealing. Our experience has shown that simulated quenching, using very fast cooling rates, is adequate for simple problems, but that simulated annealing, using a slower cooling rate, is necessary for the more complex. 3.2.3 Parameterization of the model 3.2.3.1 Data inputs and Zoning The model requires five data layers, which describe landscape vegetation conditions, as well as the different levels of zoning. The information requirements, and their function in the model, are outlined in Table 3-1. In the model, targets (or disturbance events, in this case) may be specified in each zone, as well as for the landscape as a whole. The zones are static boundaries, which do not change with time. Based on treatments, landscape conditions can change: species composition may be altered, and stands age incrementally until another disturbance or harvest event when age is reset to 1. The model is hierarchical in the sense that zones are nested at increasingly finer levels of detail. In this way, zones can also be thought of as different scales. A t the broadest level, the forest level, the zone is the entire landscape. At the next levels, habitat type, landscape unit, and special management areas, zones are defined based on increasingly finer detail, representing smaller and smaller aggregations or patches. For example, habitat type in this study is defined by slope and leading vegetation type (tree, shrub, no vegetation). Within 72 each habitat type there can be any number of leading species which define the boundaries of the landscape units within the habitat types (Figure 3-2). The former B C Ministry of Sustainable Resource Management supplied the Vegetation Resource Inventory data for this study, from which all input data were derived (Table 3-2). The dominant leading species on the landscape was lodgepole pine (Pinus contorta var. latifolia). Interior spruce (Picea glauca x engelmannii), aspen (Populus tremuloides) and Cottonwood (Populus trichocarpa) were also abundant. Stand age ranged from 1 to about 400 years. In areas affected by mountain pine beetle, the stand age of lodgepole pine in the inventory was corrected to the year when attack levels were greater than 30%. Landscape units in this case were the same as the vegetation cover to enable us to set targets for disturbances by cover species. Special management areas were not used in this analysis. 3.2.3.2 Indicators In order to set targets for the goals which comprise the objective function indicators are defined. Indicators can be real numbers or binary measures (true/false) of landscape condition. Indicators are tied to vegetation type, habitat type, and age (or time since disturbance). We set up 6 different indicators for our model runs (Table 3-3). For serai stage indicators, we have defined serai stages following the British Columbia standard method, based only on stand age, rather than structural characteristics, which are often implied in other definitions of serai stage. It should be noted that since we have 1 shrub cover type and 2 shrub habitat types, the concept o f serai stage is loosely applied to non-treed landscapes. In these complexes, serai stage becomes an indicator simply of time, that is, how long the shrub types go undisturbed, rather than any implication o f stand structure or composition transitions. At the start o f the model runs shrub age is always 0 since age data is not available for shrub vegetation types. Trees do exist on some of the habitat types classified as shrubs. These stands would have been assigned an age, and therefore a serai stage in the model. The user sets targets in the model relative to indicator, zone, and each time period. For binary indicators, the target units are in percentages. For non-binary indicators, such as cost per hectare, the target units are real values. Indicator targets can be set for every time period in the analysis, and at all levels of zoning, including the forest level. 73 3.2.3.3 Disturbance The model acts as a harvest-scheduling model, which schedules management actions or disturbances each time period, and assesses the change in the objective function resulting from each treatment to determine whether to accept or reject that treatment. To model disturbances, we simply treat disturbances as harvesting in the model. We assume that fire, the stand-replacing disturbance being modelled, occurs regardless o f stand age (Fall et al. 2004). Using this method, the model required no structural modifications to simulate the outcomes o f disturbance processes on the landscape. This structure enables us to model any stand-replacing disturbance, even those, like mountain pine beetle, which might only affect stands of a certain age. Disturbances can also be modelled at the four levels of zoning allowing the user to either choose the appropriate scale for modelling disturbance or to explore the effect o f scale on disturbance processes. In using the indicator framework, the model implements disturbance in a similar manner to a simple Markov chain. Although a Markov chain is not spatial by definition, the model allows the spatial implementation of Markov chains, using optimization techniques. A t this time however, even though disturbances occur spatially, they are not clustered; disturbances occur randomly on the landscape. This allows us only to make inferences on the impacts of disturbance frequency, rather than the number of disturbances or specific disturbance sizes. Markov models typically characterize change over time and determine state changes on the landscape in a probabilistic manner (Urban and Wallin 2002). In a Markov model, each unit of the landscape exists in a particular state at a certain time with a given probability of state change at that point in time. In a simple first order Markov model, the future state of landscape unit is dependent only on the current state: x, + , = x,P • (2) where x t is the starting condition of the landscape at time step t, and P is a transition matrix containing the probabilities that a change wil l occur (Urban and Wallin 2002). As described above, the model is implemented in a vector environment, where each unit is a patch, and thus x is the state of any given patch. Because the model is object-oriented, each patch is treated as an object, which has a current state that changes to a future state with every timestep. This is similar to a traditional Markov chain, where the current state is tracked as a row vector, except that each patch is tracked as an object in this case. In this study, the values in matrix P were calculated using historical fire data over a 50 year time period 1904-1954. In British Columbia, one of our sustainable forest management 74 goals is to mimic the natural disturbance processes with harvesting. Typically, we use historical disturbance information to determine what the rate, timing and spatial patterns o f these disturbances would be in the absence o f human management. We wanted to explore the implications of a policy where harvesting occurred at the same rate, at least on average, as historical disturbances. We were able to do this by essentially returning disturbances (fire) to the landscape using our forest management model and simple Markov chains. Previous work by Campbell (unpublished) indicated the period 1904-1954 to be a reasonable time period to extract historical fire frequencies in the study area. Fire data were obtained from two sources. The primary source was the Canadian Forest Service's Natural Disturbance Database which contains spatial information on the extent of fires from 1914-1954. These data were supplemented with the Vegetation Resource Inventory data, which allowed us to identify stand-replacing disturbances from the stand age data, following the methods of Delong (1998). Probabilities were calculated for each o f the 3 spatial scales, or zones (which describe habitat types, or landscape units) as follows: 1. Calculate the proportion o f area disturbed in each zone over the entire temporal period o f the data. In this case the data span 50 years. 2. Average the proportion calculated above for the timestep used in the model. In this study the timestep was 10 years, so it was divided by 5. This then approximates the decadal probability of disturbance. These decadal then become targets for the disturbance indicator in the model, and essentially the transition matrix. We used the same target for every 10-year time period, which is a true Markov chain because disturbance is therefore time independent. However, because targets can be set for each time period individually, there is the potential to model the effect of changes in external processes (such as climate) on disturbance, by altering the probabilities of occurrence over time. Disturbances cannot occur twice in the same time period, but can occur in the same stand multiple times within the analysis period (150 years in this study). Revisiting equation 1, the objective function for each disturbance probability in each zone at each time step in the disturbance model is: i,=\ ;=i j=\ where: g = are the disturbance goals, or disturbance probabilities (transition matrix; Table 3-4) z = are the zones, which are habitat types and landscape units in this case (3) 75 p = are the 15 time periods Pffy = the preference weight associated with a deviation above goal h in zone / in time period j D^j = the deviation above goal h in zone / in time step j Phjj = the preference weight associated with a deviation below goal h in zone i in time period j D~hij = the deviation below goal h in zone i in time period j 3.2.4 Scenarios Four different scenarios were run in order to: 1) explore the sensitivity of the model to changes in the disturbance probabilities (Scenario 1), 2) determine the effect of setting disturbance probabilities at different scales on the on the model outcomes (Scenarios 2-4), and 3) examine the potential impact on the landscape i f disturbances were to occur at the same rate as historically (Scenarios 2-4). 3.2.4.1. Scenario 1, Forest Level Sensitivity Analysis We carried out a sensitivity analysis o f the model at the forest level by setting targets and goals at 9 frequencies o f disturbance ranging from no disturbance to 8% disturbance per decade. At a disturbance rate o f between 6% and 7% per decade, an area the size of the study area would be disturbed at least once over 150 years. The historical disturbance frequency was, on average, about 7% per decade, which is a return interval of about 130 years. Disturbance return intervals were calculated as the inverse of the annual disturbance probabilities (Delong 1998). We ran the model three times for each level of disturbance in the sensitivity analysis. We chose to only change the disturbance probability parameter within the likely range of forest-level disturbance in the study area rather than for all possible ranges of disturbance in all forest ecosystems or stand types. Some might argue that this is not a true sensitivity analysis. In part to counter this argument we also evaluated whether or not the modelled disturbance rates were in fact meeting our disturbance goal (the target goal is the disturbance probability or probabilities set for each run). Based on the sensitivity analysis, and a close examination of the results for each scenario we are able to determine if the model is working correctly. The model is transparent itself, and the user is able to monitor progress in real time as it converges to a final solution. We do however recommend that researchers who use our methods to model disturbance 76 in other forest management models also undertake their own sensitivity analysis within the context of their own research. 3.2.4.2 Scenario 2, Forest Level Historical Disturbance Targets For this scenario, we ran the model five times at the historical disturbance rate o f 7.4% (Table 3-4). In the boreal forest, which has similar disturbance regimes as the sub-boreal, many researchers have demonstrated that under a stand-replacing fire regime burning independently of age, the age class distribution follows a negative exponential or reverse ' J ' distribution. We hypothesized that the current age class distribution in the study area did not follow a negative exponential distribution because o f the combined effects o f management, and mountain pine beetle outbreaks. After 150 years under a historical disturbance regime (7.4% disturbed each decade) we hypothesized that the age class distribution might be returned to such a state. We explored our hypotheses using the graphical methods of Weir et al. (2000) where the age class distribution is plotted over time as cumulative percent of the study area on a logarithmic axis. A straight line on such a graph, is indicative o f a cohesive disturbance regime, represented by specific a disturbance frequency (Weir et al. 2000): Over time, a disturbance frequency can change, and these changes are identified as separate straight lines on the graph, and the time period that the change occurred is termed the change point. A s Weir et al. (2000) found, these changes can occur due to both human policy and management as well as influences such as climate change. We hypothesized the presence o f at least one change point between 1950 and 1960 in our study area, which corresponds to the approximate time when harvesting in the area began to dominate, and fire suppression began in earnest. We believed that the modelled age class distribution would become a straight line after 150 years, since we are modelling a disturbance frequency which does not change with time. 3.2.4.3 Scenario 3, Habitat Type Level Historical Disturbance Targets A t the habitat type level, targets were set based on mean historical disturbance rates which ranged from 4.3% to 8.1% per decade (Table 3-4). Return intervals ranged from 231 years to 129 years. The model was run at the habitat type level five times. 77 3.2 A A Scenario 4, Landscape Unit Level Historical Disturbance Targets At the landscape level we set targets only for those landscape units (cover types) which were greater than 1000 ha in size. We quickly found that setting targets for smaller areas slowed the model considerably. These areas make up less than 0.1% of the study area, and thus historical disturbance rates calculated for these areas might not be representative of these species types. Targets for all other landscape units were calculated from historical disturbance rates and ranged from 2.2% to 11.9% per decade (Table 3-4), while return intervals ranged from 451 to 84 years. Again, the model was run five times at the landscape level. 3.2.4.5 Comparison of Scenarios 2, 3, & 4 We compared the output from the scenarios 2, 3, and 4 described above, which had disturbance targets set at three different levels of zoning: forest, habitat type, and landscape unit. We were interested in how setting disturbance targets for one zoning level affected the outcomes at the other zoning levels (where targets were not set), and whether or not the values were comparable. We speculated that the indicator outcomes would be different between these three scenarios, since disturbance would occur independently of the zoning levels without targets, but would be constrained at the zoning level with targets. We used Analysis of Variance (ANOVA) to test for differences in indicator values in the last time period between the three zoning levels. 3.3 RESULTS . 3.3.1 Scenario 1, forest level sensitivity analysis Results from the sensitivity analysis showed consistent model behaviour with increasing levels of disturbance (Figures 3-3 and 3-4). Disturbance was, under all disturbance frequencies, constant over time, while cumulative disturbance increased. An increase in the rate of disturbance per decade correspondingly increased the overall cumulative disturbance. As disturbance increased, so did the amount of early and young forest, while the amount of mature and old forest decreased. It was interesting to note that steady state conditions occurred for all 4 serai stage indicators. Steady state conditions typically are a manifestation of a Markov chain (Urban and Wallin 2002) where subsequent changes on the landscape, in this case disturbance, no longer result in a change in the amount of area for a particular indicator. The time when steady state 78 conditions were achieved in early and young serai stages corresponded to the time when the youngest stands at the start o f run made the transition into the young serai stage (time period 4) and then the mature serai stage (time period 10). Because mature and old serai stages lose stands only to disturbance, and never to aging based on their indicator definition, they reached steady state conditions when the youngest stands made the transition into the mature serai stage (time period 10) and old serai stage (time period 13). Thus in this model, steady state conditions clearly are highly dependent on the condition of the landscape at the beginning of the model run. It should be noted that although the indicator curves 'flatten out' indicating some level of steady state, these curves are not perfectly straight and variation wil l still occur over time. 3.3.2 Scenario 2, forest level historical disturbance targets At the historical rate of disturbance, 7.4%, cumulative disturbance reached 1,181,728 ha after 150 years (period 15) (Table 3-5). This is very close to the size of entire study area (1.2 million ha), but some stands were disturbed as many as 6 different times over the entire analysis period. After 150 years, the amount o f area in early serai stage decreased by only 2%, while the amount of area in young serai stage increased by 9%. Mature serai stage decreased the most by 26%, while old serai stages increased slightly by 2%. Two change points were evident from our graph of the current age class distribution, at roughly 60 years, and 120 years (Figure 3-5). After the first 50 years however, three change points can be seen at 50 years, 110 years, and 170 years. The latter two corresponding to the effects o f the distribution o f age classes at the beginning of the run, while the first new change is the effect o f reintroducing disturbances at historical levels. Three change points remain after the full 150 years o f the study, meaning that over this time period, the age class distribution was not fully restored to the theoretical negative exponential distribution. 3.3.3 Scenario 3, habitat type level historical disturbance Targets Disturbance rates ranged from 552 ha to about 81,000 ha per decade over the 4 vegetated habitat types, and cumulative disturbance ranged from 6,400 ha to about 919,000 ha by period 15 (Table 3-5). The highest disturbance rates in terms of area disturbed occurred in the upland treed habitat type, which was the most dominant habitat type in the study area. Upland and wetland habitat types had notably different changes in serai stage distribution , from initial conditions, to conditions at the end of the analysis period. Wetland habitat types had very little increase or . 79 decrease in serai stage indicators, all being <1%. The upland treed habitat had decreases in the amount o f young, mature, and old serai stages (12%, 8%, and 9% respectively), while the amount o f early serai stage increased by about 5%. A s expected, the upland shrub habitat decreased in the amount o f early serai stage because all shrub stands which started at age 0 would have aged over time i f not disturbed. However, only an additional 2.7% of this habitat type reached the old serai stage after 150 years. 3.3.4 Scenario 4, landscape unit level historical disturbance Targets Over the entire analysis period, disturbance rates ranged from 4 ha to about 52,000 ha per decade over all stand types, while the range in cumulative disturbance was 4 ha to about 583,000 ha after 150 years (Table 3-5). Lodgepole pine stand types had the greatest area disturbed in each time period (52,000 ha) and the greatest cumulative disturbance after 150 years (583,000 ha), followed by deciduous stands, interior spruce stands and shrub types. In the lodgepole pine stand type young serai stage increased by almost 18%, while early, mature, and old serai stage decreased (15%, 2%> and 10%). Interior spruce and deciduous stands had an increase in early serai stages o f 16% and 46%, and a decrease in young, mature, and old. A s expected, the early serai stage shrub stands decreased, and increased in the older serai stages. After 150 years, 6.6% of the shrub stands had been undisturbed. 3.3.5 Comparison of scenarios 2, 3, & 4 Overall, when compared across scenarios, we found that setting the disturbance targets in different zones did affect indicator outcomes in period 15, thus confirming our hypothesis (Table 3-6). A t the forest level, the 3 scenarios produced significantly different indicator outcomes for all except early serai stage. Looking at the forest level results over time, this same pattern is apparent (Figure 3-6). Early serai stage remains relatively constant between the scenarios, while young forest decreases, and mature and old stands increase (from scenarios 2 to 3 to 4). We looked at the age class distribution of the study area at periods 0, 5, 10, and 15 (Figure 3-7). Over time, these graphs show 4 different "spikes" in the age class distribution, which are artefacts o f the initial landscape condition being tracked as the landscape ages over the analysis period. Thus, in period 0, this signal is the spike in age class 1-21, in period 5, the spike is in age class 41-60, in period 10, it is in age class 101-120, and finally, in period 15, it is in age class 141-160. These "spikes" show notable differences between the three scenarios. For example, in 80 scenario 3, the habitat type scenario, the spike in period 5 at age class 41-60 is above 30% of the landscape, but is under in the other 2 scenarios. This appears to be true for the spikes in period 10 (age class 101-120) and 15 (age class 141-160) in this scenario, which are greater than the other 2 scenarios. Also of note, in scenario 3, the landscape unit scenario, there are more stands older than 300 years than in the other scenarios. 3.4 D I S C U S S I O N 3.4.1 Assessment of the model One of the overall objectives o f this paper was to evaluate the performance of a disturbance model which uses stochastic heuristic optimization techniques. Our approach is relatively simple compared to many simulation models which incorporate ignition and spread parameters, and which are often driven not only by autogenic characteristics such as rate o f spread, intensity etc, but also allogenic characteristics such as fuel build up, and weather effects. We did not directly model a disturbance process, but rather modelled the patterns that would result from specific historical disturbance frequencies. This is the premise of management guided by a natural disturbance paradigm, where the goal is to recreate the patterns that would be caused by disturbance using forest harvesting and silvicultural regimes. The model is therefore very useful for planners wishing to explore management scenarios or test, hypotheses, but would be less useful for someone directly studying the disturbance process (for example how different fire intensity would affect ecosystem productivity). The sensitivity analysis indicated that the model was relatively robust to changes in disturbance frequency, and that disturbances were random, within the target constraints. B y robust, we mean that changes in the disturbance frequency at 1% intervals resulted, over the range o f rates (0-8%), in similar shaped curves over time (Figure 3-4), although the values o f the curves were different for the different disturbance frequencies. We would have considered the model not robust i f a change in disturbance rate by 1 % resulted in very different indicator curves over time. The variation (standard error of the mean) in indicator outcomes between the runs (three runs for each disturbance rate in the analysis) increased with increasing disturbance frequency, which is indicative o f randomness in the disturbances. At low disturbance levels, changes in the serai stage indicators are primarily caused by stand aging, which is deterministic. However, at higher disturbance frequencies, serai stage indicators are affected more by disturbance, which is a stochastic process that occurs independently o f stand age in the model, 81 meaning there is more variation between model runs. We choose to only complete 3 runs o f each disturbance rate in the sensitivity analysis, as compared to 5 runs in the remaining scenarios. Since we were looking for general patterns in the data in the sensitivity analysis to evaluate performance, rather than making direct inferences about the outcomes, a smaller number o f runs seemed appropriate. Another relatively simple test of the model's overall performance was to compare the modelled disturbance rates with those o f the historical time period. From a modelling perspective, did the model achieve the disturbance goals in all the scenarios? While we were running the model we had speculated that there was a case where the model would not completely converge on a solution. This could occur when goals are set at increasingly smaller scales and for smaller physical areas on the landscape. For example, in a landscape unit with very little area, and thus few polygons, achieving the disturbance probability goal might in fact be impossible i f no polygons could be found to add or subtract that would exactly equal the required area disturbed. This would render the model incapable o f reproducing similar temporal patterns in disturbance on the landscape. Fortunately this did not occur, and in fact Figure 3-6 indicates that the model did reproduce the temporal patterns o f our historical data, as indicated by the area disturbed each time period. We were partially able to validate the model by comparing the age class distribution produced by our model runs at different time periods to the findings o f other research in the boreal forest. Although admittedly subjective, we would have considered there to be errors in the model i f our age class distribution(s) differed in extreme from similar research. The initial age class distribution in the study area (all species combined) (Figure 3-5) was comparable to the age class distribution of Weir et al. (2000) in the Saskatchewan boreal forest where both management and climate change have had a noticeable impact on the age class distribution. Our model runs also produced similar patterns in age class distribution at discrete time periods throughout the planning horizon as Fal l et al. (2004) in the boreal forests of Quebec. This research used a simulation model to simulate wildfire with a return interval of 100 years on the landscape. The study area in Quebec had a very similar forest management history as our study area, where the majority o f activity occurred from the 1950's onward (Fall et al. 2004). Despite differences in tree species in the two areas, which could affect the initial starting conditions due to preferential species' harvesting, both models produced "spikes" in the age class distribution that affected the outcomes o f the model runs over the entire analysis period; these again are the landscape legacies described above. Therefore, our model which simulates stochastic • 82 disturbances through optimization, achieved comparable outcomes to a more complex disturbance model, which simulates initiation, spread and termination o f forest fires. Inclusion of disturbance in any forest management model can seem like a daunting concept, especially i f a large amount of data is required to parameterize a model o f disturbance initiation, spread, and termination. Our approach demonstrates that integration of disturbance events into a forest management is possible with mirufnal data inputs. We used historical fire data to create disturbance probabilities for our study area to explore the outcomes which would arise i f we were to manage our landscape in the same manner as historical natural disturbances; this is consistent with the concept of historical or natural range o f variability which guides B C forest management policy. The source of this data however, can also come from more complex modelling efforts, using deterministic fire models such as Farsite 7 or Prometheus8 where various runs can be made for particular stand types over a range of typical climate conditions for an area to generate a probability matrix. We like the flexibility of using a probability matrix, because it is also able to incorporate expert knowledge into the model framework. Knowledge from an expert fire scientist, non-governmental organization or First Nations elder on disturbance frequency can thus be easily modelled, which is especially useful for examining hypotheses and scenario planning. 3.4.2 Implications of modelling approach Our approach follows the large body of work on modelling forest succession using simple and complex Markov chains (Baltzer 2000; Korotkov et al. 2001; Yemshanov & Perera 2002; Benabdellah et al. 2003). In the case of forest succession, the processes affecting succession-light, nutrients, moisture, and tree growth/death- are not modelled directly. Instead, each particular stand has one or more possible transition probabilities which, when modelled using Markov chains, result in the reproduction o f similar temporal patterns in succession observed (or modelled by a process model) in actual forests. As with the forest succession literature, the logical extension is a spatially explicit model using Markov chains. Our methods for this application wil l be described in a subsequent paper. When compared with efforts by other researchers along similar lines o f inquiry in disturbance modelling, the approach we have taken identifies an alternate path for using 7 http://farsite.org/ 8 http://www.firegrowthrnodel.com/index.cfin 83 disturbance modelling and simulation in support of forest management planning and analysis. Many of the concepts underlying our approach were described by Richards et al. (1999), who formulated and solved it as a stochastic dynamic prograrnming problem. These authors noted the need for further work to apply these concepts efficiently in a decision-making framework, and in our approach we have attempted to do that. Zhou and Buongiorno (2006) note the benefits of combining Markov chain approaches with optimization, and develop a solution to a loblolly pine (Pinus taeda L. ) management problem. But they employ a mathematical approach and multiple linked models to perform the analysis. In our approach, no special explicit mathematical formulation was needed to define to the problem. Rather, the disturbance simulation problem was formulated as a special case of a stochastic-heuristic harvest scheduling problem, and a harvest scheduling model was used to perform the analysis. Peter and Nelson (2005) report on an approach which links a harvest scheduling model and a disturbance model, and incorporate disturbance as an independent process between harvesting periods. Using our approach, the Markov chain - type simulation is embedded within an optimization framework. This allows, within a single model, establishing the age class distribution which would result from a given set o f disturbance probabilities, then setting that age class distribution as a management goal. Alternative management strategies to achieve those conditions using harvesting, making tradeoffs with additional goals such as harvest volume and revenue, can then be accomplished within the same model. The transferability of our approach to other users, models, and applications in part depends upon the purpose these users intend the model to serve in a particular application. Our research question was driven by a particular requirement o f B C forest management policy, namely the use o f age-class distributions resulting from historical disturbance frequencies as a basis for setting harvest levels. Our modelling approach answers that need efficiently, while requiring minimal data and model calibration. However, different information needs may dictate different modelling approaches. One aspect of our approach, which may be limiting in some applications, is how we have handled the spatial aspects o f disturbance patterns. Our approach is spatial, to the level o f the zones for which the disturbance probabilities may be set. While the proportions of disturbance, and the resulting age class distributions, are accurate to the level of these zones, within the zones they are randomly distributed. I f the purpose o f the modelling requires achievement o f specific disturbance patch sizes and configurations, our approach will not currently meet that need. Other approaches (e.g. Salvador et al, 2005; Pinol et al, 2005, 84 Mladenoff, 2004) which include some level of disturbance behaviour modelling may be preferable. 3.4.3 Scenar ios Initial starting conditions on the landscape played a more important role in the long-term distribution o f serai stage indicators than we had expected. A s discussed above, Markov chains have the property o f achieving steady state conditions at some point in time. Achievement of a steady state in the model, which is functionally a Markov chain model, is sensitive to the starting conditions of the landscape. The term "landscape legacy" has been used to describe this effect where the condition of the landscape at one point in time (whether natural or managed under a particular harvest regime) affects the landscape condition for many years into the future regardless o f the historical disturbance regime (Wallin et al. 1994; Osterlund et al., 1997; Fall et al., 2004, Nonaka and Spies, 2005). Wall in et al. (1994), for example, found that harvest practices in Oregon, which highly fragmented the landscape, created a legacy of small patches for up to 100 years after the initial harvest. In Scandinavia, Osterlund et al. (1997) found that fire suppression and clearcutting greatly changed the structure of forests from multi-storied stands to even-aged, single story stands. Nonaka and Spies (2005) demonstrated that even if wildfires were allowed to burn in the Oregon Coast Range the landscape does not return to the historical range o f variability after 100 years, primarily because of the loss o f old-growth. In the boreal forests of Quebec Fall et al. (2004) found that the cumulative impact o f harvesting, fire suppression, and natural disturbance has shaped the age class distribution on the landscape, which has implications for management many years into the future. Specifically, balancing ecological (old-growth) and economic objective becomes difficult in the transition period before an equilibrium age class distribution is reached because the current landscape has a deficit o f older forests. In this study the time before equilibrium is the transition time in our model, while the equilibrium is the desired condition. The desired condition, in this case, is the historical condition, while the transition time is 130 years (Figure 3-4) i f we assumed that tomorrow historical patterns o f stand-replacing fires once again became the dominant (and only) disturbance regime in the study area. Although the time when equilibrium is achieved may be important from a forest management perspective, the transition dynamics have important 85 implications to consider in terms o f management for the mountain pine beetle (discussed in further detail below). The transition time to equilibrium in each serai stage was largely achieved only after all initial stands had either been disturbed or had aged into another serai stage. Thus, equilibrium was reached in early stands first at 40 years, and old stands last at 130 years. In early and young serai stages the transition time occurred when all initial stands in early and young serai stage made the transition into the next serai stage. For example, all stands which initially were early serai stage at the start of the model run became young serai stage at period 4 (at 40 years). A n y new disturbance would be classified as early serai stage and thus equilibrium was achieved. B y the time the legacy stands reached the mature serai stage, most o f them had already been disturbed, and the transition time to mature was the same time as the transition to young. The transition to old occurred 20 years later at 130 years. Our comparison of the three different scenarios (2, 3, 4) where targets were set at the Forest, Habitat Type and Landscape Unit level revealed that scale does significantly affect indicator outcomes, especially serai stage. O f course, some caution should be exercised when making such a statement since at period 15, after 150 years, there were generally significant differences in indicator outcomes. This was not true in some cases, most noticeably in lodgepole pine stand types, where all indicators except old serai stage were not different between the scenarios. This was not intuitive, for reasons explained below, so we looked at the distribution o f the indicators over time in this stand type (data not shown). Although the indicator values were similar at period 15, they were not necessarily similar at other time periods. In our approach, when disturbance targets are set at the forest level, each disturbance occurs randomly on the landscape, constrained only by the target itself. Disturbance occurs independently o f the habitat type, landscape unit, and age class of the stands. When targets are set at a finer scale, using zones such as habitat types, disturbances at the scale above (forest) are constrained by the targets set at the habitat type level, while disturbances in the scale below (landscape units) still wil l occur randomly and independently o f landscape units. This demonstrates the hierarchical nature of the model, which affects the disturbance simulation process. The best example o f this is looking at Figure 3-7, where cumulative disturbance decreased between scenario 2, where targets were set at the forest zone, and scenario 4, where targets were set for each landscape unit. In scenario 4 we set targets for only a portion o f the landscape (units with > 1000 ha). Therefore, even i f our targets were met in all units, our forest level indicators would still be less than targets set (and calculated) for the entire forest, since the 86 entire forest includes the units less than 1000 ha. If we had set targets in all units however, regardless of size, our cumulative disturbance indicator at least would have been similar to those obtained using forest-level targets. Another example can be seen from Table 3-5, in spruce and deciduous stands. When the target is set at the forest level rather than the landscape unit level, the amount of landscape in early serai stage at period 15 is greater in spruce and less in deciduous than it would be when targets are set at the landscape unit level. This is due to the randomness in the model, because it does not proportionally allocate disturbance events within small ecosystem types, unless specific targets are set for those ecosystems. If understanding the effect of disturbances in these ecosystems is important, the targets must be set for these ecosystems specifically. However, this leads us to believe that the ability to model disturbance at multiple scales is an excellent tool for looking at long-term impacts on rarer ecosystems types. Because these ecosystems are small, it is often believed that a coarse-filter approach to management indeed will not necessarily preserve or maintain the integral processes of these ecosystems, because we do not account for them directly. If targets are set at the forest level, this might be the case in our model, yet because we can track any number of zones, we can assess the impact of disturbance on these processes which might act independently of the landscape as a whole, such as a large fire. Then, we can rerun the model and set specific goals for these ecosystem types if so desired. The model is thus flexible to explore landscape level, coarse-filter issues, while retaining its ability to model the rare as well. This is because the original design of the model was as a hierarchical model (Nelson 2003) allowing the flexibility to model at broad scales where the lack of detail did not impact the questions being asked of the model, and at finer scales where more detail is required. 3.4.4 Implications for management The implications of this study for sustainable forest management in the study area are: 1. Currently, landscape conditions are beyond the historical range of variability, meaning that despite management policies we are not harvesting in a manner similar to natural disturbances, 2. The treed landscape is currently older than what it would have been historically, and if historical disturbance levels were immediately reintroduced (fire only) a much younger landscape would be dominant, 87 3. If we look at specific stand types, some stands, such as lodgepole pine and deciduous wi l l be younger overall than currently, while others, such as interior spruce will be older than currently, and 4. It will take longer than 150 years to return the age class distribution to a historical age class distribution i f that were possible or desired. In the face of the current mountain pine beetle epidemic in this region, one o f the criticisms about forest management has been that our harvesting practices, when combined with our fire suppression policies, have created an increasingly older landscape that is more susceptible to both insects and fire. This belief is supported by this modelling exercise, considering the distribution o f serai stages at the start o f the model analysis, and then again at the end o f the analysis. What is striking is that the modelling results show that the amount o f mature forest dropped by over a quarter (26%) while the amount o f early and young serai stages certainly increased overall. I f early and young serai stages can be considered less susceptible to mountain pine beetle than the old and mature, then the forest does not become "beetle-proof until equilibrium is reached in young serai stands at 100 years. Thus, even i f management were to emulate the rate of disturbances in the past, it wi l l take many decades before we potentially return the forests to a state where an epidemic o f the current size cannot occur. Another important factor o f this modelling exercise is that even i f we return historical disturbances to the landscape, it wi l l take more than 150 years to have an age class distribution that wi l l truly resemble the negative exponential distribution, which is expected under a stand-replacing disturbance regime which occurs independently of stand age (Johnson and V a n Wagner, 1985). Although using the negative exponential model as a target is interesting from a research perspective, other researchers have pointed out that it is not the target for managed forests where specific age classes are more desirable for harvesting (Fall et al., 2004). I f our goal in B C is to harvest in a manner so as to emulate natural disturbances we clearly need to define what this means, and whether we apply this management strategy to the whole landscape or only part. A s we are finding in B C , large natural disturbances can occur regardless o f our management activities. Right now B C management policy potentially ignores the interaction o f management and disturbance, and assumes that management alone wi l l maintain the health o f function o f the forests. The results of this modelling demonstrate that this may not be the case. Finally, all of these implications presume that the climate is not changing. If the climate were different than in the past, and i f this change in climate affected the disturbance rate, then 88 the landscape condition will be different than that projected by the model. The effects of a changing climate can be explored with the model by modifying the disturbance probabilities up or down. Furthermore, since disturbance probabilities can be set for each time period disturbance probabilities could vary depending on the future climate scenario. For example, one climate scenario might indicate a gradual warming in the spring, which could lead to increasing fire disturbance. In period 1 the disturbance rate might be quite small, but could increase every period over the entire analysis period. 3.4.5 Future direction A clear line of further inquiry is suggested by the architecture of our model, which could significantly expand its capabilities. We have designed our model so that the individual polygons which represent the vegetation are stored as faces in an undirected planar graph, with the edges representing adjacency between polygons. In this arrangement, it becomes possible to use optimization techniques which exploit the graph architecture to incorporate patch size and other spatial metrics into the model. Thus, we believe that it would be possible to incorporate targets for the spatial configuration (minimally patch size) of the disturbances into the model through this method, without requiring the incorporation of a disturbance behaviour and spread model. Our first priority is to study the same problem, but with the addition of spatial goals. Information on the disturbance size distribution in the study area is available, as well as the spatial patterns of disturbances. The model architecture is capable of modelling a range of specific disturbance sizes, and this should allow the comparison of disturbance spatial patterns produced by the model with the historical disturbance patterns. We are also working on modelling multiple disturbances. Currently the model considers disturbance processes either independent or dependent of age class, but not both at the same time. We would like to be able to model both fire (independent of age) and mountain pine beetle (dependent) and look at the cumulative effects and interactions of these disturbance processes. 89 3.5 T A B L E S Table 3-1 Description and function of the five required data layers in disturbance model Data layer Description Funct ion in model Cover type Vegetation description. For example, leading species. Describes landscape condition Age c lass Age classes on the landscape. Describes landscape condition Landscape unit Administrative boundaries, planning units, etc. Describes zoning Habitat types Ecological boundaries, such as ecosystems or wildlife habitat. Describes zoning Special management areas Discrete planning units. For example, could be old-growth management areas, caribou corridors, etc. Describes zoning Table 3-2 Description of the data layer values used in this study Data layer Study values Cover type Lodgepole pine, interior spruce, black spruce, Engelmann spruce, deciduous (aspen/cottonwood), subalpine fir, hemlock, Douglas-fir, paper birch, shrub Age c lass 40 age classes, each 10 years in length includes anything older than 400 years) (except for 40 which Landscape unit Same as cover type Habitat types Upland treed, upland shrub, wetland treed, wetland shrub, non-vegetated Special management Not applicable. All have value of 1. areas Table 3-3 Indicators and their attributes used in this study Indicator Name Indicator Definition Serai Staae Indicators Early % area in stands 1-50 years of age Young % area in stands 51-100 years of age Mature3 % area in stands 111+ years of age Old 3 % area in stands 141+ years of age Disturbance Indicators Disturbance % area in disturbed stands in age class 1 only Cumulative % area in disturbed stands in all age disturbance classes "Note that the ages in Mature and Old serai stages overlap. Table 3-4 Disturbance targets calculated from historical wildfire disturbance rates Scenario Total area burned,1904-1954 (ha) Total area (ha) Tarqet area burned/decade (%) Disturbance return interval (years) Scenario 2 Forest 446,521.30 1,213,136.48 7.36 136 Scenario 3 Habitat tvoe Upland tree 392,427.08 968,936.61 8.10 123 Upland shrub 46,076.21 213,636.20 4.31 231 Wetland tree 2,755.51 7,096.81 7.77 129 Wetland shrub 5,262.50 23,466.86 4.49 223 Scenario 4 Landscape unit Lodgepole pine 255,011.60 657,186.66 7.76 129 Interior spruce 57,098.23 200,865.53 5.69 176 Black spruce 5,151.17 12,294.74 8.38 .119 Deciduous 99,687.31 168,142.85 11.86 84 Subalpine fir 2,264.45 20,439.77 2.22 451 Shrub 51,338.71 213,636.10 4.33 231 Table 3-5 Comparison of indicator values between Scenarios 2, 3, & 4; non-significant values are shown with ns, all others are significant. Indicator Mean Hectares (SE) Period 15 Analysis Scenario Level Cumulative Disturbance Early Serai Young Serai Mature Serai Old Serai Forest Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 1,181,728 (4,286) 1,070,573 (7,770) 1,014,533 (11,806) 317,440 (2,523) ns 323,325 (1,438) ns 321,618 (1,968) ns 420,292 (4,051) 392,278 (1,828) 374,356 (9,671) 355,715 (3,289) 377,852 (3,130) 397,476 (9,838) 180,009 (2,468) 210,906 (2,155) 258,940 (8,145) Upland Treed Habitat Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 928,423 (3,354) 918,537 (7,227) 845,721 (6,191) 288,244 (3,366) 306,433 (1,635) 296,240 (4,200) 378,534 (4,920) 372,130 (2,197) 354,890(1,103) 319,520 (3,082) 307,744 (3,917) 335,172 (554) 160,725 (1,992) 150,500 (1,863) 209,117 (3,162) Wetland Treed Habitat Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 6,614 (55) 6,455 (50) 5,249 (53) 2,018 (59) ns 2,049 (8) ns 1,896(13) ns • 2,611(80) 2,670 (51) 2,043 (93) 2,453 (58) 2,362 (39) 3,143 (96) 1,268 (86) 1,250 (58) 2,295 (81) Upland Shrub Habitat Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 196,374 (1,948) 125,971 (1,665) 149,119 (946) 17,271 (911) 12,155 (533) 19,188 (1,796) 24,122 (631) 13,787 (385) 20,539 (449) 22,328 (716) 37,779 (546) 23,995 (789) 11,768(640) 30,709 (346) 15,493 (2,245) Wetland Shrub Habitat Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 21,213 (229) 13,382 (189) 11,851 (59) 1,445 (74) 974 (55) 1,668 (34) 2,084 (193) 1,242 (113) 1,616 (23) 1,710 (115) 3,023 (165) 1,955 (70) 894 (170) 2,399 (87) 1,440 (52) Lodqepole Pine Stands Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 617,176 (3,622) ns 609,494 (4,453) ns 583,218 (9,569) ns 195,362 (6,066) ns 197,748 (3,371) ns 196,750 (2,113) ns 251,104 (7,250) ns 248,819(3,597) ns 238,726 (9,921) ns 211,754 (4,605) ns 211,648 (3,581) ns 222,740 (9,548) ns 107,565 (2,744) 103,252 (4,535) 128,795 (8,291) Interior S D r u c e Stands Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 189,489 (964) 175,458 (1,734) 138,358 (1,793) 57,482(2,170) 61,478 (1,420) 44,272 (210) 77,665 (1,781) 69,337 (1,545) 55,762 (654) 65,380 (1,283) 69,712 (1,796) 100,493 (1,293) 32,448 (935) 40,780 (2,242) 76,424 (588) Deciduous Stands Sc. 2 Forest Sc. 3 Habitat Type Sc. 4 Landscape unit 157,466 (926) 148,245 (1,448) 158,423 (1,745) 46,493 (1,246) 51,440 (696) 70,020 (404) 64,700 (1,105) 58,946(1,711) 66,135 (602) 56,826 (1,097) 57,634 (2,110) 31,864 (857) 28,744 (748) 33,891 (1,836) 16,603 (357) VO Table 3-6 Mean (SE) indicator values at time period 15 (150 years) for Scenarios 2-4, and the % change in the serai stage indicators from initial conditions (indicated in bold). Period 15 Scenario Disturbance 3 rate Cumulative Early serai Young Mature Old (ha/decade) disturbance (ha) (ha) serai serai serai (ha) (ha) (ha) Scenario 2 93,517 (0) 1,181,728 (4,286) 317,440 420,292 355,715 180,009 Forest (2,523) - (2,468) (4,051) (3,289) 2.1% +9.1% -26.0 +2.2 Scenario 3 Habitat tvoes Upland treed 80,970 918,537 306,433 372,130 307,745 150,500 (0.3) (7,227) (1,635) (1,863) (2,197) (3,917) +4.7% -12.3% -7.5% -8.6% Wetland 552 6,455 2,049 2,670 2,362 1,250 treed (0.2) (50) (8) (58) (51) (39) +0.2% +0.1% -0.3% -0.3% Upland 9,448 125,971 12,155 (533) 13,787 37,779 30,709 shrub (0.7) (1,665) -4.2% (3,460) (385) (546) +1.0% +3.2% +2.7% Wetland 1,027 13,382 974 1,242 3,023 2,399 shrub ' (0.9) (188) (55) (87) (113) (165) +0.1% +<0.1% -0.1% -0.1% Scenario 4 Landscape units Lodgepole 52,235 583,218 196,750 238,726 222,740 128,795 pine (71) (9,569) (2,113) (8,291) (9,921) (9,548) -14.6% +17.8% -3.2% -9.7% Interior 11,630 138,358 44,272 (210) 55,763 100,493 76,424 spruce (0.3) (1,793) +16.4% (588) (654) (1,293) -49.4% -17.2% +21.0% Black spruce 1,027 10,265 3,817 4,164 4,275 2,748 (1) (108) (38) (97) (78) (115) +30.9% +12.0% -42.9% -30.7% Deciduous 19,826 158,423 70,020 (404) 66,136 31,864 16,603 (0.2) (1,745) +36.5% (357) (602) (857) -7.6% -28.9% -14.3% Subalpine fir 449 6,211 1,772 2,518 16,148 14,977 (0.3) (112) (14) (59) (58) (116) +1.2% +2.2% -3.4% -1.6% Shrub 8,765 117,136 (1,568) 4,697 6,652 18,092 15,648 (0.3) (524) (851) (1,075) (1,285) -10.4% +2.8% +7.6% +6.6% " Variation due to slight deviations from target. 95 3.6 FIGURES Figure 3-1 Location of study area 9 6 Figure 3-2 An example of zoning at the a) habitat type and b) landscape unit level in the study area. Note that Non-vegetated types include water bodies such as lakes. Figure 3-3 Disturbance (a) and mean cumulative disturbance (with standard error bars) (b) indicators at 9 different disturbance rates, for Scenario 1 . 97 500000 Disturbance a) Early serai stage rate 1 2 3 4 5 6 7 8 9 101112131415 Time period (10 year intervals) 500000 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Time period (10 year intervals) Figure 3-4 Mean area (ha) in early (a), young (b), mature (c), and old (d) serai stage indicators (with standard error bars) at 9 different disturbance rates, for 3 runs of Scenario 1. 100.00 CO CD >> X J CD > O 10.00 1.00 0.10 50 100 150 200 250 300 Age (years) Figure 3-5 Mean age class distribution, presented as cumulative % of the study area, for Scenario 2, at 4 of the planning periods in the analysis. 98 3.7 R E F E R E N C E S Armstrong, G . W . (2004). Sustainability of timber supply considering the risk of wildfire. Forest Science. 50(5): 626-39. Balzter, H . (2000). Markov chain models for vegetation dynamics. Ecological Modelling. 126: 139-154. B C Ministry o f Forests. (1995). Forest Practices Code of British Columbia, Biodiversity Guidebook. Victoria, B C : B C Ministry of Forests. Benabdellah, B . , Albrecht, K . - F . , Pomaz, V . L . , Denisenko, E . A . , Logofet, D M . (2003). Markov chain models for forest successions in the Erzgebirge, Germany. Ecological Modelling. 159: 145-160. Bergeron, Y . , Leduc, A . , Harvey, B . D . , and Gauthier, S. (2002). Natural fire regime: a guide for sustainable management o f the Canadian boreal forest. Silva Fennica. 36(1): 81-95. Bettinger, P. , and Chung, W . (2004). The key literature of, and trends in, forest-level management planning in North America, 1950-2001. International Forestry Review. 6(1): 40-50. Bonar, R. L . , Lougheed, H . , and Andison, D. W . (2003). Natural disturbance and old-forest management in the Alberta Foothills. Forestry Chronicle. 79(3): 455-61. Boychuck, D . , Perera, A . H . , Ter-Mikaelian, M . T., Martell , D . L . and L i , C . (1997). Modelling the effect o f spatial scale and correlated fire disturbances on forest age distribution. Ecological Modelling. 95: 145-64. Boychuck, D . , and Perera, A . H . (1997). Modeling temporal variability o f boreal landscape age-classes under different fire disturbance regimes and spatial scales. Canadian Journal of Forest Research. 27: 1083-94. Delong, S. C . (1998). Natural disturbance rate and patch size distribution o f forests in northern British Columbia: Implications for forest management. Northwest Science. 72: 35-48. 100 Dewhurst, S. M . , and Karjala, M . K . (2003). Timber Supply Analysis Aleza Lake Research Forest. University of Northern B C , Prince George, B C . Diaz-Balteiro, L . , and Carlos Romero (2004) Sustainability of forest management plans: a discrete goal programming approach. Journal of Environmental Management. 71(4): 351-359. Fall, A . and Fall, J. (2001). A domain-specific language for models o f landscape dynamics. Ecological Modelling. 141: 1-18. Fall, S. A . , Fortin, M. - J . , Kneeshaw, D. D. , Yamasaki, S. H . , Messier, C , Bouthillier, L . , and Smyth, C. (2004). Consequences of various landscape-scale ecosystem management strategies and fife cycles on age-class structure and harvest in boreal forests. Canadian Journal o f Forest Research. 34: 310-322. He, H . S., Mladenoff, D. J., and Gustafson, E . J. (2002). Study of landscape change under forest harvesting and climate warming-induced fire disturbance. Forest Ecology and Management. 155: 257-70. Johnson, E . A . , and Van Wagner, C. E . (1985). The theory and use o f two fire history models. Canadian Journal of Forest Research. 15: 214-20. Karjala, M . K . , and Dewhurst, S. M . (2003). Including Aboriginal issues in forest planning: a case study in central interior British Columbia, Canada. Landscape and Urban Plarining. 64: 1-17. Karjala, M . K . , Sherry, E . E . , and Dewhurst, S. M . (2004). Criteria and indicators for sustainable forest planning: a framework for recording Aboriginal resource and social values. Forest Policy and Economics. 6: 95-110. Klenner, W . , Kurz, W., and Beukema, S. (2000). Habitat patterns in forested landscapes: management practices and the uncertainty associated with natural disturbances. Computers and Electronics in Agriculture. 27: 243-262. Korotkov, V . N . , Logofet, D . O . , Loreau, M . (2001). Succession in mixed boreal forest of Russia: Markov models and non-Markov effects. Ecological Modelling. 142: 25-38 101 Landres, P. B . , Morgan, P., and Swanson, F. J. (1999). Overview o f the use of natural variability concepts in managing ecological systems. Ecological Applications. 9(4): 1179-88. Lertzman, K . , and Fall, J. (1998). From forest stands to landscapes: Spatial scales and the roles o f disturbances. Ecological Scale: Theory and Applications. D . Peterson, and V . T. Parker, eds. Columbia University Press. Lockwood, C , and Moore, T. (1993). Harvest scheduling with spatial constraints: a simulated annealing approach. Canadian Journal of Forest Research. 23: 468-78. Mitchell , R. J., Palik, B . J. , and Hunter, M . L . Jr. (2002). Natural disturbance as a guide to silviculture. Forest Ecology and Management. 155: 315-17. Mladnoff, D . 2004. L A N D I S and forest landscape models. Ecological Modelling 180(1): 7-19. Nelson, J. (2003). Forest-level models and challenges for their successful application. Canadian Journal o f Forest Research. 33: 422-29. Peter, B . , and Nelson, J. (2005). Estimating harvest schedules and profitability under the risk o f fire disturbance. Canadian Journal of Forest Research. 35: 1378-88. Pinol, J. , Beven, K . , and Viegas, D . X . (2005). Modelling the effect of fire-exclusion and prescribed fire on wildfire size in Mediterranean ecosystems. Ecological Modelling. 183(4): 397-409. Richards, S.A., H.P. Possingham and J. Tizard. (1999). Optimal fire management for maintaining community diversity. Ecological Applications 9(3): 880. Salvador, R, Lloret, F. , Pons, X , and Pinol, J. (2005). Does fire occurrence modify the probability of being burned again? A null hypothesis test from Mediterranean ecosystems in N E Spain. Ecological Modelling. 188(2/4): 461-469. Turner, M . G. , Romme, W . H . , Gardner, R. H . , O'Neill , R. V . , and Kratz, T. K . (1993). A revised concept of landscape equilibrium: disturbance and stability on scaled landscapes. Landscape Ecology. 8(3): 213-227. 102 Urban, D . L . , and Wallin, D . O. (2002). Introduction to Markov models. Learning Landscape Ecology. S. E . Gergel, and M . G. Turner, eds. New York: Spring-Verlag. 35-48. Wei , X . , Kimmins, J. P., and Zhou, G . (2003). Disturbances and the sustainability of long-term site productivity in lodgepole pine forests in the central interior of British Columbia-an ecosystem modeling approach. Ecological Modelling. 164: 239-56. Weir, J. M . H . , Miyanishi, K . and Johnson, E . A . (2000). Fire frequency and the spatial age mosaic of the mixed-wood boreal forest in western Canada. Ecological Applications. 10(4): 1162-77. Wotton, B . M . , and Martell , D . L . (2005). A lightning fire occurrence model for Ontario. Canadian Journal of Forest Research. 35: 1389-401. W u , J, and Loucks, O. L . (1995). From balance of nature to hierarchical patch dynamics: a paradigm shift in ecology. The Quarterly Review of Biology. 70(4): 439-66. Yang , J., He, H . S., and Gustafson, E . J. (2004). A hierarchical fife frequency model to simulate temporal patterns of fire regimes in L A N D I S . Ecological Modelling. 180: 119-33. Yemshanov, D. and Perera, A . H . (2002). A spatially explicit stochastic model to simulate boreal forest cover transitions: general structure and properties. Ecological Modelling. 150: 189-209. Zhou, M . and Buongiorno. J. (2006). Forest landscape management in a stochastic environment, with an application to mixed loblolly pine-hardwood forests. Forest Ecology and Management. 223 (1-3): 170-182. 103 CHAPTER 4 INTERACTIONS OF BROAD-SCALE CLIMATE PATTERNS AND DISTURBANCE IN SUB-BOREAL FORESTS9 4.1 I N T R O D U C T I O N Long-term climate change, and shorter term climate cycling are different mechanisms affecting the disturbance regime and landscape of any region. It is thought that climate is a driver of larger, more severe disturbance events such as catastrophic fires (Pierce et al., 2004) or epidemic insect outbreaks like the spruce budworm (Peterson et al., 1998), but it is not known to what extent it also affect less severe disturbance events. Untangling the cumulative effects o f climate drivers at multiple time scales is complex, but is often ignored by research which focuses on only one scale. We may be therefore underestimating or overestimating potential future fire severity as the climate continues to warm. In addition, current debate on fire suppression highlights further confounding impacts, o f both climate and human management on the landscape (Johnson, et al., 2001; Girardin et al., 2004; Westerling et al. 2006). The purpose of this study was to explore how climate at long and short-term scales affected the historical and current area disturbed in the sub-boreal spruce ecosystem in British Columbia, in order to incorporate some of this climate variability into our long-term landscape planning processes. 4.1.1 A review of climate impacts at multiple temporal scales A t the landscape level, climate at all temporal scales is a major driver of forest disturbance dynamics (Wu & Loucks, 1995; Veblen, 2003). Climate changes which occur over very long time periods, such as tens o f thousands of years, can cause catastrophic disturbance regimes that result in the complete renewal or reorganization o f ecosystems. This, in fact, has occurred in the past over and over again, in response to shifts in climate patterns (Davis, 1986). Climate changes on mid-time scales, such as hundreds of years, may have an impact on disturbance patterns that cause more gradual vegetation changes. Many authors have pointed out that ecosystems do not respond as a whole to climate changes on this temporal scale, but instead respond as individual species. Interactions with other species and environmental conditions, along with adaptation (sexual reproduction) or acclimation (ecological amplitude) form new ecosystems over time (Davis & Shaw, 2001; Hansen et a l , 2001). Climate (long-term trends) and weather (daily 9 A version of this chapter will be submitted for publication. Campbell, K.A., Larson, B., and Dewhurst, S.M. Interactions of broad -scale climate patterns and disturbance in sub-boreal forests. 104 trends) patterns which fluctuate on finer time scales, such as decadally or annually, are often the drivers o f cyclic disturbance regimes (White & Pickett, 1985; Agee, 1993; Oliver & Larson, 1996). The cycles of disturbances affecting a landscape often govern the range of variability in landscape conditions (Landres et al., 1999). The landscape itself is resilient to a certain background level of disturbance, (Grimm & Wissel, 1997; Peterson et al., 1998; Beisner et al., 2003; Ehnqvist et a l , 2003) and the species on the landscape reflect o f this level of disturbance. 4.1.2 A review of short-term cycling mechanisms, Pacific Decadal Oscillation On a decadal scale, broad-scale climate fluctuations (such as E l Nino events) can affect disturbance regimes over several years or even decades (Skinner et al., 1999; Heyerdahl et a l , 2002; McKenzie et a l , 2003; Hessl et al., 2004). In the Pacific Northwest there is strong effect of the interdecadal cycles o f the Pacific Decadal Oscillation (PDO). These events are similar to E l Nino, but have a stronger effect in northern latitudes, vary decadally rather than interannually, and have the strongest signal in winter (Mantua & Hare, 2002; Wang & Schmimel, 2003). In its positive phase, conditions are warmer and windier in the Pacific Northwest, with the opposite seen in more northernly latitudes (e.g., Alaska) and in the South Pacific (Gedalof et a l , 2002). Between 45 and 50 degrees latitude the P D O in its positive phase also tends to be associated with increased storm activity (Bond and Harrison 2000). Abrupt sign changes, from positive to negative phases or vice versa, have occurred over the last century in 1925, 1947, and 1977 (Mantua & Hare, 2002). However, evidence from tree rings and corals indicate that the influence o f the P D O on climate patterns is also variable, exhibiting greater influence in the 20 t h Century than the 19 t h Century (Gedalof et al. 2002). The phases o f this cycle are determined by measuring changes in sea surface temperatures north o f 20 degrees latitude in the Pacific Ocean. The index itself has a range of about 4.0 to -4.0 and is based on the difference of the sea surface temperature from monthly mean global average sea surface temperature, constructed using principal component analysis (Mantua & Hare, 2002). 4.1.3 Examples of climate impacts at multiple scales 4.1.3.1 Long-term climate changes (10,000 years +) Periodic interglaciation is an example of a disturbance which is affected by broad scale climate changes, since they often last 10,000-20,000 years (Davis, 1986). One such glacial maximum 105 occurred about 18,000 years ago, causing large-scale extirpation of species as the glaciers advanced over most o f North America. When the glaciers began to retreat, about 10,000 years ago, tree species migration (influenced by reproductive capabilities such as seed dispersal) and adaptation led to a reorganization of species; climate patterns and differential adaptation were such that the range o f many species overlapped in the past that no longer co-occur (Davis & Shaw, 2001). At long time scales, therefore, disturbances such as glaciation, which are influenced by long-term patterns of climate, cause ecosystem reorganization. 4.1.3.2 Mid-term climate changes (100-10,000 years) On slightly shorter time scales, hundreds rather than thousands or tens of thousands of years, climate events such as the Little Ice Age occur, which can also have profound effects on disturbance and ecosystem composition and condition. The Little Ice Age began around 1250 and ended in the late 1800's (Davis, 1986). During this time, the climate was cooler and wetter, with frequent lightning activity and a short fife cycle. Weir et al. (2000) provided an example of the impact of the Little Ice Age on disturbance. After about 1890, when the Little Ice Age ended, shifts in vegetation composition and structure were noted, from species adapted to short disturbance cycles pre-1890, to those adapted to longer disturbance cycles dominating after 1890. This pattern was also observed in the inland temperate forests of British Columbia, where there was evidence o f an increased number o f fires 250-1000 years prior to today, despite cooler temperatures during this time period (Sanborn et al. 2006). 4.1.3.3 Short-term climate cycling (1-100 years) The impact of P D O on forest fires has been demonstrated for various regions in the Pacific Northwest. In Washington and Oregon ponderosa pine forests, positive PDO events affected fire size, possibly because o f shallower snow packs and warmer temperatures (Heyerdahl et al., 2002). Large fires in Washington State were associated with the positive phase o f the P D O (Hessl et al., 2004), as were fires in the Pacific Northwest (Gedalof et al. 2005). 4.1.4 Context and objectives This study used a number o f climate variables to examine potential links between climate and disturbance on multiple time scales. A t mid-term time scales, temperature and precipitation 106 trends were examined. Over the shorter-term, the relationship between the P D O and large and small disturbance years was determined. The objectives were to: 1. Create a statistical model to explain the historical variation of area disturbed (1910-1999) using climate variables which represent mid-term climate changes, and shorter term climate cycling. 2. Predict mountain pine beetle disturbance 2000-2003 using the statistical model. 3. Determine whether human management (fire suppression) or climate has been the primary cause o f area disturbed over the past 89 years. 4.2 M E T H O D S 4.2.1 Study area description The study area is located across the range o f the dry cool subzone of the sub-boreal spruce ecosystem (the SBS dk biogeoclimatic variant) in the central interior of British Columbia, and is about 1 million ha in size (Figure 4-1). The climate is relatively cool, with mean annual temperatures 1.1°-2.1° C . In the winter snow depths can reach up to 250 cm, while during the rest of the year an average of 500 mm falls as rain. There are only about 70 frost-free days a year. Lodgepole pine is the current dominant species on the landscape, but interior spruce, and aspen are also abundant. The disturbance regime is dominated by stand replacing disturbances such as fire and mountain pine beetle (Delong, 2002), with a mean return cycle o f 125 years expected (BC Ministry o f Forests, 1995). 4.2.2 Disturbance data Historical forest fire and mountain pine beetle data were provided by the Canadian Forest Service as part of their Natural Disturbance Database 1 0. The data are spatial, in vector format (polygons), defining the timing and extent o f each individual disturbance 1920-2000. Data were compiled from existing paper and digital maps into one GIS database for each disturbance type. Additional information on each disturbance, such as ignition cause for fires, and infestation severity for mountain pine beetle, are included in the database. Additional data were collected from the B C Government to extend the time o f analysis to cover the period 1904-2000. Disturbances which occurred between 1904 and 1954 were 1 0 httpV/wvvW.pfc.forestrv.ca/fires/disturbance/index e.html 107 identified from the 2003 Vegeta t ion Resource Inventory data, where stand age i n the inventory was assumed to approximate time since fire (Bergeron et a l . , 2001; De long , 1998; Johnson & V a n Wagner, 1985; Johnson & Larsen , 1991; Johnson & Gutse l l , 1994; W e i r et a l , 2000). It was necessary, however , to supplement the historical fire data at the turn o f the century wi th the disturbance data derived from age classes, since the N a t u r a l Disturbance Database currently underestimates the number and perhaps extent o f the fires i n the early part o f the century ( B . Hawkes , Canad ian Forest Service, pers. comm.) . This underestimation occurs because all fires were not necessarily recorded in the early part o f the 2 0 t h century as they were i n more recent times. On ly stands w i t h >30% mounta in pine beetle infestation were included i n the analysis. This is considered a severe infestation, and was used as a point to differentiate stands wi th lower infestation levels, i n wh ich a po r t i on o f the lodgepole pine may not be attacked. Addi t iona l mountain pine beetle attack data were obtained from government sources, cover ing the period 2001-2003. These data were used to check the capabil i ty o f the analysis to predict current mountain pine beetle activity in the area. The purpose o f this analysis was to look at overa l l disturbance trends, rather than at trends in different disturbance regimes. Therefore, a l l fire, and mountain pine beetle data were combined. M o u n t a i n pine beetle act ivi ty dur ing this time p e r i o d was, however, min imal . A l l disturbance data were spatial ly jo ined in G I S , and summaries were created by year. Graphica l summaries were created to show the different tempora l distributions o f disturbances. Considerat ion was not given to differences in disturbance type (fire vs. mounta in pine beetle) in this analysis, as a l l disturbance data were poo led into one category (annual area disturbed). 4.2.3 Climate data His tor ica l temperature and precipi tat ion data were p rov ided by Environment Canada as part o f the Adjusted H i s t o r i c a l Canadian Cl ima te Data (Environment Canada, 2005) . These data have been compiled, and when necessary, corrected so that a coherent and complete database o f historical climate data is available for research purposes. Temperature and precipitat ion data were downloaded for the climate stat ion located in Ft . St. James, B C ( 5 4 . 4 5 ° N , - 1 2 4 . 2 5 ° W , 686 m), wh ich has cl imate records for 1895-1999. This climate station is not w i t h i n the study area, but was the nearest station wi th a r eco rd long enough to cover the time o f the analysis. Loca t ion o f the climate s tat ion is shown in (Figure 4-1) . 108 Seasonal averages of maximum temperature, minimum temperature, daily rainfall, and daily snowfall were created. The data were grouped by season to reduce the number of variables in the analysis. Other variables were initially explored, such as mean temperature and daily precipitation (rain + snow), but were redundant given that they were composite of other variables. Mean annual values were calculated for each variable for four seasons: 1) Winter (January - March), 2) Spring (Apri l - June), 3) Summer (July - September), and 4) Autumn (October - December). Historical data for the P D O were obtained from the University of Washington (Mantua and Hare, 2006). These were also averaged seasonally for each year. Climate variables and associated abbreviations are listed in Table 4-1. Graphs of each variable plotted over time were used to describe trends in the broad-scale climate data (temperature, precipitation) and finer scale climate data (PDO). 4.2.4 Statistical model The statistical model was created using Splus 7.0 (Insightful Corporation, 2005). The response variable, annual area disturbed ( A A D ) was regressed against climate variables listed in Table 4-1 using a general linear model method o f Poisson regression. The Poisson distribution typically is used for count data where there are a large number o f small values and few large values, and thus is non-normal. For example, Poisson regression was used to model number o f fires (Wotton et al., 2003). Although the fire data in this study are not count data, they -fit this distribution reasonably well (Figure 4-2 and Figure 4-3) given the many small disturbance years, and the few large disturbance years. The difference between normal regression and Poisson regression is that there is a transformation of the dependent variable, in this case a log transformation through a log link function in the regression. This transformation creates a linear relationship between, the dependent and independent variables that can be modelled using regression. Originally, normal regression was explored as a means to analyze the data. However, many of the large disturbance years appeared to be outliers greatly influencing the model, and the relationship was nonlinear. I f these "outliers" were removed from the data, multiple regression could be used, but only for disturbance years where total disturbance was less than 10,000 ha. The purpose of this study was to look at the effects of climate on all disturbances, especially the larger fire years, so Poisson regression was deemed a more appropriate modelling tool. 109 4.2.4.1 Trends in area disturbed and individual climate variables Scatterplots o f area disturbed and each climate variable were used to visually examine significant relationships between both the broad scale and fine scale climate data and annual area disturbed prior to choosing a regression model. 4.2.4.2 Cross-correlation analysis Cross-correlation analysis was carried out between P D O variables (Winter, Summer, Spring, Autumn) and A A D to determine i f lagged variables should be included in the regression model. Cross correlation analysis identifies similar periodic variation in two time series (Legendre & Legendre, 1998) such as a rise or fall in area disturbed in a year which corresponds to a particularly high or low climate index value. This point where the correlation is highest between the two variables is the lag time between the variables, and significantly high values were included in the analysis. The P D O signal occurs at longer time scales than other climate variables, and the purpose o f including this index was to pick up on longer drying or wetting trends which would affect disturbance levels in the study area. 4.2.4.3 Temporal autocorrelation Temporal autocorrelation o f the response variable, A A D , was examined using the autocorrelation function available in Splus. If a significant lag time existed, it was included in the final model as an autoregressive term (Insightful Corporation 2005). The Durbin-Watson statistic was used to test for autocorrelation in the residuals. Assumptions o f regression were examined using residual plots. 4.2.4.4 Regression model The final model was chosen using backwards stepwise regression methods. Initially all explanatory variables were included in the model. The regression.was then run again, and the climate variable which explained the least amount o f variation in area disturbed was removed. This was repeated until all variables included minimized the errors (the residual deviance) o f the model, indicating a good model fit (Insightful Corporation 2005). Variables were dropped by comparing the A I C (Akaike's information criteria) statistic for each variable against the A I C 110 statistic for the entire model. If the A I C for the variable was lower than the A I C for the model, then that variable did not improve the model fit (Insightful Corporation 2005). This method o f choosing the regression model ensures that the model selected is both free of bias, and parsimonious; too few variables included in the model would make it biased, while too many is not precise (Forster 2001). Explanatory variables were also selected that had low correlation with each other, to avoid problems o f multicollinearity (Stevens, 1996). 4.2.4.5 Mode l fit and predictive capacity The predictive capacity of the model was determined by graphing predicted versus actual area disturbed over time. Residuals from the model were checked to determine i f they met the assumption o f normality and equal variances. A n approximation of the r-squared value was obtained by the following formula (Insightful Corporation 2005): r = 1 -(residual deviance/ null deviance) (4-1) 4.2.5 Predicting mountain pine beetle disturbance with the statistical model Once the model had been specified for the period prior to the epidemic M P B outbreak (approximately 1999), the model was used to predict disturbance activity 2000-2003 when the M P B outbreak accelerated in this area. A A D was the annual area disturbed by mountain pine beetle during these 4 years. 4.3 R E S U L T S The time period analyzed using the Poisson regression was 1910-1999. Years prior to 1910 were excluded from the dataset because a significant 6-year crosscorrelation between annual area disturbed and mean spring P D O index was found, Because data were not available six years prior to 1904, it was decided to shorten the analysis time period. None o f the data variables were removed, only the time period was shortened. More recent years (2000-2003) were initially excluded from the analysis because climate station data were unavailable. This was a good opportunity however, when the data became available, to test the predictive capacity of the model in years with large area disturbed by mountain pine beetle. I l l 4.3.1 Summary of climate variability 1910-1999 Between 1910 and 1999, annual area disturbed ranged from 0 ha disturbed, to a year with 68,241 ha disturbed (Figure 4-4). Median annual area disturbed was 699 ha. Most o f the disturbance activity occurred prior to 1955, with very large disturbance years in 1922, 1930, and 1932, when total disturbance exceeded 40,000 ha. O f the years with less area disturbed, 15 years (all occurring after 1950) had less than 20 ha disturbed, 5 o f which had no disturbance recorded for either fire or mountain pine beetle. Maximum temperatures ranged from - 8 ° C in the winter to 24° C in the summer. Minimum temperatures at the climate station in Ft. St. James ranged from - 2 2 ° C in the winter, to 8° C in the summer. Minimum temperatures appeared to be increasing since 1910 although no trend was tested for in this analysis. There was no apparent trend in maximum temperatures. Precipitation fell as rain in all seasons, as did snow, although only sporadically in spring and summer seasons. Both rain and snow appeared to be greater in the first part of the century, and towards the middle, between 1940 and 1975. Maximum rainfall occurred in the spring and summer, although considerable rainfall was recorded in autumn as well. Maximum rainfall was 246 mm in the summer of 1959. The largest cumulative snowfall was in 1982, when 225 mm fell in the winter. The lowest snowfall in the winter occurred in 1993 when only 29 mm fell. The P D O was in a negative phase prior to 1925, a positive phase from 1925-1947, and a negative phase from 1947-1977 (Figure 4-5). These change points were more evident in the winter and spring data, as has been noted by other research (Mantua and Hare, 2002). The pattern between 1925 and 1947 in the summer was not consistently positive as in the winter and spring seasons. Similarly, the autumn P D O index was not consistently positive from 1977 onwards as in the other seasons. 4.3.2 Statistical model 4.3.2.1 Trends in area disturbed and individual climate variables The relationships between each explanatory climate variables and the response variable, annual area disturbed were explored using scatterplots. Variables included in the model are shown in Figures 4-6 and 4-7. . Higher maximum and lower fninimum temperatures in the spring, summer and autumn appear to have a linear trend with annual area disturbed in the study area (Figure 4-6a and b). Winter temperatures had little effect on total disturbance. Years with less rain than others in the spring, and summer tended to have less disturbance than years with more total 112 seasonal rainfall (Figure 4-6c). Similarly, years with less snowfall in the autumn and winter tended to have greater disturbance than years with greater snow. Only a weak trend is apparent in the P D O index in the summer in spring, with years o f greater disturbance occurring with more negative values (Figure 4-7). 4.3.2.2 Crosscorrelation Significant crosscorrelations were found between area disturbed and P D O in the summer at a lag of 2 years, and between area disturbed and P D O in the spring at a lag o f 6 years (Figure 4-8). This indicates that there was a correlation between area disturbed and the spring P D O index 2 years and 6 years prior to the disturbance year. In other words, the P D O phase in the spring affects area disturbed up to 6 years later. These crosscorrelations were included in the model. 4.3.2.3 Temporal autocorrelation Significant autocorrelation in the response variable, area disturbed, was found at 2, 3, 8 and 10 year lags (Figure 4-9). 4.3.2.4 Regression model The final Poisson regression model chosen was: A A D = 4.78 + 0.4l(MaxSum) + 0.17(MinAut) + -0.02(RainSpr) + -0.04(SnowAut) + -0.56(PDOsum) + -0.71 (PDOsum L A G 2 ) + 0.69(PDOspr L A G 6 ) + 0 .03(AAD L A G 3 ) (4-2) Crosscorrelations were included at lag 2 for P D O summer (PDOSUITILAG2), and at lag 6 for P D O spring (PDOsprLAG6), while annual area disturbed ( A A D ) 3 time periods earlier was included as an autoregressive term. Two variables (one for current, and for the lag time period) were included for the summer P D O in order to capture both interannual, and interdecadal effects o f this cycle. Mean maximum summer temperatures (MaxSum) and mean minimum autumn temperatures (MinAut) were predictors of annual area disturbed ( A A D ) . Total spring rain (RainSpr), and total autumn snow levels (SnowAut) were also important predictors, along with the mean summer P D O index (PDOsum) values. None of the variables were correlated above 113 0.40, indicating that multicollinearity was not a problem (Kozak, 1997). The Durbin-Watson statistic for this model was 1.57. 4.3.2.5 Mode l fit and predictive capacity When predicted area disturbed was plotted against actual area disturbed over time indicated the model correctly predicted large disturbance years (Figure 4-10), but that the predicted disturbance levels were not always of the same magnitude than the actual. The model also predicted greater disturbance in some years than the actual disturbance levels, such as around 1914 and 1915. In general, however, the model predicts years of low disturbance relatively well. These results showed that using climate variables alone, the model correctly predicted the predominance of disturbances from 1910 to about 1952, and the lack o f disturbance after 1952. In many o f the years the model correctly predicted minimal disturbance, in both the 1970's and 1990's. Around 1952 is the time when, according to historical documents, fire suppression began in this area. It is interesting to note that from this time onwards, there are few years when disturbance (fire) is predicted, but not present, which we would expect to occur i f fire suppression was effective. Years where this does occur are: 1954, 1963, 1969, 1986, 1989, and 1991. However, since there are years prior to 1952 where the predicted area disturbed is also higher than the actual area disturbed, fire suppression alone may not have caused this difference. Variables not included in this model may be influencing the results (see Discussion for more information). Assessment of the residuals also indicated a reasonable fit o f the model, although some cautionary points are noted. Deviance residuals plotted against the fitted model values were somewhat randomly scattered, albeit clustered around the lower fitted values (Figure 4-11). The normal probability plot indicated potential problems with the model. The model is overfitting a few small values and underfitting for large disturbance years (Figure 4-11). The Kolmogorov-Smirnov test for normality confirmed that the residuals are not normally distributed (p=0.02). Although the assumptions were not met, the model was used as a preliminary examination o f the relationship between climate and annual area disturbed, realizing that future model refinements should be explored. The approximation of the r-squared was 0.69. 114 4.3.3 Predicting mountain pine beetle disturbance The model did not predict years with mountain pine beetle very well from 2000 to 2003 (Figure 4-12). It over or under predicts in all years except 2000. This is discussed further below. 4.4 D I S C U S S I O N This study demonstrated that in the dry cool sub-boreal spruce ecosystem type, both short-term and long-term climate variables are strong drivers of annual area disturbed. Fire was the primary disturbance in the study area 1910-1999, although 1996 marks the beginning o f the large mountain pine beetle outbreak in the area. Both temperature and precipitation are known factors in moderating annual area burned in similar ecosystem types (Agee 1993). At a broad scale, it has been assumed that these variables wi l l be responsible for increasing disturbance frequency and magnitude in the central interior of B C under climate change. This study showed, however, that relatively constant temperature and precipitation patterns over time however, do not account solely for the cycling o f disturbances apparent in the last century. The different phases o f Pacific Decadal Oscillation (PDO) , occurring over regular but shorter time scales, have a very strong impact on annual area disturbed. Large disturbance years coincide with the positive phases of the P D O , over a period o f 20-30 years, as do smaller disturbance years coincide with more negative phases (Figure 4-5 and Figure 4-7). 4.4.1 Effect of short-term climate cycling- Pacific Decadal Oscillation Previous research has demonstrated the importance o f the P D O index on fire extent (Hessl et a l , 2004), and annual area burned (Duffy et al., 2005), and other work has described the relationship between other similar regional climate indices, such as the E l Nino (Heyerdahl et al., 2002) and mid-tropospheric circulation anomalies at 500hPa (Skinner et al., 1999). Hessl et al. (2004) found in Washington and Oregon that larger fires burned in years when the P D O was in a positive phase. Duffy et al. (2005) found the exact opposite in Alaska, where the impacts of the P D O are opposite the impacts at lower latitudes. At their study site, more area burned in years when the P D O was in a negative phase, because this cool phase was associated with drier summers. This study was not able to discern whether positive or negative phases o f the P D O affected area burned alone. Instead, it appeared that the interaction of seasonal P D O and time was an important predictor o f higher disturbance years. The P D O signal has been reported to be 115 strongest in winter (Mantua & Hare 2002) or the boreal spring (Minobe, 2000), affecting local climate conditions in the northwest during these times. In this study, positive spring (April -June) PDO indices lagged 6 years were positively correlated to annual area disturbed. Hessl et al. (2004) suggested that the influence o f a positive PDO phase is the lengthening o f the fire season; snowpacks are shallower during these times, and melt faster in the spring months. This would not explain the relationship between area disturbed and spring P D O with a time lag, unless the effects could persist over several years. Cycles of the spring P D O , whether positive or negative, occur over a period o f 5-7 years (Figure 4-5). During the positive phase a longer growing season means greater fine fuel build up, accumulating over a period o f about 6 years. The disturbance event affected by the positive spring P D O occurs 6 years later, during a negative summer P D O phase (when the spring P D O also may have switched). A signal o f the P D O in the summer has, to our knowledge, not been reported in the literature. We found that in the year in which the disturbance occurred, and 2 years immediately before the disturbance occurred, the summer PDO index was negatively correlated to area disturbed (Figure 4-7). We believe that during the year when the disturbance happens, the storm activity associated with the wetter, windier negative phase o f the summer P D O could generate a higher frequency o f lightning strikes, causing a greater number of disturbances overall. Presumably, i f the summer P D O signal is weak, there is only a moderate increase in hghtning activity, but that combined with the right fire weather and fuel conditions, creates ideal situations for larger areas to burn. What is most puzzling, however, is the time lag in the summer P D O of about 2 years. This suggests that the P D O phase 2 years prior to disturbance influences the total area disturbed. H o w cooler, wetter conditions 2 years before a disturbance occurs increases the total area disturbed is an area for further research. The only supposition possible is that there is an interaction o f moisture and fuels where short-term fine fuels temporarily increase in that two-year period, but given the shorter growing season and cooler temperatures, this is unlikely. It should be noted that even though the P D O is currently in a positive phase (which shifted in 1977) this alone was not sufficient to create climate conditions suitable for fires burning in the study area. Area disturbed since 1977 continued to be lower than the first part of the century even during this positive phase. As mentioned previously, the interplay o f the P D O with temperature and precipitation variables, as well as drought and fuel availability and moisture is critical to the development o f large fire years. 116 4.4.2 Extreme disturbance events The inability o f the Poisson regression to predict large disturbance years led to some interesting questions not considered at the beginning of this analysis. It was speculated that fuel build up could be an important predictor o f fire disturbance which was not in the model, simply because we did not have the historical data on vegetation required. Over the years, many studies have been published linking fire suppression and fuel build up in the later part o f the century, which can lead to increase in annual area burned. Although we do not believe that fire suppression was ongoing during the beginning of the century, we inferred that a natural fuel build up (because o f factors occurring prior to 1904) could have caused the few large years of greater disturbance in the data. However, one study found that fuels related to stand age were not as important as seasonal weather in predicting intense fires, which lead to greater annual area burned, in subalpine forests (Bessie & Johnson, 1995). Similarly in California and Oregon, researchers found that fuel build up, thought to be due to fire suppression, did not cause an increase in fire severity (Odion et al., 2004). One study in another sub-boreal ecosystem confirms that in fact disturbances which occurred prior to 1904 were o f similar overall magnitude (in terms o f area disturbed) as those years 1904-1953, with the exception o f the 3 large fire years (Delong, 1998). Fuel moisture, wind patterns and lightning activity can also affect the amount o f area burned in an area, but none of these were explicitly included in the regression model. Warmer fall temperatures and a shallower than normal snowpack would result in less soil moisture in the spring and conditions for low fuel moisture. In addition, low rainfall in the spring and warm summer temperatures would dry out the fuels even more. Wind events and possibly lightning activity are related to P D O ; in a negative phase (such as in the summer) storm activity would result in increased lightning and winds, which in combination with the warm temperatures and dry fuels, would result in a year with high fire activity. Since climate variables should account for some of the variability in area disturbed that is the result of fuel moisture, wind and lightning, we speculated that drought, a broader scale event occurring over many years, could be the reason why some years have more area disturbed than others. Duffy et al. (2005) found that the large fires in Alaska occurred when an E l Nino event produced warm, dry winters, and when there was a period o f persistent drought. Thus although many of the variables included in the regression are linked to the shorter and longer term 117 processes such as wind, lightning, and drought which may cause larger areas to be disturbed, the explicit inclusion o f these variables could improve the predictive capability of the model. Finally, it may be that additional climate data wil l improve the predictive capacity of the Poisson regression, especially for extreme events. By necessity, to encompass as broad a range in temporal data as possible, the Ft. St. James climate station was the source for temperature and precipitation data. This climate station is near, but not within the study area. The weather patterns in this area are certain to be different than those in the study area, as evidenced simply by the differences in forest composition. Whereas the dry cool sub-boreal tends to favour single species stands o f lodgepole pine which are able to grow in such dry conditions, forests around Ft. St. James have species such as Douglas-fir and spruce which survive in wetter and warmer conditions. On average, annual temperatures in Ft. St. James are about 0.5°C warmer than climate in Burns Lake (on the northwest boundary of the study area). Looking at mean temperatures seasonally, temperatures are similar in.the winter, but warmer in Ft. St. James in the summer (data not shown). Whether the magnitude (of less than a degree) o f this difference would affect results is unknown. In this study, the lack of regional climate data probably accounts for some o f the model error, which could be improved upon in the future by testing the model with more local data over a shorter time period. 4.4.3 Fire suppression and climate impacts This study suggests that fire suppression efforts, which are widely believed to be effective, have in fact had the luck o f good weather on their side in the past 50 years. If the projections of the regression model are somewhat accurate, fire suppression may have been effective in stopping fires in some o f those years where predicted area disturbed is greater than the actual area disturbed, such as between 1988 and 1994. Otherwise, this study demonstrated that both broad and smaller scale climate in the most recent 50 years (-1950-1999) did not favour years with large disturbance, as they had in the past (1910-1950). Other researchers have found a similar effect of climate in Canada, where area burned was greatly reduced due to climate cycles and local weather conditions (Girardin et al., 2004; Johnson et al., 2001; Weir et al., 2000). 4.4.4 Mountain pine beetle The regression model of disturbance and climate did not predict mountain pine beetle disturbance from 2000-2003. Most disturbance in the original model was fire disturbance as their 118 had been very little mountain pine beetle activity in the study area prior to 1996. It was assumed that similar climate variables might drive both fire and mountain pine beetle. This might include higher than average minimum fall temperatures, low snow cover, and warm summers. However, there may be other climate variables such as precipitation which in fact have the opposite effect on the two disturbances. Mountain pine beetle is also a different kind of disturbance than fire in that each year the new area of spread is dependant on the location of the previous years' activity. It therefore has much higher spatial autocorrelation than fires, which although they spread within a year, ignite on the landscape randomly, depending on lightning strikes. Future development of such disturbance-climate models o f this nature could develop models separately for each disturbance. General trends in, for example, P D O , could then be examined. 4.4.5 Conclusions and implications Since 1977 the P D O has shifted back to a positive phase, and, perhaps with the right temperature, precipitation, and drought conditions, large fife disturbances could be more likely to happen. However, climate is also a driver of mountain pine beetle, allowing the insect to expand at such a fast rate in the past few years. The build up o f suitably aged contiguous stands o f fire origin lodgepole pine has created conditions suitable for the mass spread o f the insect. Today, the mountain pine beetle wi l l attack any lodgepole pine, regardless o f its age (C. Hawkins, pers. comm.). A few years immediately after the mountain pine beetle attacks a tree, the probability o f fire in that stand increases because o f the dead needles. Over time as the needles drop, the chance of fire decreases until new regeneration and the fallen dead trees create suitable fuel for fire to spread again. Advanced regeneration in the understory also play a role in fire hazard of mountain pine beetle killed stands. For example, balsam fir and interior spruce may increase the fire danger as they are very combustible and have the potential to act as ladder fuels. Recent research published in the journal Science links large fires in the past 3 decades to climate warming trends (Westerling et al. 2006). Westerling et al. (2006) compared fire activity and climate 1970-1985 to fire and climate post-1985, and concluded that the fire season is now longer, and that fires are much larger and occur with greater frequency than in the earlier time period of the study. The P D O shifted phases around 1977, from a negative to a positive phase, 119 which coincides with the increased fire activity (and associated droughts) reported by the authors 1985-present. L ike the research presented in this chapter, there appears to be interactions of the P D O with other climate variables which are interacting to create climate conditions suitable for large fires. B y ignoring the interaction o f climate patterns on shorter and longer time scales (more than 2 possible cycles of events like the PDO), important drivers and cumulative effects of these drivers (e.g., interaction o f P D O and temperature) of disturbances may in fact be overlooked. Thus although it is imperative to consider climate in an analysis o f disturbance dynamics, it is essential also to understand the climate processes themselves, which often can affect disturbances at multiple scales. This study also has implications for our management practices, where we try to mimic natural disturbance events. Such management requires a natural disturbance baseline from which we can measure our successes and failures at emulating natural disturbances with harvesting. This baseline, called the range of variability, is quite often quantified during a time when human influences are presumed to be minimal on the landscape. In this case, disturbances prior to the 50's might meet that criterion. B y ignoring climate effects on disturbance however, the more accurate picture o f range of variability can be potentially missed. If the most recent 50 years were removed from the picture, one might think that high disturbance years were more common than i f the data were not truncated in this manner. In turn, the implications for future forest management may be that there is a larger range of uncertainty regarding natural disturbances than is often accounted for. Just as there may be periods o f muted disturbances, certain times will have elevated disturbance patterns, such as the current mountain pine beetle epidemic, which can have long-term consequences for both the economy and the environment. 120 4.5 T A B L E S Table 4-1 Climate variable abbreviations and description Abbreviations Variable Description MaxWin, MaxSpr, MaxSum, Mean maximum temperature Maximum monthly MaxAut temperature, averaged seasonally MinWin, MinSpr , MinSum, Mean minimum temperature Minimum monthly MinAut temperature, averaged seasonally RainWin, RainSpr , RainSum, Mean daily rainfall Mean monthly 1-day rainfall, RainAut averaged seasonally SnowWin, S n o w S p r , Mean daily snowfall Mean monthly 1-day snowfall, SnowSum, SnowAut averaged seasonally PDOwin, P D O s p r , PDOsum, Pacific Decadal Oscillation Monthly PDO, averaged PDOaut seasonally 121 4.6 F I G U R E S 127 W W 126t(TW 125WW 124°0"0"W Figure 4-1 Location of study area in BC. The outside boundary of the study area, shown in white, follows the dry cool Sub-boreal Spruce biogeoclimatic zone (SBSdk). Other ecosystems within this boundary were excluded. Dark gray areas are lakes and streams. Lighter gray areas are ecosystems outside of the study. 122 16001 24001 32001 40001 48001 Annual area disturbed (ha) 56001 64001 Figure 4-2 Histogram of annual area disturbed (AAD). The x axis shows the size of disturbance as a % of the total number of years (89 years). Over 50% of the disturbance years had very small area disturbed. 123 12.H c o o -20000; 40000 ieogop ; Annual area disturbed,(ha): Figure 4-3 Lognormal probability plot of annual area disturbed. 80000 60000 T3 CD .Q 3 40000 fO TJ CO CD 20000 ifflrrll i k b J M l J L x i L l L ' i i 1 1 — r ~ — i r 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 YEAR Figure 4-4 Annual area disturbed (ha) 1910-1999. Disturbance each year includes both fire and mountain pine beetle disturbance. 125 1910 1930 1950 1970 1990 Figure 4-5 Mean PDO index by season, 1910-1999. 0 t— to TS 1.0e1 8.0e0 1 ' * • • • co 6.0e0 " O "co co 2.0e0 -O.OeO " a) 16 20 Temperature ( C) 24 w T3 1.0e1 S 8.0e0 1 CO 4.0eO -| o co CO 2.0e0 1 O.OeO ' -18.5 -16.0 -13.5 -11.0 -8.5 -6.0 Rain precipitation (mm) -3.5 T3 CD _Q -*-» CO 8.0e0 ' CO CD 0 > o "co CO 2.0e0 " b) 0.0 0.5 1.0 1.5 2.0 Temperature ( C) 2.5 T3 CD -Q i T : CO CD D J O "co I 1.0e1 i 8.0e0 2.0e0 O.OeO d) • . • • I0 0.25 0.50 0.75 1.00 1.25 1.50 Snow precipitation (mm) 1.75 Figure 4-6 Scatterplots of broad-scale climate variables (y axis) and natural log transformed area disturbed (x axis), 1910-1999: a) Mean maximum summer temperature (°C), b) Mean minimum fall temperature (°C), and c) Mean spring rain precipitation (mm) and d) Mean fall snow precipitation (mm). - 0 Spring P D O index.. -i, o ;i S u m m e r P D O Index ' • 0 " ' 1 S 3.0e3 H Ti 3,0e3 -5 5e1 --1 0 -1 ispf ing PDO.Index> •-% 0 1. S u m m e r PDO'Index Figure 4-7 Spring and Summer P D O index and the natural log of annual area disturbed (ha). efoss?C6rrelatibn;Rl6t' eross;C^rrel^idritP16t Figure 4-8 Crosscorrelation of annual area disturbed and a) summer P D O index, and b) spring P D O index. Correlation values are n the y axis. At lag 0 this is the Pearson Correlation statistic. The lines are the 95% confidence intervals. A value outside this line is a significant cross-correlation. 128 o Lag Figure 4-9 Autocorrelation of annual area disturbed over time. ACF is autocorrelation function. The lines are the 9 5 % confidence intervals. A value outside this line indicates significant autocorrelation. 129 80000 60000 CD _o 3 40000 03 20000 Actual area disturbed Predicted area disturbed 0 ^ ' iT iYmyMVi VW^^ ^ iTfTlfi'i ifiTTfiTtttftttittttTTTITTITITT 1910 1916 1922 1928 1934 1940 1946 1952 1958 1964 1970 1976 1982 1988'1994 YEAR ttttr Figure 4-10 Actual disturbance levels vs. predicted disturbance levels. C r a y bars show the actual area disturbed each year. The black line and triangles show the predicted area disturbed from the regression model. 20C00 '30000 F^itted Values a) b) Figure 4-11 a) Quantile residuals plotted against fitted model values, b) Normal probability plot of quantile residuals. 130 80000 60000 CD: JO JO ZD . 1. 40000 • 20000-o • 2000 2001 •HH H i iiilllill iilllllli i l s l • M M H i - , 1PP1P|||1^  •flMI •••wl • • H •hi 2002 2003 Figure 4-12 Actual disturbance, primarily mountain pine beetle, versus predicted disturbance from the Poisson regression model. Gray bars show the actual area disturbed each year. The black line and triangl show the predicted area disturbed from the regression model. 4.7 R E F E R E N C E S Agee, J. K . (1993). Fire Ecology of Pacific Northwest Forests. Washington, D C , U S A : Island Press. B C Ministry of Forests. (1995). Forest Practices Code of British Columbia, Biodiversity Guidebook. Victoria, B C ; B C Ministry o f Forests. Beisner, B . E . , Haydon, D . T., & Cuddington, K . (2003). Alternative stable states in ecology. Frontiers in Ecology and Environment. 1(7): 376-382. Bergeron, Y . , Gauthier, S., Kafka, V . , Lefort, P., & Lesieur, D . (2001). Natural fire frequency for the eastern Canadian boreal forest: Consequences for sustainable forestry. Canadian Journal of Forest Research. 31: 384-391. Bessie, W., & Johnson, E . (1995). The relative importance o f fuels and weather on fire behaviour in subalpine forests. Ecology. 76(3): 747-762. Bond, N . , and Harrison, D . E . (2000). The Pacific Decadal Oscillation, air-sea interaction and central north Pacific winter atmospheric regimes. Geophysical Research Letters. 27(5): 731-734. Davis, M . B . (1986). Climatic instability, time lags, and community disequilibrium. In J. Diamond, & T. J. Case (Eds.), Community Ecology. N e w York: Harper and Row Inc. pp.665. Davis, M . B . , & Shaw, R. G. (2001). Range shifts and adaptive responses to quaternary climate change. Science. 292: 673-679. Delong, S. C. (2002). Natural Disturbance Units of the Prince George Forest Region: Guidance for Sustainable Forest Management. Prince George, B C : B C Ministry of Forests and Range. Delong, S. C. (1998). Natural disturbance rate and patch Size distribution of forests in northern British Columbia: Implications for forest management. Northwest Science. 72: 35-48. 132 Duffy, P. A . , Walsh, J. E . , Graham, J. M . , Mann, D . PL, & Rupp, T. S. Impacts of large-scale atmospheric-ocean variability on Alaskan fire season severity. Ecological Applications. 15(4): 1317-1330. Elmqvist, T., Folke, C. , Nystrom, M . , Peterson, G. , Bengtsson, J., & Walker, B . et al. (2003). Response diversity, ecosystem change, and resilience. Frontiers in Ecology and the Environment. 1(9): 488-494. Environment Canada. 2005. Adjusted Historical Canadian Climate Data. Retrieved Apri l 2006 from: http://www.cccma.bc.ec.gc.ca/hccd/ Gedalof, Z . , Mantua, N . , and Peterson, D . L . 2002. A multi-century perspective o f variability in the Pacific Decadal Oscillation: new insights from tree rings and coral. Geophysical Research Letters. V o l 29(24). Art . N o . 2204. Gedalof, Z . , Peterson, D . L . , and Mantua, N . 2005. Atmospheric, climatic, and ecological controls on wildfire years in the northwestern United States. Ecological Applications. 15(1): 154-174. Girardin, M . P., Tardif, J., Flannigan, M . D . , Wotton, B , M . , & Bergeron, Y . (2004). Trends and periodicities in the Canadian drought code and their relationships with atmospheric circulation for the southern Canadian boreal forest. Canadian Journal of Forest Research. 34(1): 103-119. Grimm, V . , & Wissel, C. (1997). Babel, or the ecological stability discussions: A n inventory and analysis o f terminology and a guide for avoiding confusion. Oecologia. 109: 323-334. Hansen, A . J., Neilson, R. P., Dale, V . H . , Flather, C. H . , Iverson, L . R., & Currie, D . J. et al. (2001). Global change in forests: Responses o f species, communities, and biomes. Bioscience. 51(9): 765-779. Hessl, A . E . , McKenzie , D. , & Schellhaas, R. (2004). Drought and pacific decadal oscillation linked to fire occurrence in the inland Pacific Northwest. Ecological Applications. 14(2): 425-442. 133 Heyerdahl, E . K . , Brubaker, L . B . , & Agee, J. K . (2002). Annual and decadal climate forcing o f historical fire regimes in the interior Pacific Northwest, U S A . The Holocene. 12(5): 597-604. Holling, C. S. R. (2001). Understanding the complexity o f economic, ecological, and social systems. Ecosystems. 4(5): 390-405. Insightful Corporation. (2005). S-PLUS 7.0 Guide to Statistics, Volume 1. Seattle, W A : Insightful Corporation. Johnson, E . A . , & Gutsell, S. L . (1994). Fire frequency models, methods and interpretations. Advances in Ecological Research. 25: 239-287. Johnson, E . A . , & Larsen, C. P. S. (1991). Climatically induced change in fire frequency in the southern Canadian Rockies. Ecology. 72(1): 194-201. Johnson, E . A . , Miyanishi, K . , & Bridge, S. R. J. (2001). Wildfire regime in the boreal forest and the idea o f suppression and fuel buildup. Conservation Biology. 15(6): 1554-1557. Johnson, E . A . , & Van Wagner, C . E . (1985). The theory and use of two fire history models. Canadian Journal of Forest Research. 15: 214-220. Kozak, A . (1997). Effects of multicollinearity and autocorrelation on the variable-exponent taper functions. Canadian Journal of Forest Research. 27(5): 619-629. Landres, P. B . , Morgan, P., & Swanson, F. J. (1999). Overview of the use o f natural variability concepts in managing ecological systems. Ecological Applications. 9(4): 1179-1188. Legendre, P., & Legendre, L . (1998). Numerical Ecology. Second English Edition. Developments in Environmental Modelling. 20: 1-853. Mantua, N . J. C , & Hare, S. R. C. (2002). The Pacific Decadal Oscillation. Journal of Oceanography. 58(1): 35-44. Mantua, N . and Hare, S. 2006. Pacific Decadal Oscillation Data. Retrieved Apr i l 2006 from: http://www.jisao.washington.edu/pdo/ 134 McKenzie, D . , Peterson, D. W. , Peterson, D. L . , & Thornton, P. E . (2003). Climatic and biophysical controls on conifer species distributions in mountain forests o f Washington State, U S A . Journal of Biogeography. 30: 1093-1108. Minobe, S. (2000). Spatio-temporal structure of the pentadecadal variability over the North Pacific. Proceedings in Oceanography. Al (2): 381-408. Odion, D .C . , Frost, E.J . , Strittholt, J.R., Jiang, H . , Dellasala, D . A . , & Moritz, M . (2004). Patterns of fire severity and forest conditions in the Western Klamath Mountains, California. Conservation Biology. 18(4): 927-936. Oliver, C. D . , & Larson, B . C. (1996). Forest Stand Dynamics. New York : John Wiley & Sons, Inc. Peterson, G . , Al len , C. R., & Holling, C . S. (1998). Ecological resilience, biodiversity, and scale. Ecosystems. 1: 6-18. Pierce, J. L . , Meyer, G . A . , & Jull, A . J. T. (2004). Fire-induced erosion and millennial-scale climate change in northern ponderosa pine forests. Nature. 432: 87-90. Sanborn, P., Geertsma, M . , Jull, A . J .T . , and Hawkes, B . 2006. Soil and sedimentary charcoal evidence for Holocene forest fires in an inland temperate rainforest, east-central British Columbia, Canada. The Holocene. 16(3): 415-427. Skinner, W. R. C . M . , Stocks, B . J. C . M . , Martell, D . L . C . M . , Bonsai, B . C . M . , & Shabbar, A . C. M . (1999). The association between circulation anomalies in the mid-troposphere and area burned by wildland fire in Canada. Theoretical and Applied Climatology. 63(1): 89-105. Stevens, J. (1996). Applied Multivariate Statistics for the Social Sciences. Mahwah, N . J . : Lawrence Erlbaum Associates. Veblen, T. T. (2003). Historic range o f variability of mountain forest ecosystems: Concepts and applications. Forestry Chronicle. 79(2): 223-226. Wang, G . , & Schmimel, D. (2003). Climate change, climate modes and climate impacts. Annual Review of Environment and Resources. 28: 1-28. 135 Weir, J. M . H . , & Johnson, E . A . M . K . (2000). Fire frequency and the spatial age mosaic of the mixed-wood boreal forest in western Canada. Ecological Applications. 10(4): 1162-1177. Westerling, A . L . , Hidalgo, H . G . , Cayan, D. R., Swetnam, T.W. (2006). Warming and earlier Spring increases western U . S . forest wildfire activity. Published online July 6 2006; 10.1126/science.l 128834 (Science Express Research Articles) at: www.scienceexpress.org White, P. S., & Pickett, S. T. A . (1985). Natural disturbance and patch dynamics: A n introduction. The Ecology of Natural Disturbance and Patch Dynamics. S.T.A. Pickett & P.S. White, eds. 27-42. Wotton, B . M . L . , Martell, D . L . L . , & Logan, K . A . L . (2003). Climate change and people-caused forest fire occurrence in Ontario. Climatic Change. 60(3): 275-295. Wu, J., & Loucks, O. L . (1995). From balance o f nature to hierarchical patch dynamics: A paradigm shift in ecology. The Quarterly Review of Biology. 70(4), 439-466. 136 C H A P T E R 5 D A T A A C C U R A C Y Although Vegetation Resources Inventory (VRI) data is commonly used in B C , especially for timber supply analysis its reliability and accuracy is not typically reported in a transparent fashion. Specifically, the government and/or the contractors do not provide an error assessment as part o f the metadata. So, although the data is widely used, there is very little understanding o f the error associated with this data. In addition, the Natural Disturbance Database is a new compilation of data, and it is important to better understand the potential for error in this database. Field data collected by U N B C M S c student, Patience Rakochy, were used to assess the accuracy o f the V R I and M P B data. Data were collected from 48 randomly located M P B attacked stands in and around the Cheslatta Community Forest on mesic and submesic SBSdk site series. A l l stands were pine-leading, and between 100-160 years o f age. Each sample plot was located using a GPS, allowing the plot information to be linked to a GIS. The plot level data described above provides a means of assessing the accuracy of the V R I . data, which is used in turn to validate the N D D fire data. The limitations o f the available data dictate this approach, and the results o f this accuracy analysis should be viewed as preliminary and used with caution. The V R I and M P B accuracy was assessed by assuming that the field plots reflected the "true" stand conditions. A n y stands that did not have similar characteristics as the plot data were therefore assumed to be incorrectly classified. Errors in stand age, stand composition in the V R I dataset, and attack severity in the M P B dataset were identified. The V R I data had been classed into 10 year age classes, while the field data were in 20 year age classes. The V R I data were reclassified accordingly for comparison. Stands 2 or more age classes (40 years) different than the field data were considered incorrect. Considering the coarse grain of stand ages in the inventory (which are largely derived from aerial photos) making the assumption that a one-age class difference was correct did not seem unreasonable. Stand composition, as identified by leading species, was considered correct only i f identical to the field plot since it is a general stand characteristic more readily discernible from aerial photos (it either had the same leading species or it did not). The fire dataset was assessed by assuming that the stand age class (using 10-year age classes) o f the V R I stand was correct. If the date o f a fire was inconsistent with the current stand age then the fire date was considered incorrect. For example, i f a fire date was recorded in the dataset as 1930 and the current V R I stand was age class 15 or (150-160 years old), then fire date 137 was considered error. Because both datasets had spatial extent (the field data above were only points) then the error was based on area, rather than number o f stands. Dataset error was therefore the area of incorrectly classified stands with fire disturbance expressed as a % o f the total area burned (over all time). A l l error reported below were "false negatives" or errors o f omission, that is, absence o f the value when it should have been present. "False positives" or errors o f commission (presence of the value when it should have been absent) were not assessed because the data were not available for these calculations. Field data collected measured only presence data (of a particular age class, species type, and mountain pine beetle infestation level) so a complete picture of absence was not possible. 5.1 S T A N D A G E Only 15 o f the original 48 stands from the field could be used to verify the V R I age data. Stands in the V R I were, classified based on current age of the live canopy. When the field data were collected, stand age was recorded as the age o f the three trees most representative of stand age (recorders interpretation) even i f the trees were dead. Since most o f the pine were dead, stand age in the field data is the stand age at time of death. Only stands which were contained a live component at time of the V R I classification could be compared to the field data. In the end, only 15 o f the 48 field plots were able to be compared to the V R I data for age class comparisons. However, all 48 stands could be used to verify the accuracy of the cover type data. O f the 15 V R I stands, 60% were within 1 age class of the field data age classification, the remaining 40% of the stands were not (Table 5-1). A l l differences between V R I and the field data showed the V R I classification to be older than the field data. The magnitude o f the error was surprising, but in hindsight should not have been given the methodology used to assemble the V R I data. Inventory ages are estimated from aerial photos, sometimes weighted by basal area, and based on an interpreter's experience with similar stand types. Some of the V R I is ground truthed with field plots, but not the entire inventory. I f there were a few larger, older trees, perhaps left as remnants from a previous disturbance, the stand age could be overestimated. Since all the stands in the V R I were older than the field data, this suggested bias in the V R I towards older stands. That is, based on this error assessment, it can be said that V R I classification consistently overestimates stand age. Furthermore, the field data ages were purposely collected so as to represent the mean stand age, neither older nor younger than 138 the overall stand (Rakochy 2005). Clearly the V R I data were not very accurate in terms of stand age, although the error is systematic bias rather than random error. The greatest impact of this source of error is on the identification of historical disturbances using the V R I age data, as used in Chapters 2 - 4 . Disturbances were identified based on stand age, between 51 and 100 years old (translating into disturbances between 1904-1953). If these stand ages are incorrect, as shown from the error assessment, 40% o f the time, then potentially almost half of the disturbances did not occur during this time period. These disturbances would have happened post-1954. The implications o f this is that disturbances are potentially overestimated in the historical time period, since the overestimated V R I age would imply stand origin in a disturbance that occurred earlier than was really the case. Looking at the results from Chapter 2, this would suggest that the difference between total disturbance 1954-2004 and total area disturbed 1904-1953 could be much less. For Chapter 3, the age class distribution o f the landscape at the start of the model runs would be similarly biased towards older stands about 40%> of the time. This would result in the area o f stands in the mature and old age classes being overestimated. Therefore, there would not be as much o f a difference between current conditions and those projected by the model for 150 years from now. For example, in Scenario 2, a 26% reduction in mature serai stage occurs by period 15 as compared to the starting condition; the observed errors would imply that this difference is overestimated. Annual area disturbed as reported in Chapter 4 for 1904-1953 would also be affected by this source of error. From a management perspective, return intervals may in fact be longer' than we have previously believed, especially in pine ecosystem types. The landscape would also be older, with fewer early and young stands. Overall, disturbance may have changed less in the past 100 years than we had previously thought. Briefly, i f this accuracy assessment is true the conclusions o f the chapters would be affected as follows: 1. Chapter 2 - Disturbance frequency in most recent 50 years may in fact be higher than historical times. 2. Chapter 3 - The long-term landscape would be older than estimated from the modelling exercise. 3. Chapter 4 - Could potentially change the importance or influence of climate variables in the Poisson regression model because there would be less area burned than was reported in the historical time period. 139 The reliability o f this error assessment is limited by the small number o f stands available for comparison. Further investigation o f the error characteristics of the V R I seems warranted, however. This should be performed using the original V R I data, not the final post-processed dataset as was used in this assessment in combination with additional field data. Errors in the V R I and related datasets could provide a misleading view of current conditions and disturbance history, leading to flawed management and policy decisions. 5.2 C O V E R T Y P E Cover type in the V R I data were classified correctly 85% of the time (n=48 stands; Table 5-1), with 15% of the stands classified as spruce leading (and one as shrub) when they were pine leading. These errors seem relatively minor, considering the amount of dead pine on the landscape in the study area. I f stands were dead at the time of the inventory they would be classified as non-pine leading, especially if there was a significant understory. Given this possibility, and that no data is without error, the V R I data appears accurate in terms o f stand composition. 5.3 M O U N T A I N P I N E B E E T L E D A T A Mountain pine beetle attack levels were classified correctly 80% of the time (n = 48 stands; Table 5-1). Twenty percent of the stands were classified as having attack levels < 30% when in fact they were greater. This could be because the mountain pine beetle data were collected earlier (pre-2005) than the field data. Since the mountain pine beetle attacks in waves, stands with lower attack levels have higher attack levels over time. If however, this is untrue the total number of hectares killed by mountain pine beetle in the most recent decade in the analysis would be underestimated 20% of the time. 5.4 FIRE D A T A Only 46% of the fire data from the Natural Disturbance Database were classified correctly when compared to the V R I age class data (Table 5-1). About 36% o f the fire data identified fires in V R I stands that had age classes older than the time since fire. Another 18% o f the fire data identified fires in V R I stands that had age classes younger than the time since fire. Given inaccuracies of the V R I dataset, this does not necessarily mean that 54% o f the fire data were incorrect. A portion of this error (up to 39%, or 14% o f the 36%>) can be attributed to 140 V R I inaccuracy, but only in older stands. The remaining 61% (or 22% of the 36%>) of the inaccuracy (as a rough estimate) might also be attributed to the method of mapping fires prior to the 1950's. Fire boundaries may be rough approximations o f the fire's location, and could include patches of unburned remnant stands. Some error could also be attributed to digitizing errors between the two data sources, although this is likely small. On the other hand, stands which are younger since the time since fire are less problematic than the older stands because an intuitive explanation can be given. If the stand did not promptly regenerate after fire, some of this error may be a regeneration delay, sometimes longer than a decade. Activities such as salvage harvesting or rehabilitation o f fire origin stands are other reasons why the stand is younger than the time since fire. Furthermore, it is likely there was some error that can be attributed to fire reporting inaccuracies in the early part o f the century. For the purposes of quantifying error in each o f the 3 studies in this dissertation (Chapters 2-4), the fire data has an approximate overall accuracy of ± 18-22%. About 22% of the error is from underestimating the time o f fire, but ignoring error from the V R I inaccuracies, and 18%> is from overestimating the time o f fire. The implications are that fire disturbances could be lower or higher than reported. However, given the magnitude o f some o f the disturbances, it is unlikely to change the overall temporal pattern in disturbances. Certainly, it would be preferable to have more accurate data, but given the difficulties of obtaining spatial historical data, using the available data has allowed an approximation of disturbance dynamics in the study area. 5.5 R E C O M M E N D A T I O N S R E G A R D I N G E R R O R Any researcher should exercise great caution when using spatial data, especially if the error is not reported from the source. But the risks must be weighed against the benefits of having even imperfect spatial data. Clearly, the benefits are the availability of spatial data with a large number o f attributes for the entire province, even for areas where it is impossible or impractical to collect detailed field data. The risks are the types o f inaccuracies reported above, wit the potential implications for analysis and management decision-making which I have discussed. Government has not, to our knowledge, committed itself to making these inaccuracies transparent to data users and therefore there is a risk of flawed analyses and bad decision-making as a result. In this study, it was crucial to use spatially delineated data in order to reconcile (spatially) many data sources, and thus the benefits outweighed the potential risks. 141 Figure 5-1 shows % area disturbed, by decade, for each disturbance type (from Chapter 2), but with the addition of error bars, based on the error assessment described above. A l l error bars were generated based on the calculation of error for age, cover type and mountain pine beetle. When viewed in this manner, while it appears that error potentially affects the amount of area disturbed, it does not significantly change the overall temporal pattern in the data. The final results and recommendations of this dissertation, therefore, can be considered valid in spite of the potential data errors which have been identified. 142 5.6 TABLES Table 5-1 Error quantification for stand age, cover type and mountain pine beetle data. The number and percent of incorrectly classified stands, as well as the number and percent of correctly classified stands is presented. Stands are incorrect if they had a "False negative" that is had the absence of the classification when it should have been present. Total # Stands Incorrect Stands Correct Stands (False negative) (True positive) # # % # % Stand age 15 6 40 9 60 Cover type 48 7 15 41 85 Mountain pine beetle 48 6 13 42 87 Table 5-2 Error quantification for fire. The hectares and percent of incorrectly classified stands, as well as the hectares and percent of correctly classified stands are presented. Stands are incorrect if they had a "False negative" that is, had the absence of the classification when it should have been present. Also shown is the area and % of stands which were older and younger than the year of fire. Total area Stands Incorrect Stands Correct burned (False negative) (True positive) ha ha % ha % All fire 131,772 71,131 54 60,641 46 Stands older than year of fire 47,653 36 Stands younger than year of fire 23,478 18 143 5.7 FIGURES • Fire • Harvest • MPB • Unknown Figure 5-1 % area disturbed, by decade for each disturbance type (from Chapter 2). Error bars are as follows: ± 20% for fire (averaged of range of 18-22%), - 40% for unknown, and + 20% for MPB. Error was not assessed for harvesting. The error bars were calculated from an error assessment comparing digital data to field plots, and comparing separate digital datasets. 144 C H A P T E R 6 SUSTAINABLE F O R E S T M A N A G E M E N T , N A T U R A L DISTURBANCES AND C L I M A T E C H A N G E 6.1 I N T R O D U C T I O N The goal o f this dissertation is to provide insight into the historical variation in natural disturbances in the dry cool spruce ecosystem type in the central interior of British Columbia. Recently, this area has been affected by a severe mountain pine beetle outbreak, killing most all of the mature lodgepole pine (over 80%), which make up over 50% of the stands in this ecosystem. A s a direct result o f the outbreak, management practices in the area have come under close scrutiny. Fire suppression, in particular, is a practice which many scientists believe has led to an overabundance of mature, highly-connected host stands for the mountain pine beetle, contributing to the magnitude and severity of the outbreak. B C mandates management of forests in a manner compatible with natural disturbance regimes, under the umbrella concept of sustainable forest management. Several policy documents (e.g., the B C Biodiversity Guidebook, the Lakes Land and Resource Management Plan, and the Morice Sustainable Forest Management Plan) provide guidance with specific indicator targets for each ecosystem, often written in a regional, rather than provincial, context. The idea is that, even though the majority o f fires wi l l be suppressed, timber harvesting wi l l provide a substitute mechanism o f disturbance, such that there is a natural level of disturbance over time. The anticipated result is that the forests will be healthy. A policy of natural disturbance management (harvesting similar to disturbances, and fire suppression) has been in place for over 10 years in B C , yet there remain many gaps in our understanding of natural disturbance dynamics and in management policy that need to be addressed. In the study, historical disturbance variability has not previously been quantified. The impacts of our current policies on the current age class distribution are unknown, and the role o f climate in disturbance temporal variability has not been closely examined. The studies in this dissertation bridge some of these gaps in our knowledge at the landscape level. Using new disturbance information from the Canadian Forest Service, stochastic modelling techniques, and climate indicators, the three studies build upon the results o f each in succession. In Chapter 2, I explored the historical range of variability in disturbances, and compared this to current disturbances and landscape conditions. In Chapter 3, I used the historical range of variability and a forest management model to deterrnine possible impacts on the landscape i f we were to manage (harvest) at the same rate as historical disturbances. Finally, 145 in Chapter 4, I examined the link between disturbances in the sub-boreal and broad and smaller scale climate variables. Because the studies were completed in chronological order as they appear in the dissertation however, the conclusions of each chapter relate only to that particular study and the one preceding it. This final chapter places the results o f the individual studies into an overall context. 6.2 S U M M A R Y OF M E T H O D S In the first study (Chapter 2), the historical range o f variability in fire and mountain pine beetle was quantified for the period 1904-1953, and 1954-2004. Disturbance indicators (e.g., % area disturbed, and disturbance return interval) and landscape condition indicators (e.g., % early serai stage) were calculated. Indicators for both time periods were compared, and then used to determine the compatibility o f current management policy with the historical disturbance regime. The main assumptions of this study re that: 1) there is a natural regime of historical disturbance resulting in a natural landscape condition, which occurred in the absence of substantive human influence, and 2) current disturbance and landscape condition reflects an era of considerable human intervention, and is not natural (if humans are considered unnatural). The second study (Chapter 3) explored the potential impacts o f a historical disturbance regime as i f it occurred on the landscape today using a stochastic simulation through optimization model. The purpose of this study was to explore the impact of disturbance on the age class distribution o f the landscape i f management activities truly mimicked the historical disturbance frequency. This was accomplished by simulating forest fires using Markov chains and a forest management model. Fire probabilities were generated from the historical fire data, and the model was run over 15 decades (150 years), assuming that the disturbance rate was constant for the entire analysis period. The third and final study (Chapter 4) was designed to look at the link between climate and disturbance over the 100 year time period (1904-2004), thereby introducing another dimension which had not been considered in the previous 2 studies. While Chapters 2 and 3 assumed that climate was basically stable over the entire study period of 100 years, and that humans were what caused visible and quantifiable changes to the landscape. Chapter 4 took the opposite view that human impacts were niinimal (at this scale) and climate was variable. Both broad scale climate and fine scale climate drivers were included in the analysis. A t the broad scale, the Pacific Decadal Oscillation has been linked to fire activity in the Pacific Northwest, 146 while temperature and precipitation are drivers o f disturbance at finer temporal scales. A Poisson regression analysis was used to further elucidate the relationship between annual area disturbed and climate variables. 6.3 K E Y F INDINGS 1. Climate influences changes in fire disturbance and landscape condition more than management The driving agents o f change of the disturbance regime in the study area are broad and fine scale climate processes, not management activities such as harvesting or fire suppression. This point was illustrated in Chapter 4, where the results from the Poisson regression model in the latter half of the 20 t h century were predicted to have lower annual area disturbed than the previous 50 years, indicating climate conditions were not as conducive to fire during this period. Reduced fire activity during this period has often been attributed to fire suppression, but this study supports other recent research which shows how smaller fire years are related to changes in temperature, precipitation, and the Pacific Decadal Oscillation. This research also shows the strong effects of the P D O , which are variable over many decades, and how these effects can overshadow potential global climate change impacts. In this manner it contradicts the recent findings of research in the which found that large fires in the southwestern U S A were related to broad-scale climate changes, rather than the effects of climate cycles such as the P D O and E l Nino. In the next 50 years, climate may once again be conducive to large burn years; this was modelled in Chapter 3. A s disturbances occur every decade in the model, the landscape becomes much younger than it is today, moving towards an equilibrium o f stands concentrated in the younger age classes, with fewer in the mature and older age classes. Given that the mountain pine beetle has destroyed vast tracts o f mature pine, the cumulative effects o f multiple disturbances could be result in a much larger area disturbed than seen in the past 100 years. However, it is also a possibility that the current mountain pine beetle outbreak is replacing large fires as a disturbance mechanism at this time, especially since it is thought that mountain pine beetle killed trees lose their fire hazard for many years once the needles have dropped. 2. Quantification of historical range of variability must include climate effects 147 The results from Chapter 4 demonstrated the importance of including climate factors when trying to quantify the historical range o f variability in disturbance. The time period when disturbances should be considered when quantifying the historic range of variability should not be constrained to the beginning of the 20 t h Century, but rather should extend at least 100 years from the present (1906-2006). This is because fire disturbance has not been as manipulated by fire suppression efforts as previously thought. Rather, climate has been effective at reducing the area burned in the most recent 50 years. Thus even though human management activities impact the landscape, i f we are looking at fire disturbance alone, even today we are experiencing a "natural" frequency o f disturbance. This implies that disturbance frequency is in fact lower than reported in Chapter 2, while the return intervals are longer. Similarly, the amount o f early serai stage should be less than reported. Given the mountain pine beetle outbreak however, and the possibility of other pests such as spruce bark beetle, the level of disturbance may be relatively stable, whereas the agent of disturbance has changed. These agents may or may not have different climate triggers, which this research was unable to quantify. Caution must be taken when trying to quantify the range of variability in disturbances because of these many confounding factors. 3. There is a need to change the way B C sustainable forest management policy is developed by changing the way we look at the historic range of variability By replicating the methods used to quantify the historical range o f variability in disturbances and landscape conditions, Chapter 2 demonstrates the need for further and ongoing refinement of our policies as new information on disturbances becomes available. For example, guidelines on return intervals and disturbance frequencies need to be tied to species specific guidelines, in order to ensure that the age class distribution o f each species is maintained. Commercial species often are targeted over other species, leading to an unnatural age class distribution in non-commercial species, which become older and older over time. In addition, forest management policy suggests that the only way to quantify baseline conditions is to study disturbances that occur in areas free from human influence. B y necessity, these disturbances must be historical disturbances, since few areas on Earth today are untouched by humans. At a broad scale, Chapter 4 showed how humans have had less influence on disturbance dynamics in the past than climate has. This suggests that quantification of disturbance variability under static climate conditions, or at least climate conditions which can be documented, is more important than quantification of these disturbances in the absence of 148 human management. Furthermore, the idea that Aboriginal people had little impact on disturbances is increasingly being questioned, meaning that considering disturbances prior to European settlement to be without human influence may not be correct. Finally it should be recognized that we cannot harvest at the same rate as stand replacing events such as fife, nor should we attempt to do so. Chapter 2 demonstrated that even under a policy allowing harvesting at rates similar to natural disturbances, harvest levels did not approach the average area disturbed between 1904-1953. Were it possible, it should still not be attempted, because the results of Chapter 4 indicate that under certain climate conditions, fires (or mountain pine beetle) will be beyond human control, and wi l l disturb forests despite best efforts of us to keep it in check. B y harvesting at elevated levels the landscape could easily be pushed beyond its level o f resilience (historic range of variability). 4. Continue research into prescribed burning, and consider cumulative effects After the large fire year in many parts of the province in 2003, a report was commissioned by the Government of British Columbia to look into why so much area was burned, especially in the wildland-urban interface areas. The report recommends increased use of prescribed burning and thinning techniques to reduce the fire hazard in many o f these interface areas. Although this wi l l certainly protect our cities and homes, it is a treatment method which should be applied with caution in areas with minimal human settlement. Chapters 2 and 4 showed how disturbance types and frequencies have changed over the past 100 years, shifting from a high frequency, fire dominated system, to one with overall lower disturbances, except with large spikes o f mountain pine beetle activity. Introducing prescribed fires or additional harvest thinning might elevate disturbance beyond the range o f variability, especially i f catastrophic events, whether they are fire or mountain pine beetle, are not mitigated by the management activity. Further research into cumulative effects of disturbances, including restoration and harvesting, should be undertaken. 5. The age class distribution never approaches the theoretical negative exponential A key finding o f Chapter 3 was that even under a constant stand replacing fire regime for 150 years the age class distribution never followed a negative exponential. The starting condition o f the landscape prevented the landscape from ever reaching a negative exponential distribution. Based on the findings o f Chapter 4, it appears that the current condition of the landscape is a direct result of climate effects on disturbances. Therefore, even though this area is believed to 149 have a stand-replacing disturbance regime, this means many years or decades o f low disturbance, followed by periods of high disturbance frequency. The age class distribution thus reflects these spikes, of either high or low disturbance, where larger areas wil l be in younger or old classes respectively, at least on average. These spikes were evident in the results from the simulation model and had been found by other researchers studying the boreal forest. Thus, although the negative exponential is a theoretical model o f the age class distribution o f a stand-replacing disturbance dominated landscape, it is only a loose approximation. I f the age class distribution is used, as it was in Chapter 2, to identify periods with a cohesive disturbance regime, error could be incorrectly introduced without a thorough understanding o f the landscape to which it is being applied. In this case, where the disturbance regime fluctuates with climate, I posit that both periods o f high and low disturbance frequency are part of the disturbance regime. Figure 2-2 supports this theory. In this figure I had identified 3 different disturbance regimes in the past century- 1874-1953, 1954-1994, and then 1994+. In fact, this is one entire disturbance regime, given the results from Chapter 4, and the return to an elevated disturbance regime supports this, when looking at the data from such a broad scale. 6.4 C O N C L U D I N G R E M A R K S The final results o f this dissertation research present a landscape level picture o f natural disturbance dynamics in the dry cool sub-boreal spruce ecosystem. The main conclusion of this thesis is that any analysis of the historical range of variability in disturbances, especially i f it is being used to guide management policy, must include the effects of climate on disturbances. Even with the inherent uncertainties o f climate cycles and long-term climate change, we should not always assume that we can use the past to manage in the present. Frequently in undertaking research on natural or historical range of variability, we concern ourselves most with the possibility that humans have impacted the disturbance regime in our study area. This dissertation demonstrates that overall climate effects are much stronger and have a greater affect on disturbance dynamics than human activities, even in today when our current levels of forest management may be unprecedented. 150 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0074961/manifest

Comment

Related Items