Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characteristics of historical forest fires in complex mixed-conifer forests of southeastern British Columbia Cochrane, Jared Douglas 2007

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

Item Metadata

Download

Media
831-ubc_2007-0365.pdf [ 6.49MB ]
Metadata
JSON: 831-1.0100776.json
JSON-LD: 831-1.0100776-ld.json
RDF/XML (Pretty): 831-1.0100776-rdf.xml
RDF/JSON: 831-1.0100776-rdf.json
Turtle: 831-1.0100776-turtle.txt
N-Triples: 831-1.0100776-rdf-ntriples.txt
Original Record: 831-1.0100776-source.json
Full Text
831-1.0100776-fulltext.txt
Citation
831-1.0100776.ris

Full Text

CHARACTERISTICS OF HISTORICAL FOREST FIRES IN COMPLEX MIXED-CONIFER FORESTS OF SOUTHEASTERN BRITISH COLUMBIA  by JARED DOUGLAS COCHRANE B.Sc, University of Victoria, 2002  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE  in THE FACULTY OF GRADUATE STUDIES (Geography)  THE UNIVERSITY OF BRITISH COLUMBIA August 2007 © Jared Douglas Cochrane, 2007  Abstract Forest fires in southeastern British Columbia are considered the dominant natural disturbance to have shaped forest structure. In the mixed conifer montane forests of the Dry Cool Montane Spruce (MSdk) biogeoclimatic subzone, we have limited understanding of the characteristics of fire which have resulted in current forest structure. A better understanding of fire in these forests is needed to improve forest management and ecosystem restoration strategies which seek to emulate natural disturbance. To assess the historic role of fire in mixed conifer forests and to test the null hypothesis that fires do not vary spatially or temporally, this study uses cambial fire scars to analyze the historic frequency of fire in 20 stands that are statistically representative of complex, mixed-conifer forests in the southern Rocky Mountain Trench of British Columbia. Determining the date of cambial injury on a tree is an important objective of ecological research that determines the timing of disturbances such as fire, tree falls, or human modification of trees. Methods to determine scar dates require either a full stem cross-section or a partial crosssection of the wounded area. The latter method is less destructive however it is rarely used in British Columbia due to potential violations of established standard of care procedures regarding wildlife and danger trees. I developed procedures and protocols that provide a.standard of care that was reviewed by WorkSafeBC and found to meet the intent of the Occupational Health and Safety Regulations ensuring the health and safety of workers. These procedures allow large, old trees of interest to researchers to contribute to the ecosystem over the long term and prevent the creation of dangerous trees that may threaten the safety of other forest users, making a valuable contribution to future research using cambial injuries in British Columbia. Spatial variation in fire is an important driver of forest heterogeneity at the stand and landscape scale. Using cambial fie scars on trees sampled at 20 study sites, I determined that fire  frequency varied considerably between and among studied plots. I expected fires would be more frequent in plots with southern aspects than northern aspect plots. Instead, using logistic regression, I found fires to be more frequent on plots with northern aspects plots than southernaspect plots. Plot elevation, slope angle and solar radiation significantly influenced fire frequency, while plot aspect and latitude did not. Differences in season of fires were predominantly the result of differences in phenology between the two most commonly sampled tree species, Douglas-fir {Pseudostuga menzeisii var glauca (Beissn.) Franco) and western larch [Larix occidentalis Nutt.). Temporally, fire was most frequent during the time period of documented European settlement and the least frequent during the modern fire suppression period. My research results have important implications for forest management strategies which emulate natural disturbance to promote ecological resilience. Historically, fire occurred in the complex montane forests at much higher frequency than is currently classified using the Natural Disturbance Types (NDT). As a result, forest management strategies are likely preventing the occurrence of low- to moderate-severity disturbances in these forests, reducing their ecological resilience. Future management strategies should incorporate the variability observed in this study at the stand and landscape scales and return low- to moderate severity disturbances to these stands more frequently. Ecosystem restoration in these forests should be prioritized on stands that have deviated the most from the historic frequency of fires; particularly in stands which are at lower elevations and closer to valley bottom of the Rocky Mountain Trench.  Table of Contents Abstract  '.  ii  Table of Contents  iv  List of Tables  vi  List of Figures  vii  Acknowledgements  ix  Chapter 1. Introduction  1  The Study of Forest Fires  1  Defining Fire as an Agent of Disturbance  1  Describing Specific Fire Regimes  3  Mixed-Severity Fire  3  Spatial and Temporal Variation in Mixed-Severity Fire Regimes  4  Quantifying Fire Regimes  6  Mixed-Severity Fire in British Columbia  9  Research Questions and Hypothesis  11  Spatial Controls Temporal Controls Objectives and Thesis Outline  11 12 13  ;  Chapter 2. Striking a Balance: Safe Sampling of Partial Stem Cross-Sections in BC  14  Introduction  14  Methods  18  Preliminary Site Assessment Assessment of Scarred Trees and Pre-Sample Documentation Cutting of Partial Cross-Section .' Verifying Safety Conditions Signs and Communication with Other Forest Users  18 20 23 23 24  Discussion  26  Chapter 3. Fire History of Complex Mixed Conifer Stands in the MSdk of Southeastern BC  29  Introduction  29  Study Area  30  Methods  33  Research Design and Site Selection Field Sampling Determining Fire Years and Seasons Fire History Statistics Logistic Regression of Fire Intervals  33 36 38 40 42  iv  Results  •  Fire History i. . /'/'. /'//. iv. v. vi.  Plot Characteristics Fire Records Fire Season Plot-Level Fire Frequency Sub-regional Fire Frequency Regional Fire Frequency  45 45 45 45 46 51 53 56  '.  :  Influence of Site Physical Attributes on Fire Intervals  58  Discussion  60  Mixed-Severity Fire Measures of Fire Frequency Spatial Variation in Fire Frequency Temporal Variation in Fire Frequency  60 62 66 69  Conclusions  71  Chapter 4. Looking Forward: Suggestions for Sustainable Forest Management in the MSdk Subzone of Southeastern British Columbia  73  Introduction  73  Critical Fire Regime Attributes  74  Quantity on the Landscape Change in Quantity Variability of Fire and Drivers Past Severity of Fire Restoration  ;  :  Conclusions  Literature Cited  •  :  74 76 77 78 79 81  ,  83  V  List of Tables Table 1.1 Definitions of disturbance regime descriptors (Pickett and White 1985)  2  Table 3.1 Characteristics of the 20 sampled plots in southeastern British Columbia, stratified by aspect. General stand composition describes the species composition of the canopy and subcanopy tree strata. Seasonal solar radiation is the total modeled solar radiation for April through October 48 Table 3.2 Summary of fire-scarred material sampled from 20 plots in southeastern British Columbia, stratified by slope aspect. Total fire-scarred trees includes all live trees, snags, logs and stumps with external fire scars 49 Table 3.3 Summary of fire history statistics for 18 of 20 sampled plots in southeastern British Columbia, stratified by aspect. The fire recording period includes the time from first fire to death of last recording tree. Total fire scars, fire intervals, range of intervals and all fire frequency values were calculated for the full recording period 50 Table 3.4 Comparison of six models on physical attribute influence of plot-level fire intervals. Independent variables included time since fire (TSF) and five physical plot attributes. Models were constructed using a forward step-wise procedure. Maximum likelihood estimates are presented as p-values and odds-ratio (in parentheses); models were assessed using the -2log likelihood (-2log) and Akaike's Information Criterion (AIC) 60  VI  List of Figures Figure 2.1 Schematic of a tree scarred by six fires, showing a catface with a deep cavity and a firescarred partial cross-section. The horizontal dimensions (width and depth) are indicated on the section (Heyerdahl and McKay 2001) 16 Figure 2.2 Field data form for partially sectioned trees that provided the necessary documentation for each stage of the process 19 Figure 2.3 i) Two-dimensional illustration of measurements used in the completion of the field data form, ii) Three-dimensional illustration of the measurements and cuts used in the partial crosssection process 22 Figure 2.4 Warning signs for i) individual trees with a partial section removed ii) entrance points to sites with partially sectioned trees. Signs were 6" X 6" and 24" X 12" respectively 25 Figure 3.1 Study area in the southern Rocky Mountain Trench of southeastern British Columbia. Total area (shaded grey) is 1.4 million hectares : 31 Figure 3.2 Location of 20 sampled plots in southeastern British Columbia, marked with stars. The darker grey area is the MSdk subzone and all identified stands with northern and southern aspects are in blue and red, respectively 35 Figure 3.3 Variation in season of fire scars. Season was determined for 194 scars ("AH", left). Scars were stratified by plot aspect (center, left) and elevation (center right) and by tree species (right). Season was divided into three categories, earlywood, latewood and dormant, based on the position of the fire scar tip in the annual ring. Dormant season scars were located between two annual rings and were considered to have occurred in the fall of the previous year, due to favourable fall burning conditions in this region , 47 Figure 3.4 Composite fire intervals for 18 plots. The shaded box represents the 25th to 75th percentiles, whiskers are the 10th to 90th percentiles and outlying values are shown as black circles. The horizontal line in each box is the median fire interval; triangles are mean fire intervals; squares are Weibull median fire intervals; diamonds are modeled mean fire intervals 52 Figure 3.5 Fire-scar record from 1645 to 2003 for plots with southern aspects (P04 to P20) in the montane forests of the southern Rocky Mountain Trench. Horizontal lines show the time span of each plot-level fire chronology and triangles indicate a fire scar., where hollow triangles show years when one tree was scarred and solid triangles are years when >1 tree was scarred. Gray vertical dashed lines indicate fire years recorded at >10% of plots and black vertical lines when >25% of plots recorded a fire <. 54 Figure 3.6 Fire-scar record from 1501 to 1971 for all plots with northern aspects (P01 to P19) in the montane forests of the southern Rocky Mountain Trench. Horizontal lines show the time span of each plot-level fire chronology and triangles indicate a fire scar, where hollow triangles show years when one tree was scarred and solid triangles are years when >1 tree was scarred. Gray vertical dashed lines indicate fire years recorded at >10% of plots and black vertical lines when >25% of plots recorded a fire 55  VII  Figure 3.7 Frequency of fire in the montane forests of the southern Rocky Mountain Trench. Plots are grouped by aspect (top and middle) and all plots combined (bottom); n is the total number of fire years for each group. Regional fire years (^3 plots burned in the same year) occurred in 1706, 1718,1831,1847,1886,1888 and 1889. In the bottom panel, the alternating gray and white bands represent the pre-European settlement periods II (1690 - 1774) and I (1775 -1859), European settlement period (1860 -1944) and fire suppression period (1945 - 2005). For each period, the mean fire interval in years is in parentheses 57 Figure 3.8 Probability of fire through time for forests of different elevations (top), slope angles (middle), and solar radiation (bottom). Curves were calculated using logistic regression model M6 in equation 7 and four values for each attribute that represent the range of values observed in the 20 sampled plots. For each curve, the fire interval corresponding to a probability of 0.5 is the modeled fire interval. Modeled fire intervals increased with increased plot elevation, plot slope angle and modeled plot solar radiation 59 Figure 3.9 Distribution of fire intervals for each plot and calculated measures of central tendency. Mean is shown as a triangle, modeled mean as a diamond, Weibull median as a square and outliers are black circles. Modeled means were calculated using logistic regression 65 Figure 4.1 Historic range of fire intervals (gray bars) for 18 sampled plots in the montane forests of southeastern British Columbia. Crosses indicate the time since last fire at each plot 80  viii  Acknowledgements  A tremendous amount of support from individuals and organizations made this research possible and I have many amazing people to thank for their contributions. Research was funded by Tembec Inc and Canfor Forest Products Ltd under the Forest Investment Account and the Forest Science Program and additional funding came from an Ecological Integrity grant from Parks Canada. In kind support on this project was provided by the Tree Ring Lab at UBC, BC Ministry of Forests and the Protection Branch, BC Timber Sales, Tembec Inc, Canfor Forest Products Ltd and Parks Canada. In no particular order, I would like to thank Rick Fiddis, Dave Gleave, Jason Hall, Chad Smith and the Initial Attack Crews from the Invermere BCMOF Protection Branch. Your contributions to this project were tremendous. Not only did you help me build my operational knowledge of fire but you were always willing lend assistance to me in any way that you could. From Tembec, I thank George Richardson and Kari Stuart-Smith for including me on this project. From Parks Canada, thanks to Rick Kubian for your advice and knowledge, and for encouraging me, and this project, to expand. Thanks to Robert W. Gray for all your input on the project and for showing me the way of the fire-scar. Field work for this project was considerable, requiring long hours and numerous very early mornings. The quality and amount of material collected and processed would not have been possible without the tremendous effort received from Brad 'the bear' Gooderham, thank-you. In the lab, Shane McCloskey and Amanda Stan provided encouragement and laughter that made going to 'work' every day a pleasure. Trevor Jones arrived at precisely the right moment to assist in making the analysis portion of this project much stronger and he always seemed to be there with the answer! I thank whoever placed me in the same office as Cory Dobson and Carmen Wong, as this provided laughter, friendship and kept me sane. Carmen, your insight was significant and I'm proud to think that my thesis was done the 'Wong' way. I thank my committee members, Peter Marshall and Brian Klinkenberg, for providing critical assistance and feedback, which greatly improved this thesis. To my advisor, Lori Daniels, I am very grateful. As an advisor, your mentorship and guidance was invaluable and as a friend, you made this a great experience. I thank my amazing and inspiring parents, whose love and encouragement have guided me from the beginning.  ix  To my Jillian, you were an inspiration and a partner to me throughout. I cannot imagine doing this without your love, support and encouragement. The highlight my masters was marrying you, I am truly blessed.  "No doubt one of the things that the history books of the future will record is the present awakening of man to the value of all the other living things that share the world." Roderick Haig Brown (1908-1976)  Chapter 1. Introduction  The Study of Forest Fires Forest ecosystems are highly dynamic, influenced by processes at the global, regional and local scale. Until recently natural disturbances were considered the most dominant process shaping our forests, influential at all three scales, but they have been replaced by forest harvesting (World Resources Institute 2000). Thus, forest management practices are now shaping our forests. Given the past dominance of natural disturbance, current forest management practices have begun to use concepts such as natural-disturbance-based management (NDBM) (Drever et al. 2006) or the historic range of variability (Landres et al. 1999) to manage forests for ecological objectives. Understanding the role of natural disturbance in shaping forest ecosystems becomes essential, as does the successful integration of this knowledge into management practices. As forest fires are the dominant natural disturbance in most forest ecosystems in. western North America understanding their history has been the focus of much research effort.  Defining Fire as an Agent of Disturbance Pickett and White (1985) define a disturbance as any relatively discrete event in time that disrupts ecosystem, community, or population structure and changes resources, substrate availability, or the physical environment. A disturbance regime is the spatial and temporal characteristics affecting a defined landscape (Table 1.1), defined by the characteristics of multiple disturbance events through time. Conceptualizing forest fires as a disturbance is not difficult. The effects of forest fires became popular culture with the iconic story of Smokey Bear, who was orphaned when:  1  "In May, a vast fire swept through a tinder-dry forest  after five days 17,000  acres were burned, countless animals killed, grass scorched and the water producing and storage capacity of the land was badly hurt!" The True Story of Smokey Bear, USDA (1960)  Not only did Smokey's story convince North America to be careful with fire (perhaps too careful, but that's a story for another day), it also illustrated some of the characteristics of forest fires that define it as a disturbance. Smokey's fire burned at a particular time of year, to a certain size and with a visible intensity, characteristics which had a resultant effect on the forest ecosystem. Several key characteristics of fire(s) and its effects on a forest landscape over a period of time are used to describe a fire regime: spatial (size or extent, distribution), temporal (frequency, return interval, rotation period, timing), magnitude (intensity, severity) and synergism.  Table 1.1 Definitions of disturbance regime descriptors (Pickett and White 1985).  Characteristic Spatial  Descriptor | Definition Size or extent Distribution  Temporal  Magnitude Synergism  Frequency Return interval Rotation period Timing Intensity Severity  Area disturbed. Can be expressed as per event, per time period Spatial distribution, including relationship to geographic, topographic, environmental and community gradients Mean number of events per time period The inverse of frequency; mean time between disturbances Mean time needed to disturb an area equal to that of the study area The seasonality of the disturbance Physical force of the event per unit area, per unit time Impact on the organism, community, or ecosystem Effects on the occurrence of other disturbances  2  Forest structure and composition are strongly linked to the characteristics of a fire regime on a landscape (Agee 1993), where spatial and temporal variability in fire creates the vegetation and community structure in fire-dominated ecosystems. Quantifying this variability within a fire regime becomes important in determining how the forest is shaped by fire, its synergistic feedbacks and the influence of exogenous factors such as topography or climate. Ultimately, linking of fire regime descriptors to the physical attributes of a forested landscape is a primary focus of fire history research that aims to guide forest management, conservation and restoration.  Describing Specific Fire Regimes Mixed-Severity Fire A primary means of describing the impact of fire as a disturbance on the landscape is by the severity of the fire regime, where severity is reflected in the effect of fire on the dominant vegetation. In coniferous forests, high-severity fires kill a majority of the overstory trees and, lowseverity fires result in a majority of the overstory trees surviving (Arno et al. 2000). Fire severity is often inversely proportional to the frequency of fire on a landscape, where higher frequency fires are more likely to be of a lower-severity. In the mixed-conifer forests of western North America, recent research has shown that fire regimes often included a spatial and temporal combination of low- and high-severity fires (Arno et al. 2000; Baker et al. 2007; Buechling and Baker 2004; Taylor 2000). This dynamic mix of fire severities across a landscape results in a uniquely diverse forest structure and is referred to as a mixed-severity fire regime. Although a mixed-severity regime may include individual fires of low- or high-severity, the majority of the fires over time will have a range  3  of severities within a single event (Arno and Fiedler 2005). This latter type of disturbance creates an especially patchy mosaic of forest structures within stands and across the landscape.  Spatial and Temporal Variation in Mixed-Severity Fire Regimes Inherent in the definition of mixed-severity fire regimes is a dynamic combination of both spatial and temporal variation. Variation in mixed-severity regimes and their characteristics is determined by a set of physical and ecological conditions and processes that operate on a range of spatial and temporal scales (Lertzman et al. 1998). Spatially, fire regimes vary at regional and local scales. Regionally, variation in climate has been found to be an important top-down influence on fire frequency, but this is dependent upon spatial scale and geographical context. Within the Blue Mountains of Oregon and Washington, USA, Heyerdahl et al. (2001) found southern watersheds (44°N - 45° N) had significantly lower fire intervals than northern watersheds (45° N - 46°N) and this was consistent with longer and drier fire seasons, differences in vegetation and increased summer lightning strikes associated with southern latitudes. At a larger scale, Heyerdahl et al. (2007) noted that fire frequency was comparable between similar low elevation ponderosa pine-dominated forests at latitudes of -45°N, -47° N and -50°N (Heyerdahl et al. 2001; Heyerdahl et al. 2007; Wright and Agee 2004). It appears that for variation in latitude to have an influence on fire frequency there must be associated variation in climate which drives differences in vegetation characteristics and the potential for fire. Locally, fire frequency has been found to be influenced by bottom-up controls. This is most commonly due to topographical variation which influences local climate, the abundance and combustibility of fuels, and fire behaviour. Numerous studies have found site aspect to influence fire frequency (Beaty and Taylor 2001; Heyerdahl et al. 2001; Taylor 2000; Taylor and Skinner  4  1998, 2003). Sites with southern aspects typically have higher fire frequency as a result of increased solar radiation, lowered fuel moisture contents and differences in fine fuel combustibility. Elevation has also been found to influence fire frequency. At higher elevations, fuel moisture is generally greater and fires are less frequent than at lower elevations. However, the elevational range over which sampling is conducted is important, as this trend was observed over a range of 600-800m (Heyerdahl et al. 2001; Taylor 2000; Wolf and Mast 1998), but not over a range of 227m (Heyerdahl et al. 2007). The relationship between topography and fire frequency can involve complex interactions that facilitate or prevent fire spread. At the watershed scale, Heyerdahl et al. (2001) found that in steep terrain that facilitated fire spread, fire frequencies were highest on southern aspects and low .elevations. Capturing spatial variation in fire regimes is strongly dependent upon the scale over which the control operates. Previous research shows that top-down and bottom-up controls on spatial variation operate at different scales. With this in mind, fire regime studies should be carefully designed to ensure that meet the desired objectives and allow testing of hypothesized controls on fire. Temporally, fire regimes can be influenced by both variation in climate and variation in land-use. Inter-annual to decadal variations in climate have been shown to have a strong influence on the extent and severity of fires (Heyerdahl et al. 2002; Veblen et al. 2000). Heyerdahl et al. (2002) found fire extent to be suppressed during La Nina years in the Pacific Northwest, while Veblen et al. (2000) found fire occurrence to increase with La Nina events in the Colorado region. These results are consistent with geographic climate variation associated with El Nino-Southern Oscillation (ENSO) across North America. Significant changes in fire frequency have also been linked to changes in land use associated with European settlement of North America and increased  5  fire suppression success with the advent of aerial suppression (Brown et al. 1999; Fule et al. 1997; Taylor 2000; Veblen et al. 2000).  Quantifying Fire Regimes Research techniques to quantify fire frequency for either low- or high-severity regimes are well documented (Agee 2004). Commonly, frequency is calculated as the mean time between disturbances as a fire return interval, an inverse of frequency. High-severity fires typically kill the majority of mature trees in a stand resulting in distinct stand cohort boundaries across the forested landscape that can be visually identified in aerial photographs or on the ground and then dated using dendrochronology. Time-since-fire distributions are then produced for the landscape, from which the fire interval can be extrapolated (Johnson and Gutsell 1994; Reed 1994; Van Wagner 1978). Cohort boundaries can provide a reasonable indication of fire extent however the size of older fires may be obscured by the occurrence of more recent, overlapping fires. The season of a particular fire may be determined using fire scars found on trees located at the perimeter of a fire or by using recent fire boundary maps (Beaty and Taylor 2001). Although effective for high-severity fires, this research approach has limited ability to capture low- to moderate-severity fires that do not leave distinct cohort boundaries on the landscape. And as a consequence fire intervals could be overestimated. Low-severity fires typically result in cambial scars on mature trees that survive fire. Fire scars may be cross-dated using dendrochronological techniques and analyzed to reconstruct the fire return interval for the scarred tree and the stand (Arno and Sneck 1977; Daniels and Watson 2003; Dieterich and Swetnam 1984). The season of a particular fire may be discernable by analyzing the position of the fire scar tip within the seasonal growth ring (Baisan and Swetnam 1990; Dieterich and Swetnam 1984). Within a stand or landscape, the size or extent of a fire can  6  be estimated by determining the spatial distribution of sampled trees that had fire scars with the same date caused by an.individual fire (Agee 1993). The mixture of low- and high-severity fires found in a mixed-severity regime requires a combination of sampling techniques to reconstruct the low- and high-severity components of fire history. In recent mixed-severity studies, the type of information used to reconstruct fire regimes varied among plots depending on the objectives of the particular study and the specific forest type (Buechling and Baker 2004; Heyerdahl et al. 2001). Within a landscape, distinct, single cohorts of trees indicate patches that were burned by high-severity fires, while multi-cohort, multi-aged patches may indicate that multiple fires burned at a severity below the threshold required to kill all the trees. A combination of cambial fire scars, stand age structures, tree growth histories and spatial analyses may be used to reconstruct fire for the full range of severities in a mixed-severity regime.  Recent Advances in Fire History Research Traditionally, in fire history research, the goal of sampling was to obtain long and complete records of fires (Baker and Ehle 2001). This focus on the temporal component of fire regimes resulted in targeting, in which stands with high fire-scar densities were specifically chosen for research. While this method does produce long fire histories, it may result in a biased estimate of fire return interval if the targeted study sites are not representative of the entire landscape. Targeted sampling is an example of mensurative pseudoreplication, where the physical sampling space is more restricted than the inference space. Specifically, for targeted fire history studies, the calculated fire interval may be lower than actual mean fire interval of the landscape, due to the emphasis on areas with a high frequency of low-severity surface fires and long periods without  7  high-severity fires. Pseudoreplication in fire history studies can be avoided by clearly defining the sample population and determining objective methods for the placement of sample plots prior to sampling. A second potential source of bias is the selection of fire-scarred trees for sampling. Within a single stand, older trees and trees with multiple fire scars are often preferentially selected for sampling as another means of obtaining long and complete records of fire. Traditional justification for this method was that trees are a natural archive of historical data and sampling multiple-scarred trees creates the most complete fire history with the highest sampling efficiency (Arno and Sneck 1977; Brown et al. 2000). VanHorne and Fule (2006) assessed the impact of different methods of sample tree selection by comparing stand-level fire intervals calculated from preferentially selected trees versus all fire-scarred trees within a stand. They found that targeted sampling resulted in a fire interval that was within one-year of the census fire interval for their study site. This result suggests that older trees and trees with multiple scars can be preferentially sampled to reduce the number of sampled trees at each study site without biasing estimates of fire frequency. Sampling intensity is a third potential source of bias when quantifying fire regimes. Sampling intensity, defined here as the area sampled for fire scars, influences the calculated fire interval and affects the comparison of intervals between stands (Baker and Ehle 2001, VanHorne and Fule 2006). When sampling fire scars, the size of the fire that caused the scar is not taken into account so scars located within a sample plot may result from small fires that burned entirely within the plot or they may be from larger fires that burned only part of the plot. Therefore, a larger area sampled increases the chance of encountering additional fires. With each additional fire date in the record, the stand-level fire interval is reduced. Large study sites are more likely to have more and shorter fire intervals than small study sites, which is a significant potential source of bias. To avoid  8  bias and allow direct comparison among sites when sampling multiple stands, the area sampled should remain consistent throughout the study.  Mixed-Severity Fire in British Columbia Although mixed-severity fire regimes are a potential disturbance process within the mixed conifer forests of British Columbia, there has been little research effort and few publications on this topic (Heyerdahl et al. 2007; Wong et al. 2003). Instead, the majority of fire research in B.C. has focused on high- or low-severity fire regimes in forest types that occupy approximately 50% of the provinces forested area. While mixed-conifer forests occupy approximately 15% of B.C.'s forested area, they contain a high diversity of flora and fauna that are likely well adapted to the historic natural fire regime (Bunnell 1995). Determining how fire regime characteristics vary is essential for identifying the role of fire in the long-term dynamics of mixed conifer ecosystems. Furthermore, understanding and quantifying these fire regimes is important for scientifically-based and certifiable sustainable forest management practices in such forest types. Southeastern British Columbia (BC) contains a high-percentage of mixed-conifer montane forests within the Dry Cool Montane Spruce (MSdk) subzone, as classified by the BC biogeoclimatic ecosystem classification (BEC) system. The BEC system is a hierarchical classification scheme that uses regional, local and chronological integration with climatic, vegetation and site classification to describe the ecosystems of B.C. (Pojar et al. 1987). In the Nelson Forest Region of southeastern B.C., the MSdk subzone occupies the mid slopes of the southern Rocky Mountain Trench and the valley bottoms of the neighbouring drainages. Latesuccessional stands on zonal sites are dominated by hybrid white spruce {Picea glauca Parry x engelmannii) and subalpine fir {Abies lasiocarpa (Hook) Nutt.). Stands on other sites are dominated by Douglas-fir (Pseudostuga menzeisii var glauca (Beissn.) Franco) and western larch  9  {Larix occidentalis Nutt.) Large, early-successional, even-aged stands of lodgepole pine {Pinus contorta var latifolia Engel.) result from high-severity fires (Braumandl 1992). Spatial extent and distribution of serai stages and the specific stand attributes are important determinants in the composition and structure of vertebrate faunal communities (Bunnell 1995). As such, understanding and quantifying fire regimes and the effects these disturbance have on vegetation patterns becomes important for understanding and conserving biodiversity. Forest management agencies in BC use Natural Disturbance Type (NDT) to classify and describe the disturbance regimes of BEC subzones. The MSdk subzone is currently classified as NDT 3, indicating a fire-dominated disturbance regime with return intervals of 150-200 years for stand replacing fires (BCMOF 1995). Research used to classify the fire regime of the MSdk quantified fire for the even-aged lodgepole pine stands using techniques for high-severity fire. This classification is now applied to all stands within the MSdk, despite the fact that approximately 25% of the MSdk has obviously different structural characteristics, such as multiple cohorts of Douglas-fir and western larch. The multi-cohort structure of these stands suggests that they are misclassified and that research is needed on these stands to determine the characteristics of their fire regime. My proposed research will investigate the controls driving spatial and temporal variability in the fire regimes of mixed conifer forests in southeastern British Columbia, within the Rocky Mountain Trench. To test for these controls I utilized cambial fire scars to reconstruct fire history and potential changes in fire frequency. To avoid pseudoreplication and bias in my results, I used a stratified-random sampling approach. My sample population was clearly defined prior to sampling and sample plots were consistent in size and objectively placed within each study stand.  10  Research Questions and Hypothesis Spatial Controls Topographic variation has been shown to be an important control on spatial variation in fire frequency for mixed conifer forests, when controlling for species composition, elevation and soil type (Beaty and Taylor 2001; Heyerdahl et al. 2001; Heyerdahl et al. 2007; Taylor 2000; Taylor and Skinner 1998, 2003). At the latitude of my study area (~50 N) the warmest slope aspect is e  southwestern and the coolest aspect is northeastern. I expect the greatest difference in fire regimes to occur between slopes with the warmest and coolest aspects. However, given the topographic complexity of my study area, other sources of topographic variation may significantly influence fire frequency. To examine spatial controls on fire regimes, I addressed the following research questions and hypotheses: 1.  Is there a significant difference in fire return intervals between sites with northeastern and southwestern slope aspects? Ho: There will be no difference in the fire return intervals for sample sites with a northeastern aspect when compared to the intervals for southwestern aspects. HA1: Fire intervals will be longer on sites with northeastern aspects than the southwestern sites. The cooler aspect will result in a higher fuel moisture contents which will decrease the likelihood of fire ignition and spread. HA2: Fire intervals will be shorter on sites with northeastern aspects than the southwestern sites. The reduced occurrence of fire on the cooler aspect sites allows for an accumulation of surface fuel, which may result in higher intensity surface fires and increase the likelihood of scarring on trees.  11  2.  Do topographic factors other than aspect significantly affect fire frequency? Ho: Elevation, slope angle, solar radiation, latitude and aspect have no significant influence on fire frequency for individual plots. HA1: Increased elevation will result in lowered fire frequency as the fire season is shorter and higher elevation and higher elevation stands are farther away from the high-frequency fire regimes found at valley bottom. Fire frequency will be higher with increased slope angle and solar radiation as these factors are positive influences on fire behaviour. Fire frequency will be lower at higher latitudes, associated with a shorter fire season and lower temperatures. Plots with a south-western aspect will have higher fire frequency, as hypothesized in question 1.  Temporal Controls Significant changes in fire frequency have been associated with changes in land use since European settlement of North America and increased fire suppression since the advent of aerial suppression (Brown et al. 1999; Taylor 2000; Veblen et al. 2000). The popular perception that decades of fire suppression has promoted unnatural fuel accumulation and subsequently large, unprecedented and severe wildfires was developed primarily from low-elevation ponderosa pine forests. In contrast, montane and subalpine forests may not yet have deviated greatly from their historical range of fire intervals (Schoennagel et al. 2004). In this study, I examine the temporal variation in the fire regime between 1690 and 2005. 3) Has fire frequency changed in response to documented changes in land-use? Ho: There will be no significant change in fire intervals or percent of scarred trees over the period of study.  12  HA1 : Variation in fire intervals and percent of scarred trees will correspond with time periods of historical changes in land use: i.  Fire Suppression  1945-2005  ii.  European Settlement  1860-1944  iii.  Pre-European Settlement I  1775-1859  iv. Pre-European Settlement II  1690-1774  Objectives and Thesis Outline In this study, I have used a landscape-scale approach to address a lack of scientific knowledge on forest fires in the mixed conifer forests of southeastern British Columbia. I aim to describe fire regime characteristics for stands with complex forest structure in the Dry Cool Montane Spruce zone (MSdk) and to provide forest managers with a stronger understanding of the role of fire in these stands. Chapter 2 presents a new sampling protocol that I developed in the course of this research for extracting partial stem cross-sections with cambial fire scars. This method strikes a critical balance between the need for sound, wood samples with fire scars for research, the desire for non-destructive sampling methods to conserve large, old wildlife trees, and the need to ensure safe access for future forest users. Chapter 3 presents my analysis of fire history at 20 stands that are representative of the landscape. It quantifies and compares fire regime characteristics to identify significant controls of spatial and temporal variation in the historic fire regime of my study area. Key research findings are summarized and the implications of my research for sustainable forest management, conservation and restoration are presented in Chapter 4.  13  Chapter 2. Striking a Balance: Safe Sampling of Partial Stem Cross-Sections in British Columbia  Introduction  Researchers in a wide array of disciplines use the date of cambial injuries on a tree to explain dynamic processes, such as fire history (Heyerdahl et al. 2001; Taylor and Skinner 2003; Van Home and Fule 2006), stand development (Delong et al. 2005) and the interactions between climate and disturbance regimes (Daniels in review; Swetnam and Betancourt 1990; Veblen et al. 2000) . Injury to the cambium of a tree will result in the formation of woundwood as enhanced cell division occurs at the margins of the injury causing wound closure as the tree works to restore the continuity of its vascular cambium around the circumference of the stem (Smith and Sutherland 2001) . This process causes a distinct morphology that is termed a catface scar. In particular, lowto moderate-severity fires can cause fire scars on the leeward side and lower bole of a tree (Gutsell and Johnson 1996). When properly applied, dendrochronological techniques can be used to analyze cambial fire scars and determine the exact year of fire occurrence and in many cases, the season in which the fire burned (Dieterich and Swetnam 1984). Traditionally in BC the analysis of fire scars is conducted on complete cross-sections from trees, a method which requires that the tree be felled. In well replicated studies, where >10 trees are sampled per site and multiple sites are sampled, the impacts of this destructive sampling can be significant at stand and landscape scales.  A version of this chapter has been submitted for publication. Cochrane, J.D. and Daniels, L.D. Striking a balance: safe sampling of partial stem cross-sections in British Columbia. BC Journal of Ecosystems and Management.  Often, fire history researchers aim to obtain the longest and most inclusive record of fires possible and, thus, the oldest fire-scarred trees in the stand are of the greatest interest. However, these large, old trees can be rare at stand-to-landscape scales and contribute significantly to structural diversity, habitat availability and long-term ecological functioning of the forest (Hansen et al. 1991). For example, the open catface scar on the bole of the tree may provide important foraging and/or nesting habitat and as these large trees senesce and decay they will influence the habitat and resource availability of the surrounding ecosystem (Franklin et al. 1987).  The conservation of large, old, scarred trees conflicts with research objectives to collect scientifically informative and sound samples. In light of.this conflict; numerous studies outside of BC have used partial stem cross-sections as an alternate sampling method (Brown et al. 2000; Heyerdahl and McKay 2001; McBride and Laven 1976; Van Home and Fule 2006). Rather than sampling the entire cross-sectional area of the tree, only a portion of the stem containing fire scars is removed (Figure 2.1). Sampling is then limited to a single lobe of fire scars on a restricted portion of the tree, allowing the sampled tree to survive and contribute to the ecosystem.  Use of the partial-cross section method has been limited in BC due to potential violations in the standard of care regarding the management of wildlife and danger trees. Specifically, the WorkSafeBC occupational health and safety regulation 26.11 (1) states that: "if work in a forestry operation will expose a worker to a dangerous tree, the tree must be removed." To differentiate danger trees from wildlife trees and assist in the management of wildlife trees, a multi-agency Wildlife Tree Committee (WTC) was created in 1985, with the mandate to promote the conservation of native wildlife trees and associated stand-level biodiversity in a safe and operationally efficient manner in forest and park environments (Wildlife Tree Committee 2005). The  15  WTC created guidelines and procedures that are used to determine whether a tree is dangerous to workers under various levels of disturbance and the steps and safety procedures for mitigating hazard (Manning et al. 2002).  Figure 2.1 Schematic of a tree scarred by six fires, showing a catface with a deep cavity and a firescarred partial cross-section. The horizontal dimensions (width and depth) are indicated on the section (Heyerdahl and McKay 2001).  A dangerous tree is one that is hazardous to people or facilities because of location or lean, physical damage, overhead hazards, deterioration of limbs, stem or root system or a combination of these (Wildlife Tree Committee 2005). Determining if a tree is dangerous involves a five step process, which is conducted by all forest workers prior to commencing their operations to ensure that the work area is safe:  16  Step 1: Determine level of ground disturbance or tree disturbance and type of work activity. Step 2: Conduct a site assessment overview Step 3: Conduct a tree assessment. Step 4: Make the appropriate safety/management decision. Step 5: Provide documentation and communicate safe work procedures.  In order to prevent the creation of dangerous trees by partial-sectioning we modified the above process using information provided by the WTC to create our own procedures and five-step process that applies directly to partial cross-section sampling. These procedures were reviewed by WorkSafeBC and found to meet the intent of Occupational Health and Safety Regulations to ensure the health and safety of workers. Our process was designed to meet the following safety criteria:  •  stem damage is < 25% of the trees cross-sectional area,  •  circumference removed is < 25% of the total circumference, and  •  shell thickness of sound wood is > 30% of the radius  The purpose of this chapter is: (1) outline procedures for the removal of partial crosssections that adhere to the established safety criteria of the Wildlife Tree Committee (WTC) and (2) provide a research method that is sound and less destructive than using only full stem crosssections. These procedures provide a standard of care that allows large, old trees of interest to researchers to contribute to the ecosystem over the long term and prevents creation of dangerous trees that may threaten the safety of other forest users.  17  Methods Using the WTC guidelines a five-step process was established that guides a researcher when extracting partial cross-sections in British Columbia. To meet provincial regulations, these steps must be conducted by a certified danger tree assessor, in the silviculture and forest harvesting module, and all chainsaw work must be done by falter certified by WorkSafeBC. A field data form to be completed during the five-step process for each tree partially sectioned tree was also created (Figure 2.2). Both the five-step process and documentation are explained in detail below.  Preliminary Site Assessment The preliminary site assessment uses the wildlife/danger tree assessment procedures to ensure the work area around a scarred tree is safe and that sampling may proceed. This step assesses the site as a whole, including the trees that are not directly affected by partial sectioning. This work area is defined as all trees within 1.5 tree lengths of active falling.  18  Danger Tree Assessment for Partially Sectioned Trees 1. General Information Date:  U T M Coordinates:  '  Plot #: Tree #.  Tree Class: Wildlife Value:  Species:  Tree Height:  , .  2. Defects, state S (safe) or D (dangerous) Hazardous Top Dead Limbs '  Witches Broom  D  Sloughing Bark  D  D  Butt and Stem Cankers...  D  D  Fungal Fruiting Bodies  Split Trunk  • D  Stem D amage  0  Tree Lean  — t1  Visual Root Inspection  D  3. Partial Section Information a.  Tree Diameter Minus Bark:  b  Total Cross-Sectional A r e a  c.  20% of Total Cross-Sectional Area:  d.  Cross-Sectional AreaRemoved:  e  Percentage Removed DfTntal:  f  Total Circumference.  g  Percent Circumference Remaining  h.  Required Shell Thickness:  i.  Average Stemwood Shell Thickness:  Flagged  •  Photographed  D  Sketch:  •  Figure 2.2 Field data form for partially sectioned trees that provided the necessary documentation for each stage of the process.  In determining the level of ground disturbance, it's important to remember that potential danger increases with the level of disturbance. Few activities that cause high levels of ground disturbance are appropriate around potentially dangerous trees or where exposure to people is constant or of long duration (Wildlife Tree Committee 2005). In conducting fire history research, the greatest level of disturbance occurs when a tree is cut; therefore the assessment was based on the impacts associated with tree falling. In addition, other activities, such as timber cruising, logging or fire fighting, that may occur in the stand and the level of potential exposure associated with each activity were noted.  19  In the site assessment overview, factors such as stand history and condition, potential for flooding and windthrow, and general forest health indicate the overall characteristics of the stand. Information on site and stand factors provides useful clues as to the condition and potential danger of individual trees (Wildlife Tree Committee 2005). In particular, hazard indicators such as, root/stem disease, evidence of significant windthrow or stems with a height/diameter ratio >100, are of importance, as these indicate a greater potential for stem failure.  Standard tree assessments are conducted for all trees within the work area that are not being considered for partial sectioning. Trees that are deemed dangerous to the active falling area must either be removed or have a no-work-zone (no-work-zones are generally 1.5 times the length of the hazard) placed around them prior to beginning the assessment of potential sample trees, ensuring the work area is safe.  Assessment of Scarred Trees and Pre-Sample Documentation A preliminary assessment of a potential sample tree is necessary to ensure it remains safe after the partial section is removed. Trees are assessed for physical indicators of danger (Section 2 in Figure 2.2). If any dangerous defects are found, the tree is no longer a candidate for partial sectioning and the tree must be felled in order for it to be sampled.  For potential sample trees that did not have dangerous defects, a series of measurements are taken prior to sampling to ensure the approved safety criteria for partial sections were achieved. Tree diameter at sample height is measured and converted to diameter without bark by measuring depth of the thickest portion of bark on opposite sides of the tree and subtracting both measurements from the diameter (Figure 2.2, Part 3a). Diameter without bark represents the critical support structure of the tree and is used to calculate the total cross-sectional area of the  20  tree (Figure 2.2, Part 3b). The aim in this study was to sample < 20% of the cross-sectional total area (Figure 2.2, Part 3c), which provided a 5% margin of error when determining the dimensions of the partial section.  To determine the best position and orientation of the partial section, the physical characteristics of the scar lobe(s) were examined and the best orientation of the horizontal width and depth of the sample was selected to optimize the number of fire scars (Figures 2.1 and 2.3). Only trees with visible scar lobe(s) were sampled, which allowed the width of the section needed to capture all scars to be determined. Using this width, the appropriate depth to ensure the sample was < 20% of the cross sectional area of the tree was calculated and the boundaries of the width and depth of the sample was marked on the trunk of the tree using faller chalk. The certified faller could clearly see where the cuts needed to be to remove the partial cross-section. If all visible scars in a partial cross-section that was < 20% of the total cross-sectional area of the tree could not be captured, then the tree was felled for sampling or another tree was selected.  21  0  (f) Total circumference (dashed line) A " O ~* ~ ~~ !>  (b) Total cross-sectional area (space within dashed line) (h) Required shell thickness (15% of Diameter) (a) Diameter minus bark Horizontal depth Hortzonta width *  mm  ii)  Cross-sectional area removed < 25% of the total area at sample height •Horizontal depth  j Vertical cut  Vertical thickness jr_ {avg.6cm}  Figure 2.3 i) Two-dimensional illustration of measurements used in the completion of the field data form, ii) Three-dimensional illustration of the measurements and cuts used in the partial crosssection process.  22  Cutting of Partial Cross-Section To safely remove a partial section requires four cuts into the tree using a chainsaw, two vertical cuts and two horizontal cuts (Figure 2.3). The vertical cuts are done first by boring the chainsaw into the tree, where the tip of the saw enters the tree first and does the cutting. Boring is i  a dangerous and difficult procedure as there is a greater chance of chainsaw kickback and increased vibration. The vertical cuts are done first to reduce the chance that the sample will break due to vibration. The horizontal cuts are positioned such that the vertical thickness (Figure 2.3ii) of the section averages 6 cm. Care must be taken to prevent the tip of the chainsaw bar from cutting beyond the vertical cuts, which would weaken the integrity of the tree after sampling. Measuring the desired cut dimensions on the chainsaw bar prior to cutting is recommended, which provides a helpful visual cue for the faller to ensure the partial section is not too large.  Verifying Safety Conditions After the partial section is removed, measurements must be taken to ensure the sampled tree meets all three approved safety criteria. First, the actual horizontal depth and width of the partial cross-section is measured and its cross-sectional area (width x depth) is calculated (Figure 2.2, Part 3d). This is converted to a percentage of total cross-sectional area removed from the tree at sample height (Figure 2.2, Part 3e) to ensure the sample does not exceed 25% of the total area of the tree.  To ensure < 25% of the total circumference was removed, the circumference of the tree (Figure 2.2, Part 3f and Figure 2.3) and the portion of the circumference removed at sample height (Figure 3) are measured. The percent circumference removed and the percent remaining are calculated (Figure 2.2, Part 3g).  23  Ensuring the thickness of sound wood forming the shell of the tree is >30% of the radius (Figure 2.2, Part 3h), requires assessing the stemwood of the partial section and two increment cores. The average stemwood shell thickness (Figure 2.2, Part 3i) is determined by measuring the amount of solid wood visible in (a) the partial cross-section and (b) increment cores taken on both sides of the stem at equal distances from and at the same height as the partial-section. The percent shell thickness is calculated as the average measured thickness divided by half the diameter or radius of the tree measured inside the bark (Figure 2.3, Part 3a).  Partially-sectioned trees that failed any of these three criteria were considered unsafe and felled. For trees that were less than 4.0m in height, the hazard is believed to be low enough that a clearly flagged no work zone around the tree and danger tree flagging on the tree would mitigate the hazard.  Signs and Communication with Other Forest Users To ensure the safety of other forest users, signs are installed at research sites and on partially sectioned trees and the location of all modified trees is communicated to management agencies. In this study, to clearly indicate that a tree had been partially sectioned, a 6" X 6" sign was posted on the stem of tree (Figure 2.4i). The tree was flagged with 'Danger Tree' ribbon and spray painted above and below the partial cross-section. In addition 24" X 12" signs were posted at an obvious entrance point (e.g. near roads or landings) of all stands in which trees were modified (Figure 2.4ii).  24  Wildlife Tree  WARNING T r e e S t r u c t u r e Has B e e n M o d i f i e d  Do Not Fall Tree Limit Exposure to Tree  r  WARNING  This Area is Part of a Research Study on the Historic Role of Fire in Our Forests  Trees May Have Been S t r u c t u r a l l y M o d i f i e d t o retain high v a l u e wildlife trees. .  Enter Site At Own Risk Please Do Not Disturb Active Study Area This project is in partnership with the University of British Columbia, Tembec Industries Inc., Canfor Forest Products Ltd., B.C Ministry of Forests and Range, Parks Canada and B.C. Timber Sales.  TREE R|NG LAB  Partes Canada  Paras Cmntfc]  9CTtab#t''3rLlj|li#  Figure 2.4 Warning signs for i) individual trees with a partial section removed ii) entrance points to sites with partially sectioned trees. Signs were 6" X 6" and 24" X 12" respectively.  25  The exact location (UTM coordinates) and a description of each tree should be recorded and added to maps of the area using GIS software. This information should be reported to all forest licensees working in the area (e.g., Tembec, Canfor and BC Timber Sales in this study area) and the B.C. Ministry of Forests and Range. These agencies will record the location of the partiallysectioned trees on maps of the area and inform forest workers of their existence.  Discussion In the summer of 2006 20 plots were sampled in southeastern British Columbia for fire history. Within these plots, we found that -40% of the samples were from downed logs or stumps that required no falling. Of the standing trees, -10% were safely sampled using partial crosssections. The morphology of the catface scar was the primary indicator of whether a useful partial section could be removed safely from a tree. All of the trees from which a partial section was successfully removed had very little decay associated with the catface and a relatively large gap between the outer scar lobes. The calculated cross-sectional area of trees with a decayed catface (Figurel) should reflect the loss due to decay and, as a result, the partial section will need to be smaller. The distance between scar lobes was critical because trees without open scars could not be partially sectioned effectively. From a safety perspective, the amount of cross-sectional area that can be removed is limited to < 25%. For the partial samples to be useful for fire history research, the area that is removed must contain the tips of the fire scars. Trees with open catface scars will show the location of scar tips so that the researcher is able to determine if a safe partial section can be removed and the needed fire information is collected.  26  All of the methods used in these procedures were conservative. In particular, the method used to calculate the cross-sectional area removed from the tree overestimates the true area removed, as the largest horizontal width and depth was used in the calculation. The reason for using these measurements is two-fold: 1) it errs on the side of caution and 2) it is the method used in previously published literature on the mortality rates associated with partial cross sections (Heyerdahl and McKay 2001).  An important component of these procedures is the flow of information between researchers and other forest users. By documenting the location of partially-sectioned trees, future users of the forest, such as wildland fire fighters or forest industry workers, can be made aware of their existence. For this research, spatial data layers were created that could be inserted directly into GIS mapping applications to facilitate communication with our collaborators and the BC Ministry of Forests and Range.  This study is part o f an ongoing research project documenting the viability o f partial cross-sections for research o f forest dynamics. A s additional trees are sampled, they w i l l be added to the databank. Partially-sectioned trees w i l l be monitored through time to determine rates o f failure and mortality relative to other living trees and snags. Long-term research o f partially sectioned ponderosa pine i n Oregon  found mortality  rates to be <10% and failure rates to be nominal (Heyerdahl and M c K a y 2001). Extrapolating these results to the Douglas-fir and western larch that were sampled may not be appropriate due to potential differences in the physiological response to wounding, however all the trees sampled had significant resin secretions similar to those o f ponderosa pine. In addition, Heyerdahl (2006, personal communication) revealed that  27  none o f their 138 partially-sectioned trees had broken at sample height after 11 years. Over time, the impacts o f partial sections on the tree survival and the composition and structure  o f the  surrounding  stand  will  be  assessed  for  trees  in this  study.  28  Chapter 3.  Fire H i s t o r y of C o m p l e x M i x e d  Conifer  Stands in the  MSdk  of  Southeastern B r i t i s h C o l u m b i a  Introduction Knowledge of the natural disturbances likely to affect forested landscapes is essential to any forest management plan, whether the objective is timber production, wildlife conservation or wilderness management (Agee 1993). Fire history research examines fire processes responsible for stand structure in forested landscapes through the study of fire regime parameters: fire size, timing and frequency, and severity. Historically, research has focused on fire as either a highseverity (stand-destroying) or low-severity (stand-maintaining) disturbance within a specific landscape. However, recent research has shown that in the mixed-conifer forests of western North America fire regimes often include a spatial and temporal combination of low- and high-severity fires, referred to as a mixed-severity fire regime (Agee 1993; Arno et al. 2000; Baker et al. 2007). The spatial and temporal variability of a mixed-severity fire regime is a vital attribute of a forest ecosystem and the resultant patterns can be used to guide forest management (Landres et al. 1999). Although mixed-severity fire regimes have been shown to be a key disturbance process in many mixed conifer forests, there has been little research within this forest type in British Columbia. Many montane forests of southeastern B.C. have highly complex stand structures, consisting of multiple age cohorts of numerous conifer species, notably interior Douglas-fir {Pseudostuga menzeisii var glauca (Beissn.) Franco), western larch {Larix occidentalis Nutt.) and lodgepole pine (Pinus contorta var latifolia EngeL). Occupying mid-elevations in the Rocky Mountain Trench and the surrounding area, these complex stands often are surrounded by even-  29  aged stands dominated by lodgepole pine. I hypothesize that these forests result from a mixedseverity fire regime. However, studies conducted in these forests have either focused on highseverity fire regimes in subalpine forests or on low-severity regimes near valley bottoms. In general, these fire regime studies were limited in their sampling efforts and scale and very few have been published (Wong et al. 2003). The objective of this chapter is to quantify fire frequency and increase the understanding of spatial and temporal variation in the fire regimes of the complex, mixed-conifer stands of southeastern British Columbia. Fire frequency is examined at multiple scales to address the following research questions: (1) Is fire frequency similar in mixed-conifer stands with similar stand structure? (2) Spatially, does fire frequency vary between slopes with contrasting warm or cool aspects? Do topographic factors other than aspect affect fire frequency? and (3) Temporally, has fire frequency changed in response to documented changes in land-use? To address these questions, I quantified and compared fire frequency at the plot, sub-regional and regional scales using fire scars collected from 20 randomly selected plots from across the landscape.  Study Area The study area represents 1.4 million hectares (ha), located in southeastern British Columbia, bounded by the town of Golden to the north and the Canada-United States border to the south (Figure 3.1). This area includes the Invermere and Cranbrook Timber Supply Areas (TSA), which are bordered by the Purcell Mountains to the west and the Rocky Mountains to the east. It is dominated by the north-south oriented southern Rocky Mountain Trench, which is divided equally into the north-flowing Columbia River and the south-flowing Kootenay River. Highway 93/95 runs the length of the southern Rocky Mountain Trench from Cranbrook to Golden.  30  Highways Overall Study Area 1.4 Million Hectares  Figure 3.1 Study area in the southern Rocky Mountain Trench of southeastern British Columbia. Total area (shaded grey) is 1.4 million hectares.  In the southern Rocky Mountain Trench and tributary valleys that comprise the study area, glacial drift is widespread along valley bottoms and gentle lower slopes, but steeper slopes consist of rocky outcrops and colluvium derived from hillslope geomorphic processes (Meidinger and Pojar 1991). Surficial materials are discontinuous and bedrock, commonly including calcareous shale, limestone, dolomite, siltstone, quartzite and argillite, is exposed in many places (Braumandl 1992). Dominant soils are eutirc brunisols or gray luvisols (Valentine et al. 1978). Climate of the study area is classified as Cordilleran, with the Rocky Mountain Trench having a semi-arid climate as the Pacific air-stream must descend to enter the trench (Hare and  31  Thomas 1979). Local climate is strongly influenced by topography as elevation in the trench ranges from 650m at valley bottom to 3400m at ridge top. Winters at valley bottom are cold with a mean January temperature of -6.5°C, while summers are warm with a mean July temperature of 17.9°C. Yearly mean precipitation is 439mm with 77% of this falling as rain and 60% of this falling between May and August (Wasa Lake station 49° 49' N 115° 37' W at 930 m.a.s.L, 1971-2000, Environment Canada 2006). The mixed-conifer montane forests in this study are located in the Dry Cool Montane Spruce (MSdk) subzone, as classified by the biogeclimatic ecosystem classification (BEC) system. The BEC system is a hierarchical classification scheme that uses regional, local and chronological integration with climatic, vegetation and site classification to describe the ecosystems of BC. (Pojar et al. 1987). The MSdk subzone occupies the mid slopes of the southern Rocky Mountain Trench and the valley bottoms of the neighbouring drainages. Late-successional stands on zonal sites are dominated by hybrid white spruce (Picea glauca Parry x engelmannii) and subalpine fir {Abies lasiocarpa (Hook) Nutt.). Stands on other sites are dominated by Douglas-fir and western larch. Large, early-successional, even-aged stands of lodgepole pine result from high-severity fires (Braumandl 1992). Forest management agencies in BC use Natural Disturbance Type (NDT) to classify and describe the disturbance regimes of BEC subzones. The MSdk subzone is currently classified as NDT 3, indicating a fire-dominated disturbance regime with return intervals of 150-200 years for stand replacing fires (BCMOF 1995). The study area was first occupied by the Ktunaxa Nation 4,000 - 10,000 years ago. It forms a portion of their traditional territory, which now includes the Shuswap and Akisq'nuk bands. Fire was an integral part of Ktunaxa use of the landscape (Lucas 2007), although the timing and specific purposes for fire in this area are not known with certainty. In the Ktunaxa territory to the  32  south, researchers have documented oral history on the use of broadcast burning to maintain open stands, improve hunting, clear land, improve grazing and communication, and direct links were found between short fire intervals and high levels of First Nation land-use prior to 1860 (Barrett and Arno 1982). European settlement began in the study area in the 1820's when a small number of people settled on the banks of the Columbia and Windermere Lakes (Society 1972). The European population began to grow in the early 1860's when gold was discovered near present day Fort Steele and Invermere (Bancroft 1890; White 1988). These discoveries sustained miners for less than a decade and the European population of the southeast Kootenays decreased to only 11 settlers in 1865. After new discoveries of other mineral deposits near Kimberley and the completion of the southern railway, the European population increased in the early 1880's (White 1988). The town of Golden at the northern extent of the study area did not possess the mineral deposits for a mining, boom and its European population began to grow in the 1880's as a result of the steamboat era on the Columbia River associated with the operation of the Canadian Pacific Railway (Society 1972).  Methods Research Design and Site Selection In this research, I utilized a stratified-random approach to sample the mixed-conifer stands with multi-cohort structure in the Dry-Cool Montane Spruce (MSdk) subzone (Pojar et al. 1987). Sample sites and plots within sites were placed objectively using GIS analyses, allowing my results to be extrapolated to the entire sampled population.  33  Throughout the province, the BC Ministry of Forests and Range has classified vegetation cover using aerial photographs and a classification scheme that identifies homogenous forest stands. These stands have been delineated, classified by features such as species composition, age class and height class, and the boundaries digitized. The resulting forest cover maps provide a database of "polygons" (discrete patches of forest) that can be searched for compositional and structural attributes of interest. In this study, I searched all forest cover polygons in the Invermere TSA and a portion of the Cranbrook TSA (grey area in Figure 3.1) for the following attributes in order to identify all potential sample polygons:  i. Dry Cool Montane Spruce (MSdk) biogeoclimatic subzone; ii. Documented presence of at least two separate cohorts; or iii. Oldest tree cohort established before 1860; and iv. Not logged between 1950 - 1999.  The entire study area contained 83,535 polygons and 5812 of these met the criteria of potential sample polygons, a total area of 62,573 hectares (ha.). These polygons constituted 25% of the area in the MSdk subzone, with a mean polygon size of 13ha. For the purpose of sampling, I further limited my search to polygons that were greater than three hectares in size to reduce the influence of forest edges on a one hectare sample plot. The total number of potential sample polygons was now 4012, with a small reduction in total area to 60,602ha. Aspect directly affects fire behaviour through variations in the amount of solar radiation and wind that a plot will receive (Pyne et al. 1996). To test for the influence of aspect on fire frequency, I stratified potential sample polygons into southern or 'warm' aspect (165° - 185°) and northern or 'cool' aspect (345° - 105°) strata, representing 5.6% and 6.2% of the total area of the MSdk subzone, respectively (Figure 3.2). From these two strata, polygons were randomly selected and  34  plotted on a map to determine access and land ownership or the licensee. Selected polygons were sampled if they were within 750m of an active road and under the licensee/ownership of Tembec Industries Inc., Canfor Forest Products Ltd., British Columbia Timber Sales or Parks Canada.  NJ  A ft  ft  J  n v e r m e t  fr^ ft  X ', ft ft  Plot Location ft  MSdk Subzone Northern Aspect Stands \  14,355ha 5.6%  Southern Aspect Stands I  HP\S  (1261)  15,719ha, 6.2% (1446)  ft  Cranbrook »  0 10 20  40  60  1 Km  Figure 3.2 Location of 20 sampled plots in southeastern British Columbia, marked with stars. The darker grey area is the MSdk subzone and all identified stands with northern and southern aspects are in blue and red, respectively.  GIS software was used to objectively determine the UTM coordinates for plot centre. Using ArcMap™ 9.1,1 superimposed a 1.0 hectare circular plot at the centre of each polygon. If this plot  35  was entirely within the polygon and not greater than 500m from the road, the sample plot was placed at these coordinates. The location of plots was modified under three conditions: (1) if the plot was greater than 500m from the road, the plot was shifted closer to the road for easier access; (2) if the plot boundary extended outside the polygon boundaries, the plot was shifted to the closest position at which the entire plot was within the polygon; and (3) in situations where the 1 .Oha circular plot could not be placed within the boundaries of the polygon, I established multiple smaller sub-plots that totaled a search of 1 .Oha. The sample plot was located in the field using a handheld GPS unit and the UTM coordinates for plot centre. I confirmed that the characteristics of the polygon and sample plot matched the desired attributes indicated in the Forest Cover database. Plots were further scrutinized using the following criteria:  i. Plot was safe to sample; ii.  If the plot had been logged, the remaining stumps needed to be large enough (stump height >30cm) to determine if trees had been scarred, so that fire history could be represented; and  iii.  If the plot had been logged and fire scars were evident on stumps, the stumps had to be sound enough to extract samples or the fire scars had to be present in the remaining living trees.  Field Sampling From each plot centre, I measured eight 56m radii using either a handheld laser unit or, in dense stands, a surveyor chain. The radii were clearly flagged so that plot boundaries were discernable. Slope angle, aspect and elevation were recorded for each plot. Digital photographs were taken in the four cardinal directions from plot centre. Within each plot, all fire-scarred trees,  36  stumps, snags or logs (hereafter "fire-scarred trees") were identified and marked in numerical sequence. For each numbered fire-scarred tree, I recorded the species, location, whether it was alive or dead, estimated age and the number of external fire-scars. At each plot I sampled up to 10 fire-scarred trees. From all identified fire-scarred trees, I randomly selected five to sample. To capture the longest possible fire record, I also sampled the "best" five fire-scarred trees in each plot, which were subjectively selected as trees with the oldest fire scars, the most fire scars, and easy and safe to sample (Arno and Sneck 1977; Brown et al. 2000). This sampling scheme resulted in 10 of 20 plots having less than 10 samples taken, as six plots had the best five trees also randomly selected and four plots contained less than 10 firescarred trees. Fire-scarred trees were sampled using a gas-powered chainsaw either by cutting a complete stem cross-section or, when possible, a partial stem cross-section was removed using the procedure outlined in Chapter 2. Samples were collected to obtain the longest and most complete fire history for each tree. This required multiple samples from different heights along the stem of some trees. Cross-sections were removed systematically, so that for each section I visually identified the fire scars and selected the sample that provided the best fire, scar information. Samples were marked with the plot number, tree number, tree species and whether the tree was alive or dead. This information was also recorded on an individual tree sample sheet, along with the date, height of sample, diameter at breast height (where applicable), the number of fire scars observed, the distance of the tree from the sample point and its azimuth from centre. Digital photographs were also taken of each tree prior to sampling. Topographic properties associated with each plot likely had a strong influence on fire history by affecting the potential for fire ignition and fire behaviour. Slope steepness affects solar radiation and fuel moisture, influencing ignition potential, and has a direct effect on flame length  37  and rate of spread of a fire (Pyne et al. 1996).' Local changes in elevation, and thus temperature, will affect the rates of fine fuel production and decomposition (Moore et al. 1999b; Trofymow et al. 2002), length of fire season (Pyne et al. 1996) and likely the relative exposure of a plot to either lower elevation, higher frequency fire regimes or higher elevation, lower frequency regimes. To test for the influence of these properties on fire history, I recorded slope angle (S) in degrees at each plot centre. Plot elevation (£) in meters was obtained using 1:50,000 topographic maps for the area. Seasonal global solar radiation values (G) were calculated for each plot for the period of March to October, using the ArcView 3.3 GIS extension Solar Analyst® 1.0 and the appropriate digital elevation models (resolution: 25m x 25m). Global solar radiation includes both direct and diffuse radiation, both of which directly influence the amount of solar radiation a site receives (Dubayah and Rich 1995) and directly influence the fire regime through changes in ignition potential and fire behaviour. Seasonal values of solar radiation were calculated using the total yearly radiation received over the seven month period surrounding the summer fire season (April October). Topographic characteristics were categorized for all plots and the percentage of plots in each category was calculated and graphed to show the distribution of these characteristics.  Determining Fire Years and Seasons Samples were air dried and sanded to a maximum grit of 600 so that the cell structure was visible under magnification and annual rings could be accurately crossdated (Stokes and Smiley 1968). Fragile samples were mounted on boards prior to sanding. Samples were visually crossdated (Yamaguchi 1991) using regional ring-width chronologies developed for Douglas-fir, western larch and ponderosa pine trees in the study area (Daniels et al. 2006). The ring-width series of the majority (67%) of samples were measured to 0.001mm accuracy on a Velmex measuring bench  38  interfaced with a computer and statistically crossdated using the program COFECHA (GrissinoMayer 2001a). Fire scars on cross-dated samples were assigned an exact year and, when possible, a season. Each sample was assessed independently, then compared to other samples form the same study plot. In the first assessment of fire scars, only scars that were positively identified as caused by fire were included. In the second assessment, these positive fire dates were compared to the season and year of smaller, minor scars on samples from the same plot to verify they were caused by fire. Fire season was determined by the position of the fire scar tip within the annual growth ring and were classified as earlywood, latewood, dormant or unknown season (Baisan and Swetnam 1990; Dieterich and Swetnam 1984). A dormant-season fire is a fire scar that occurs between two annual growth rings. While little data on the phenology of cambial growth exists for my study area, Heyerdahl et al. (2007) found the cambial growth of interior Douglas-fir in the Stein Valley, B.C. ceased between mid-June and mid-August. According to the 1950 to 2007 fire records for the Invermere TSA (BCMOFR 2007), the greatest number of fires start in July and August and the majority of lightning fires (54%) occur in August. Therefore, I assumed the majority of dormantseason fire scars resulted from fires that burned late in the growing season after the annual growth ring was complete. For the 194 scars for which I determined the season of fire, I visually compared the proportion of earlywood, latewood and dormant-season fires in plots with southern versus northern aspects and at low (1097 to 1189 m.a.s.l.) versus middle (1250 to 1341 m.a.s.l.) versus high (1554 m.a.s.l.) elevations. I also compared the season of fire scars for western larch versus Douglas-fir.  39  Fire History Statistics /'. Plot-Level Fire Intervals To examine the effect of spatial scale, I calculated fire history statistics at the plot, subregional, and regional scales. At the plot scale, I used both the individual.tree and composite fire records. For individual trees, I calculated the mean number of scars per tree and individual tree fire intervals (ITFI) using the procedure outlined by Kou and Baker (2006). The ITFI was calculated by first calculating the mean fire interval for each sampled tree using all scar-to-scar intervals from that tree. These means were then averaged among all sampled trees, weighted by the number of intervals used to estimate each mean. All fire dates were combined into composite fire chronologies for each study plot from which I calculated summary statistics of mean and Weibull median fire intervals, using the program FHX2 (Grissino-Mayer 2001b). I estimated composite fire intervals for all plots as both the mean and Weibull median for number of years between all recorded fires. Estimating a mean fire interval assumes that the population are normally distributed, which is rarely true. Johnson (1979) first suggested the use of a Weibull distribution to quantify fire intervals, as its increased flexibility was better suited for non-normally distributed fire intervals. For mixed-severity fire regimes that include frequent, low-severity fires and infrequent high-severity fires, as hypothesized for my study area, fire interval data may not be represented by either a normal or Weibull distribution. The flexible power of the Weibull distribution may be limited for populations which are bimodally distributed, which is likely the case for mixed-severity fire regimes. Therefore I conducted Kolmogorov-Smirnov (K-S) goodness-of-fit tests between plot-level fire interval data and the normal and Weibull distributions using SAS 9.1.3.  40  /'/'. Sub-regional Fire Frequency At the sub-regional scale, I grouped plots according to slope aspect into northern and southern strata and calculated summary statistics using the program FHX2 (Grissino-Mayer 2001b). I plotted the composite fire record for the two strata and identified common fire years. Mean and Weibull median fire intervals were compared between aspect strata using the statistical software SAS 9.1.3 and a non-parametric Wilcoxon two-sample test.  /'/'/'. Regional Fire Frequency The regional-scale analyses assessed the fire scar data from all plots combined. To identify common fire years, recorded fires were graphed with respect to year for each aspect strata and for all plots, and fire years that were common to >10% and >25% of all plots in each strata were also graphed. In many areas in the western United States fire frequency has been shown to vary in response to changes in human land use (Beaty and Taylor 2001; Taylor and Skinner 2003; Veblen et al. 2000). To test for changes in fire frequency associated with changes in land-use in the southern Rocky Mountain Trench, I calculated and compared mean fire intervals and percent fire-scarred trees (Grissino-Mayer 2001b) for four periods: i.  Fire Suppression  1945-2005  ii.  European Settlement  1860-1944  iii.  Pre-European Settlement I  1775-1859  iv. Pre-European Settlement II  1690-1774  The start of European Settlement period was determined using documented changes in land use for the study area and the start of the Fire Suppression period was determined as the end of World War II when aerial fire suppression became highly effective. This created an 84-year time span of European Settlement.and, to facilitate comparisons, 84-year time spans were used for both  41  of the Pre-European Settlement periods. Mean fire intervals and their standard-deviation were calculated for each period. Paired comparisons were performed between all periods using the temporal analysis module of FXH2, which included a test for homogeneity of variances and t-tests to compare the means (Grissino-Mayer 2001b). The percentage of fire-scarred trees for all plots was also calculated and used to compare the temporal trends in fire occurrence over the entire study area (Veblen et al. 2000).  Logistic Regression of Fire intervals I used logistic regression modeling (SAS 9.1.3) as an alternate approach for quantifying fire frequency, as this approach does not require assumptions about the statistical distributions of fire intervals. I developed a logistic regression model to address three objectives: (a) quantify fire intervals from composite fire-scar records at the plot scale; (b) test for biophysical factors influencing fire intervals among plots using the regional dataset; and (c) compare fire intervals between aspect strata at the sub-regional scale. For the logistic regression model, a fire interval was defined as the number of years between two consecutive fires in the composite' fire record for each plot. In this model the fire intervals are the dependent response variable and each interval is binomially distributed, whereby each year between two consecutive fires is assigned the value zero and each year after the second fire is assigned a value of one. A fire interval is therefore the number of years that i  dependent variable equals zero and the binomial transition from zero to one indicates the occurrence of fire. The logistic regression model (equation 1) uses these binomial variables for all fire intervals to parameterize a probability curve of fire at each plot:  '  l-e  x  42  where, P is the probability that fire will occur and M is the model of fire likelihood, containing the x  r  independent variable(s). In the simplest case, the model of fire likelihood (Mi) includes only a single independent variable:  M =a + p t l  (2)  r  where f is time since fire in years, and a  and (3 are calculated regression coefficients. The  parameters for the probability curve of each plot can be used to estimate plot-level fire frequency. Fire is modeled to occur when the probability of fire is 50% (Pr - 0.5) and knowing a and p\ equations 1 and 2 can be used to solve for tos Time since fire when P, = 0.5 is a measure of central tendency for the plot-level fire intervals, heretofore the 'modeled fire interval'. To test for the influence of physical attributes on fire interval length, I combined all composite fire records into a single data set, which included 272 fire intervals from the 18 study plots. The physical attributes of each plot were added as independent variables to the model of fire likelihood. Plot-level fire intervals are likely influenced by the interaction between topography and climate. Elevation, slope angle and aspect drive fire behaviour, which determines intensity and rate of spread (Pyne et al. 1996). These plot attributes were incorporated into the model in a forward step-wise process, as follows:  = a + P -t + j3 A  M  2  x  M =a + p -t +  (3)  2  fi A + B3E  (4)  M =a + P t + p E + P L  (5)  M =a + &- t + P E + P S  (6)  3  4  x  2  r  5  3  3  4  s  M =a + p t + P E + p S + p G 6  r  3  5  6  (7)  43  where, A is class variable for plot aspect (1=Southern and 2=Northern), S and E are the slope angle and elevation of each plot, L is plot latitude (UTM coordinates) and G is the modeled solar radiation. Each parameter was tested for its effect on the linear function, by calculating maximum likelihood estimates, which were reported as p-values and odds-ratio estimates, using SAS 9.1.3. Only parameters that significantly contributed to each model were included in subsequent models. The resulting models were compared in order to identify the model with the lowest values for the fit statistics of -2log likelihood and Akaike's Information Criterion (AIC). The final model (Me) was used to test for the effect of varying topographical attributes on the length of plot-level fire intervals. Elevation, slope angle and solar radiation were each modeled for four different values, representing the range of values for the plots. For each attribute, the predicted probability of fire was calculated yearly for the four values, independently of the other attributes, using a time since fire that ranged from zero to 100. Mean values for the attributes not being tested were used in the model. The effect of each attribute on fire interval can be interpreted by comparing the four curves; fire is modeled to occur at P = 0.5 and difference in fire interval length between the curves at this point illustrates the effect of varying each attribute. To compare the probability of fire between northern and southern aspects, I stratified the data by aspect and combined all fire intervals for each stratum. I used an Odds Ratio Estimate (SAS 9.1.3) to compare the probability of fire occurring in the southern strata to the probability of fire occurring in the northern strata.  44  Results Fire History /'. Plot Characteristics The 20 sampled plots ranged in elevation from 1097 metres above sea level (m.a.s.l.) to 1554 m.a.s.l., representing the full elevational range of the Montane Spruce biogeoclimatic zone in this study area (Table 3.1). The northern-most plot location was 50° 40' 27"N by 115° 54' 09" W, while the southern-most plot was at 49° 24' 05"N by 115° 56' 26" W (Table 3.1). Slope angles ranged from 5° to 41° and modeled solar radiation from 6.3X10 Watt hours per metre squared 5  (WH/m2) to 11.1X10 WH/m (Table 3.1). 5  2  All plots had multiple aged strata of conifer tree species, but the specific species composition of the canopy and sub-canopy varied among the plots (Table 3.1). The canopy in the majority of plots (85%) contained dominant or veteran western larch (85%) and/or Douglas-fir (95%) trees in addition to other species. Eight plots contained hybrid spruce within the stand, but only plots 16 and 22 had spruce present in the canopy as dominants or co-dominants (Table 3.1). Lodgepole pine was present in the canopy of 12 stands and only plots 16 and 22 contained subalpine fir (Table 3.1).  //'. Fire Records External fire scars were present in trees at 18 of the 20 plots (Table 3.2). I found no visible evidence of fires, such as fire scars or charred logs and snags, at two of the study plots located on slopes with northern aspects. For plots with external scars, the density of fire-scarred trees ranged from 5/ha to 122/ha, with more scars present on live trees than snags, stumps or logs (Table 3.2). The mean (± standard deviation) number of fire-scarred trees per plot was 34.0 ± 26.7 from all plots, 42.1 ± 36.0 for the plots with northern aspects and 27.5 ± 12.3 for plots with southern 45  aspects. Western larch and Douglas-fir were the most common fire-scarred species; lodgepole  s pine and ponderosa pine also had fire scars but were less abundant (Table 3.2). I collected a total 164 samples from different trees. The majority of these samples were from dead trees (66%) and the most common tree species were western larch (52%) and Douglas-fir (38%). I successfully cross-dated 149 of 164 samples (Table 3.3). For 67% of these samples I used the statistical program COFECHA to verify scar dates and for the remaining 33% I used visual cross-dating methods. Analysis of the cross-dated samples yielded a total of 272 fire scars and 76 fire years (Table 3.3).  iii. Fire Season I determined the fire season for 194 (71%) of the 272 fire scars based on the intra-annual position of the scar (Figure 3.3). The majority of these scars occurred at the ring boundary (44%) as dormant season fires, which likely occurred in late-summer or fall as most lightning-caused fires in this area burn in August and burning conditions remain favourable into the fall (BCMOFR 2007). All other scars occurred during the formation of the earlywood or latewood (35% and 2 1 % respectively), indicating fires in spring or early summer.  46  80 4  >A//  Aspect  Elevation  Species  S2  »  6  n = 49  0  <*-  o  U)  JS c  n = 82  40  ^ • i Earlywood I I Latewood 1 Dormant  n = 112  n n = 117  d)  u l_  0)  °" 20  '  '  T~  ^r"  " V "  r  Figure 3.3 Variation in season of fire scars. Season was determined for 194 scars ("AH", left). Scars were stratified by plot aspect (center, left) and elevation (center right) and by tree species (right). Season was divided into three categories, earlywood, latewood and dormant, based on the position of the fire scar tip in the annual ring. Dormant season scars were located between two annual rings and were considered to have occurred in the fall of the previous year, due to favourable fall burning conditions in this region.  Fire season was similar between the northern- and southern-aspect strata (Figure 3.3). Fire season varied with plot elevation, where 66% of fires cars formed during the dormant season at the middle-elevation plots and 47% of scars formed in the earlywood of trees at the high elevation plots. Fire season also differed between the two major tree species, where 59% of fire scars in Douglas-fir were dormant-season scars, but the majority (43%) of scars formed in the earlywood of western larch (Figure 3.3).  47  Characteristics of the 20 sampled plots in southeastern British Columbia, stratified by aspect. General stand composition describes the species composition of the canopy and sub •canopy tree strata. Seasonal solar radiation is the total modeled solar radiation for April through  Table 3.1  October. Plot Number  Souther n Stratum  .  Northern Stratum  Plot  Slope  Elevation  Slope  /WH/m2\  General Stand Composition Canopy of western larch and Douglas-lir; sub-canopy of western larch and lodgepole pine  04  50° 00' 53"N  115° 30'27"W  1372  30  190  1106296  05  49° 24' 05"N  115° 56'26"W  1554  24  246  1002162  07  50° 12' 20"N  115° 28'49"W  1310  41  182  1051978  09  49° 35' 43"N  115° 56' 54"W  1143  14  256  953176  11  50° 40' 27"N  115° 54' 09"W  13  50° 27' 55"N  115° 36'17"W  14  50° 23' 24"N  18  49° 44' 35"N  20 21  Fire killed (2003) canopy of western larch; fire killed sub-canopy of western larch and lodgepole pine Fire killed (2001) canopy of western larch and Douglas-fir; fire killed sub-canopy of Douglas-fir and lodgepole pine; not salvage logged Canopy of western larch and Douglas-fir; sub-canopy of western larch and Douglas-fir; two remnant ponderosa pine snags  1554  31  185  1105213  Canopy of Douglas-fir; sub-canopy of Douglas-fir, lodgepole pine and hybrid spruce  1143  28  209  944033  Canopy of Douglas-fir and western larch; sub-canopy of Douglas-fir and lodgepole pine  115°34'33"W  1250  41 '  262  816576  Canopy of Douglas-fir and western larch; sub-canopy of Douglas-fir and lodgepole pine  115° 53' 52"W  1250  7  228  1018553  Canopy of western larch; younger cohort of Douglas-fir, lodgepole pine and hybrid spruce  50° 3 1 ' 12"N  115° 40' 59"W  1341  11  260  1003827  Canopy of western larch, Douglas-fir, hybrid spruce and lodgepole pine; sub-canopy of Douglas-fir, hybrid spruce and lodgepole pine  50° 39' 24"N  116° 16'21"W  1113  5  210  1017144  .  Canopy of Douglas-fir; sub-canopy of Douglas-fir, hybrid spruce and lodgepole pine Canopy of western larch and Douglas-fir; sub-canopy of predominantly Douglas-fir with some western larch Canopy of western larch and Douglas-fir; sub-canopy of western larch, Douglas-fir and lodgepole pine  01  50° 09' 00"N  115° 41 30"W  1128  12  70  916374  02  50° 0 5 ' 0 1 "N  115° 50' 14"W  1158  31  12  632397  06  50° 33' 13"N  116° 14' 09"W  1311  3  791079  Canopy of Douglas-fir trees; sub-canopy of Douglas-fir and hybrid spruce; some trembling aspen in'stand  10  50° 11'40"N  115° 48' 30"W  1189  24  32  761513  Canopy of western larch and Douglas-fir; younger cohort of Douglas-fir and lodgepole pine; some trembling aspen and birch  12  50° 31' 20 "N  115° 33' 53"W  1219  30  12  657998  Canopy of layer dominated by Douglas-fir with some western larch; sub-canopy of western larch, Douglas-fir and hybrid spruce; some birch in stand  15  50° 39' 30"N  116° 17' 08"W  1158  20  61  869142  .  '  . 22  Canopy of Douglas-fir; sub-canopy of Douglas-fir Canopy of western larch, Douglas-fir and hybrid spruce; sub-canopy of hybrid spruce, subalpine fir and lodgepole pine Canopy of western larch and Douglas-fir; younger cohort of Douglas-fir, lodgepole pine and hybrid spruce  16  49° 44' 57"N  115° 31'41"W  1490  37  58  934962  17  49° 35' 48"N  115° 57'19"W  1097  7  92  916495  19  50° 09' 28"N  115° 42'02"W  1189  28  72  799919  Canopy of western larch and Douglas-fir; sub-canopy of Douglas-fir  692315  Canopy of Douglas-fir and hybrid spruce; sub-canopy of hybrid spruce, Douglas-fir and subalpine fir  22  oo  Plot Latitude  Seasonal Solar Radiation  50° 10' 46"N  115° 13'35"W  1358  6  33  Table 3.2 Summary of fire-scarred material sampled from 20 plots in southeastern British Columbia, stratified by aspect. Total fire-scarred trees includes all live trees, snags, logs and stumps with external fire scars.  Fire Scarred Trees Live  Dead  Western . . Larch  „ . .. Doug as-fir  34  8  26  17  12  5  0  5  5  07  16  0  16  5  8  09  26  6  20 •  18  3  11  40  31  9  -  37  Plot Number  Total  04 05  Southern Stratum  All Northern All Plots  Sampled Trees  Lodgepoe Pine  Ponderosa „. Pine  K  3  ., , Unknown  1 2  17  12  5  4  13  35  23  12  12  17  6  18  25  8  17  . 25  20  50  25  25  4  18  27  21  27  14  13  -  24  3  275  127  148  90  132  40  2 1  01  24  12  12  12  9  1  02  34  21 -  13  23  7  4  c  With Multiple g r  c g r s  9  5 4  2  10  8  3  10  7  10  4  3  14  . . . . Total Sampled T  10  5  .  13  All Southern  Northern Stratum  Species  1  7  3  9  4  9  3  9  3  10  1  11  93  42  1  10  8  10  4  06  12  4  8  -  ' 11  1  8  2  10  37  11  26  26  10  1  10  6  12  122  98  24  98  10  13  10  3  15  12  5  7  -  12  16  -  -  17  19  2  17  18  19  77  38  39  20  22  -  -  |  1  7  -  -  1 57  -  337  191  146  197  612  318  294  287  -  116 248  | |  1  20 60  |  3  3  I  14  7  4  9  3  -  71  30  164  72  •  Table 3.3  Summary of fire history statistics for 18 of 20 sampled plots in southeastern British Columbia, stratified by aspect. The fire recording  period includes the time from first fire to death of last recording tree. Total fire scars, fire intervals, range of intervals and all fire frequency values were calculated for the full recording period. Crossdated Sample s  Fire Recording Period  Fire Scars  Total Intervals  Range of Intervals  Individual Tree Interval  04  9  1664 - 2006  14  4  42-137  05  4  1869 - 2003  9  3  19-94  10  1645 - 2001  23  6  09  10  1718 - 2006  26  11 .  10  1718 - 2006  13  7  14  9  18  Plot Number  Mean Fire Interval  Weibull Median Interval  Modeled Fire Interval  138.0  80.3  77.5  88.2  83.0  " 44.7  37.7  57.4  8-125  88.1  59.3  45.0  74.3  11  4-38  33.1  16.8  15.5  18.0  15  4  18-123  74  49.8  40.9  67.1  1662 - 2006  13  6  20-65  61.2  43.3  43.4  44.1  1739 - 2006  13  3  16-108  35.5  50.0  40.6  57.6  9  1720 - 2006  13  3  36-85  81.5  68.3  69.1  68.9  20  8  1831 -2006  14  2  31 -55  43.0  43.0  21  9  1831 -2006  10  2  27-28  27.0  27.5  150  1-44  14  8-96  68.1  29.8  25.3  36.0  12-128  72.7  52.8  45.4  56.0 34.2 60.5  07  Southern Stratum  All Southern 01 02  Northern Stratum  .  85 1509-2006  9  1707 - 2006  9  33 15  5  43.5 28.0 56.8  06  7  1861 -2006  9  3  8-59  17.5  31.3  27.1  10  10  1643 - 2006  17  4  20-89  59.7  60.0  58.8  12  8  1869 - 2006  12  2•  17 - 33  . 37.5  25.0  -  25.8  15  5  1749 - 2006  5  1  110  110  -  110.5  17  7  1629 - 2006  13  8  5-109  65.8  38.9  33.0  45.5  19  9  1621 -2006  18  7  15-85  59.3  38.1  36.0  40.6  122  1-41  -  44.8  272  1-41  All Northern All Plots  64 149  |  |  iv. Plot-Level Fire Frequency At the plot level, sampled trees had up to seven fire scars, with 72 of 164 trees having multiple fire scars from which I calculated the ITFI (Table 3.2). The mean intervals for individual trees ranged from 8 years to 232 years and the ITFI for plots ranged from 17.5 to 138 years (Table 3.3). At the plot level, when all fire dates were combined, the full fire period for all plots spanned from 1509 to 2006 and the common fire period, when all plots were recording fire, spanned from 1869 to 2001 (Table 3.3). The number of fire intervals ranged from one to 14 (Table 3.3), with an average (± standard deviation) of 4.9 ± 3.3 fires per plot. The shortest fire interval was four years and the longest was 137 years (Table 3.3). The time since last fire at all plots ranged from three to 123 years and for nine of 18 plots the time since last fire exceeded the longest fire interval in the corresponding plot-level fire record. Mean composite fire intervals at the plot level ranged from 17 to 80 years and Weibull median fire intervals from 16 to 78 years (Figure 3.4). Weibull median composite fire intervals could not be calculated for four of the 18 plots, as one plot had a single interval (110 years) and three plots had two intervals, allowing only for the calculation of mean fire intervals. Modeled fire intervals ranged from 18 to 110.5 years and in all cases were greater than both the mean and Weibull median values (Table 3.3).  51  160  Southern Aspect  140 (A  Composite  V  Northern Aspect Composite  120  Mean Fire Interval  •  Weibull Median Fire Interval  O  Modelled Mean  •  Outliers  100  >  o  80 60  ii  40 20  O  H  04  —i  P04 P05 P07 P09 P11 P13 P14 P18 P20 P21  1  P01  1  1  1  1 — :  1  1  —  P02 P06 P10 P12 P15 P17 P19  Figure 3.4 Composite fire intervals for 18 plots. The shaded box represents the 25 to 75 percentiles, whiskers are the 10 to 90 percentiles and outlying values are shown as black circles. The horizontal line in each box is the median fire interval; triangles are mean fire intervals; squares are Weibull median fire intervals; diamonds are modeled mean fire intervals. th  th  th  th  Estimation of fire frequency using different measures of central tendency resulted in different estimates of fire frequency at the plot level (Figure 3.4), which reflects the distributions of the interval data and the number of intervals per plot. Although I found 14 of 17 plots to fit a normal distribution (P>0.05, K-S test) and all plots to fit a Weibull distribution (P>0.05, K-S test), the calculated D-statistics were relatively low. For example, the D-statistic for the normal distributions, with P>0.05, ranged from 0.117 to 0.379 and for the Weibull distribution from 0.105 to 0.399, while the critical value for a D-statistic is 0.563 ( a = 0.05) when the number of observations is ^5, which was true for 11 of 18 plots (Conover 1999).  v. Sub-regional Fire Frequency Historical fire patterns were similar between the two aspect strata. At the 10 plots from the southern aspect stratum, I observed 150 fire scars and 41 fire intervals (Table 3), with the first fire in 1645 and the last fire in 2003 and a common fire period from 1869-2001 '(Figure 3.5). At the eight plots from the northern aspect stratum, I observed 122 fire scars and 44 fire intervals (Table 3.3), with the first fire in 1509 and the last fire in 1971 and a common fire period from 1869-2006 (Figure 3.6). Eight fire years were common to >10% of the southern-aspect plots, with two fire years common to >25% of plots (Figure 3.5); six fire years were common to >10% of the northernaspect plots, with two fire years were common to >25% (Figure 3.6). Mean and Weibull median fire intervals were not significantly different between northern and southern aspect plots (S=52.0 and S=30.0 respectively, P > 0.05). Modeled fire intervals were found to be significantly longer for southern aspect plots than for northern aspect plots (Odds Ratio Northern vs. Southern = 0.667).  53  Southern Aspect  >25% Scarred 1600  1700  1800  1900  2000  Fire Year Figure 3.5 Fire-scar record from 1645 to 2003 for plots with southern aspects (P04 to P20) in the montane forests of the southern Rocky Mountain Trench. Horizontal lines show the time span of each plot-level fire chronology and triangles indicate a fire scar., where hollow triangles show years when one tree was scarred and solid triangles are years when >1 tree was scarred. Gray vertical dashed lines indicate fire years recorded at >10% of plots and black vertical lines when >25% of plots recorded a fire.  54  Northern Aspect P01  t  WW7—V-  P02  I II I  P06 P10  -Hi-  - N7J  P12  t  P15  • I • V-  P17 I  P19  I  .  Jh^r?—*—Y  P21 >25% Scarred 1500  1600  1700  t  t  m  -4—4-  >10% Scarred  II  1800  1900  2000  Fire Year  Figure 3.6 Fire-scar record from 1501 to 1971 for all plots with northern aspects (P01 to P19) in the montane forests of the southern Rocky Mountain Trench. Horizontal lines show the time span of each plot-level fire chronology and triangles indicate a fire scar, where hollow triangles show years when one tree was scarred'and solid triangles are years when >1 tree was scarred. Gray vertical dashed lines indicate fire years recorded at >10% of plots and black vertical lines when >25% of plots recorded a fire.  55  vi. Regional Fire Frequency Years in which more than two plots recorded a fire were considered regional fire years and this occurred seven times over the fire period: 1706, 1718, 1831, 1847, 1886, 1888 and 1889 (Figure 3.7). The mean fire interval was lowest during the European settlement period (3.8 ± 3.1 years) and greatest during the fire-suppression period (9.6 ± 6.8years) with intermediate intervals during the pre-European settlement periods I (5.9 ± 5.2years) and II (4.3 + 3.3years) (Figure 3.7). The only two periods that were significantly different were the European settlement period and the fire suppression period {t = -2.31, P<0.05). The percentage of fire-scarred trees was significantly lower for the pre-European settlement period II compared to the European settlement period (f = -2.32, P<0.05), but did not change significantly during the fire suppression period.  56  Northern Plots  n = 45  Southern Plots  n = 42  '2  0 4  3  •—  H  2  All Plots  n = 76  (4 3)  1400  1500  1600  1700  (5.9)  1800  P8; '  6> |  1900  2000  Fire Year Figure 3.7 Frequency of fire in the montane forests of the southern Rocky Mountain Trench. Plots are grouped by aspect (top and middle) and all plots combined (bottom); n is the total number of fire years for each group. Regional fire years (^3 plots burned in the same year) occurred in 1706,1718,1831,1847,1886,1888 and 1889. In the bottom panel, the alternating gray and white bands represent the pre-European settlement periods II (1690 - 1774) and I (1775 -1859), European settlement period (1860 -1944) and fire suppression period (1945 - 2005). For each period, the mean fire interval in years is in parentheses.  57  Influence of Site Physical Attributes on Fire Intervals Of the six models I developed, model IVta best explained plot-level fire intervals as indicated by the low -2log likelihood and AIC values (Table 3.4). This model included the parameters elevation, slope angle and solar radiation; slope aspect and latitude did not significantly contribute to the models. Plot elevation contributed the most to the model'O.e., it reduced the -2log likelihood the most). Increased elevation, slope angle and solar radiation all caused a significant increase in fire intervals at the plot scale (Figure 3.8). For elevation, the modeled fire interval increased from 47 to 61 years with the increase from 1100m.a.s.l. to 1550m.a.s.L. As solar radiation increased from 6.3X10 WH/m to 1.1X10 WH/m fire intervals increased from 41 to 58 years. The greatest 5  2  6  2  increase in fire interval due to a single topographic attribute was related to slope angle. As slope angle increased from 0° on flat sites to 45° on the steepest slopes in the study area, fire intervals increased from 36 to 67 years.  58  0.0  J  ,  1  1  ,  1  1 -  0  20  40  60  80  100  Fire Interval (years) Figure 3.8 Probability of fire through time for forests of different elevations (top), slope angles (middle), and solar radiation (bottom). Curves were calculated using logistic regression model M6 in equation 7 and four values for each attribute that represent the range of values observed in the 20 sampled plots. For each curve, the fire interval corresponding to a probability of 0.5 is the modeled fire interval. Modeled fire intervals increased with increased plot elevation, plot slope angle and modeled plot solar radiation.  59  Table 3.4  Comparison of six models on physical attribute influence of plot-level fire intervals. Independent variables included time since fire (TSF) and five physical plot attributes. Models were constructed using a forward step-wise procedure. Maximum likelihood estimates are presented as p-values and odds-ratio (in parentheses); models were assessed using the -2log likelihood (-2log) and Akaike's Information Criterion (AIC). Asterisk marks insignificant values.  Model Mi  M  2  M  3  M, M  5  M  6  Test Statistics  Independent Variables TSF <0.0001 (0.966) <0.0001 (0.966) <0.0001 (0.964) <0.0001 (0.964) <0.0001 (0.964) <0.0001 (0.963)  Strata  <0.0001 (0.741) 0.3158* (1.065) -  Elevation  -  <0.0001 (1.003) <0.0001 (1.003) <0.0001 (1.002) 0.0001 (1.001) •  Latitude  Slope  Radiation  •2log  AIC  8315  8319  -  -  -  8284  8290  -  -  -  8159  8167  0.3242* (1.000)  -  -  8159  8167  -  8109  8117  8084  8083  -  <0.0001 (1.02) <0.0001 (1.026)  <0.0001 (1.00)  Discussion  Mixed-Severity Fire Montane Spruce forests in southeastern British Columbia have multiple lines of evidence supporting the hypothesis that the historic fire regime included a combination of low-to-moderate severity and high-severity fires. Low-to-moderate severity fires are evident in the current forest structure and the fire scar record. Twenty-five percent of the MSdk subzone in this area has an uneven-aged forest structure, consistent with the structure of mixed-conifer forests studied in other parts of western North America that exhibited evidence of low-to-moderate severity fires (Arno et al. 2000; Beaty and Taylor 2001; Heyerdahl et al. 2001; Heyerdahl et al. 2007; Taylor and Skinner 2003; Veblen et al. 2000; Wright and Agee 2004). In the studied stands, the fire scar record was long, including 76 fire years between 1509 and 2006, and variable among stands.  60  Fire scars were abundant at 18 of 20 sampled plots and were found primarily on western larch and Douglas-fir trees. Numerous individual trees possessed multiple fire scars, including trees with diameters as small as 2.7 cm (at a height of 90cm above the ground) when they were first scarred and thin-barked lodgepole pine. Multiple fire scars on small trees and on lodgepole pine indicated the occurrence of repeated, low-severity fires. Within single plots, fire intensity and/or spread was variable as not all trees recorded the same fires. Between plots, the periods of the fire records varied considerably. Plots with short fire records provided indirect evidence that high-severity fires occurred at some plots relatively recently, killing recorder trees and consuming fire-scarred coarse woody debris (Ehle and Baker 2003). Occurrence of high-severity fires in the MSdk subzone is further supported by the large percentage of the forest area that is considered even-aged and dominated by lodgepole pine (Braumandl 1992), although some of this area may be due to modern silvicultural practices or intentional fires set to clear land during the mining era at the turn of the last century. Only two of the 20 sampled stands had no external fire scars and possessed stand features which suggested they developed in the absence of fire. Hybrid white spruce was dominant or co-dominant in both stands. It is a thin-barked species with a low ground-to-crown height, characteristics which greatly increase the likelihood of tree mortality in a fire (Kobziar et al. 2006). The large diameter of these trees suggests they were older, which may indicate any of the following: a) a fire-free period of similar length, perhaps occurring by chance, b) that only very lowintensity fires that did not scar trees have occurred or c) these plots are areas of refugia from fire or locations where shifts in fire behaviour reduce the chance of fires burning. The former situation may be related to topographic properties (Agee 1993; Heyerdahl et al. 2001) or site moisture (Everett et al. 2003; Taylor and Skinner 2003). Assessment of.the age structure of these and  61  surrounding stands could identify distinct cohorts to estimate time since the last stand-replacing fire or if low-severity fires had burned but left little visible evidence.  Measures of Fire Frequency Fire frequency or fire return intervals quantify temporal aspects of a fire regime and are used to assess changes in the regime and to guide forest management including timber harvest, ecological restoration and hazardous fuels mitigation. The most appropriate measure of fire frequency is a debated topic within fire history research (Baker and Ehle 2001; Kou and Baker 2006; Van Home and Fule 2006), as fire frequency varies with spatial scale and is highly dependent upon research design and statistical methods of analysis. I found that fire frequency varied considerably between the individual tree and composite methods. At the individual-tree scale, mean fire intervals were the longest and represented the frequency of fire at single point within the stand. At the plot-level, the individual tree fire interval (ITFI) and the three composite measures (mean fire interval, Weibull-median interval and the modeled fire interval) represented fire frequency as a point estimate for each plot within the landscape. All four measures of central tendency varied in their estimation of fire frequency (Table 3.3, Figure 3.4), with the ITFI nearly always the greatest, followed by the modeled fire interval, while the mean and Weibull-median fire intervals were the smallest and generally similar. It was expected that the ITFI would be the greatest as this method of calculation does not include all fire years from a plot. Rather, it is calculated using only those trees with >1 fire year and assumes that trees are perfect recorders of fire. Composite measures of fire frequency use all fire years from the plot and assume that all fires burned over the entire plot. I believe that for this study the composite method better represents the true frequency of fire at the spatial scale of the 1.0 ha study plot. Moreover, I suggest that few fires that significantly influence stand structure are smaller  62  than 1.0 ha in size and, therefore, using a composite fire record would not bias the calculated fire frequency or misguide forest managers. A primary goal of my research was to develop a robust research design that would ensure the study sites were representative of fire frequency across the complex stands of the MSdk subzone in southeastern B.C. Baker and Ehle (2001) pointed out that the criteria used to select sites and stands in many ponderosa pine studies in the western United States were largely focused on the temporal component of the fire regime. Sites were targeted for sampling when they contained older trees, high fire-scar densities, trees with multiple scars and relict, un-harvested stands. The targetted method of site selection can significantly bias the fire interval to shorter values which may not be representative of the entire landscape. To avoid a similar bias in this study, I used a GIS to stratify forest cover data and identify the population of interest, which included stands with complex stand structure where logging had not occurred recently. I randomly selected stands to sample from this population and objectively placed the sample plot in each stand. Therefore, the reported fire frequencies are an unbiased estimate representing the entire population-of stands with complex forest structure. Consistent with Baker and Ehle's (2001) observation that targeted sampling may overestimate the actual frequency of fire in the landscape, fire frequencies at my study sites were generally longer than those recorded at 10 sites in the southern Rocky Mountain Trench that were selected using a targeted sampling design (Daniels et al. 2007). Stands sampled by Daniels et al. (2007) were targeted to represent forest composition and structures typical of old-growth management areas and specific wildlife habitat for the northern goshawk {Accipiter gentilis). Areas containing numerous Douglas-fir and western larch veteran trees with multiple fire scars were preferentially selected for sampling. As a result, the targeted research plots reported by Daniels et al. (2007) represent only a subset of the structurally complex forests in the landscape. Shorter fire  63  intervals associated with these stands are not representative of the full range of variation in the landscape. Forest managers need to be made aware of the differences in calculated fire frequency resulting from different sampling methods and apply research results knowing the assumptions and limitations of the data. A third limitation when quantifying fire frequency are the assumptions underlying the statistical models used to quantify fire intervals. Typically, the mean or median fire interval is used to summarize the average interval between fires and normal or Weibull probability distributions are most commonly used (Fall 1998). In this study, the use of either of these two distributions was limited by: a) the small number of intervals ( < 5 ) at more than half of plots; and b) fire intervals were bimodally distributed for over half of the plots (Figure 3,9). When the number of intervals is low, there is low power for the K-S test to adequately test the fit of a probability distribution. Subsequently, nearly all plot-level fire interval data were assessed as not significantly different from a normal distribution and none were significantly different from a Weibull distribution. Evidently, the bimodal distribution of fire intervals associated with a mixed-severity regime are not accurately summarized by the normal or Weibull distributions. I proposed logistic regression modeling as a more accurate measure of central tendency, as it required no assumptions about the distribution of the data. Modeled fire intervals were within six years of the means and Weibull medians for plots that were not bimodally distributed. For plots with a bi-modal distribution the modeled fire intervals were greater than the means and medians. This result suggests that the use of probability distributions to summarize fire frequency for some mixed-severity regimes may underestimate the true frequency.  64  Southern Aspect  o c  • CT  P04  P05  P07  Mean  v  Composite  P11  P09  c  Modeled Mean  P13  c  Weibull Median  P14  •  Outliers  P18  P20  P21  0) >  20 40 SO SD 100 120 UD  0  20 40 CO SO 100 120 140  0  20 40 B0 SO 100 120 140  0  20 40 60 80 100 120 140  0  20 40 SO SO 100 120 140  0  20 40 60 SO 100 120 140  0  20 40 60 80 100 1 20 140  0  20 40 60 SO 100 120 140  0  20 40 60 80 100 120 140  0  20 40 GO 3D 100 120 140  Fire Interval (years)  NOrthem ASpeCt  Composite  v  Mean  o  Modeled Mean  P12  P15  a  Weibull Median  •  Outliers  O  c  0 3 CT (V  P01  P02  03 >  P10  P06  P17  P19  - J"  re lap n 0  20 40 SO 80 100 120 140  0  20 40 60 80 100 120 140  O 20 40 60 SO 100 120 140  0  20 40 60 SO 100 120 140  0  20 40 60 80 100 120 140  0  20 40 60 80 100 120 140  0  20 40 SO SO 100 120 140  0  20 40 60 SO 100 120 1  Fire Interval (years)  Figure 3.9 Distribution of fire intervals for each plot and calculated measures of central tendency. Box plots represent the 25 to 75 percentiles, whiskers show the 10 to 90 percentiles and outlying values are shown as black circles. The horizontal line in each box is the median while the mean is shown as a triangle, modeled mean as a diamond and Weibull median as a square. Modeled means were calculated using logistic regression. th  th  Ol  th  th  Spatial Variation in Fire Frequency Fire frequency varied considerably between and among studied plots. Fire frequency is known to be strongly dependent upon the influences of both top-down and bottom-up controls (Heyerdahl et al. 2001; Lertzman et al. 1998). This study was primarily focused on the influence of bottom-up controls associated with physical plot attributes of aspect, elevation, latitude, slope angle and modeled solar radiation. Fires were not more frequent on southern aspect sites than northern aspect sites when the mean fire intervals or Weibull median intervals were compared. However, the results of the logistic regression indicated fire intervals were significantly longer for southern aspect plots, a surprising result. Previous studies in similar forest types in western North America have found that aspect significantly influences fire frequency (Heyerdahl et al. 2001; Heyerdahl et ah 2007; Taylor and Skinner 2003). Specifically, stands with southern aspects had shorter fire intervals due to increased solar radiation and lowered fuel moisture contents. In my study area, longer fire intervals at the southern- aspect plots could be the result of one of the two scenarios. Firstly, high frequency may have resulted in low fire intensities which, in turn, reduced the number of resultant fire scars at the plots with southern aspects. This would result in more fire scarred trees within the northern plots than the southern plots, which was indeed the case. On average, there were 12 more firescarred trees per plot in the northern stratum than the southern stratum (Table 3.3). Alternately, aspect is not a strong determinate of fire frequency in the montane forests of southeastern B.C. The larger variation in fire frequency observed between plots rather than between aspect strata provides strong support for the latter scenario (Figure 3,4). The total number of fire scar samples, fire scars and years with fire scars were very similar between the northern and southern aspect  66  strata (Tables 3.2 & 3.3), while the number and length of fire intervals at the plot level vary considerably (Figures 3.5, 3.6 & 3.7). Instead of aspect, I suggest that other topographic controls have a dominant influence on fire frequency. Using logistic regression, I found elevation to be the most important attribute influencing fire frequency, where increased elevation caused reduced fire frequency (Figure 3.8). This was expected as snow cover duration lengthens with increasing elevation, increasing the fuel moisture content and shortening the period during which fuels could support fire (Pyne et al. 1996). I believe that plots found at lower elevations and nearer to the middle of the Rocky Mountain Trench likely had increased exposure to adjacent, higher frequency fire regimes. Current forest structure (stands of ponderosa pine and Douglas-fir) in conjunction with some research (Gray et al. 1998) suggests that low-severity fires occurred historically in the valley bottom of the Trench. This forest structure does not extend to all surrounding watersheds though, which may decrease the likelihood of fire spread from a low-severity regime. In other areas of western North America, fire frequency in ponderosa pine and mixed-conifer forests has been found to decrease with elevation (Taylor and Skinner 2003; Wolf and Mast 1998), although some exceptions to this trend have been documented (Heyerdahl et al. 2007). Increased slope angle and solar radiation were also associated with decreased fire frequency in the logistic regression models (Figure 3.8); both results are surprising. As Heyerdahl et al. (2001) pointed out, slope angle does not drive the long-term likelihood of fire occurrence at a given point on the landscape. However, increased rate of spread and fire intensity are associated with increased slope angle (Albini 1992; Mcalp'ine et al. 1991) and will result in greater fire severity. This may influence the location of complex, uneven-aged stands versus even-aged stands in the landscape, but it does not explain the differences in fire frequency among stands with similar, complex structures.  •  67  Decreased fire frequency was associated with increased solar radiation. This lends further support to the possibility that a high frequency of fires resulted in relatively low intensities and less fire-scarring. Alternately, sites receiving high levels of solar radiation may have greater rates of evapot'ranspiration resulting in lower soil moisture availability.  On such sites, reduced plant  productivity would limit fuels accumulation and the spread of fires following ignition, resulting in fewer fire scars. Surface fire intensity, a critical factor in fire-scarring (Gutsell and Johnson 1996), is determined primarily by the amount of fuel consumed and the rate of fire spread (Brown and Davis 1973). Further research in these montane forests is required to determine if fuel accumulations and fuel properties differ with respect to plot aspect and if these differences are visible in the fire scar record. Latitude did not contribute significantly to explaining fire frequency between plots. Regional gradients in fire frequency, decreasing frequency with increasing latitude, have been identified for some forests in North America (Brown and Shepperd 2001), but not for all (Heyerdahl et al. 2007; Wright and Agee 2004). A lack of difference in fire frequency associated with latitude in this study suggests that for these montane forests, vegetation and fire frequency do not vary significantly at the regional scale, between the northern and southern extents of the study area, as hypothesized in Heyerdahl et al. (2001) to occur in the Blue Mountains of Washington. Fire season is an important determinant in fire severity and species response to fire (Agee 1993; Kauffman and Martin 1990). Across all plots, fire season varied most among plot elevation groups and between species groups (Figure 3.3). At mid-elevations most scars were in the dormant season, while most scars at high elevation sites were in the earlywood. It is possible that fires burned at different times in these two groups of forests. However, seasonality of fire scars reflects both differences in length and start/end dates of the fire season and tree phenology at different elevations. As previously noted, fire season is shorter at higher elevations and this is  68  reflected in more middle elevation plots recording fires in the dormant season; however it does not explain why the high elevation plots had slightly more fires in the early season. Environmental factors affecting the length of growing season, such as temperature, photoperiod, thermoperiod and amount and periodicity of rain, are selective forces (Spurr and Barnes 1980). At highelevations these factors trigger later growing seasons to prevent damage by late frosts, which, could cause the same fire to be recorded in the latewood at low-elevations and earlywood at highelevations. This hypothesis could be tested using elevational transect sampling for fire history where seasonal fire scar position should vary for the same fire year with tree elevation. Douglas-fir and western larch differed in the season of recorded fire, which is likely due to species-specific differences in physiology. While there is no data on the specific timing of cambial growth patterns for these two species in my study area, other research in southern B.C. noted Douglas-fir starts radial growth in mid-May, with cessation in mid-June to mid-August and concluded that Douglas-fir ceases radial growth earlier in the season than other coniferous species (Heyerdahl et al. 2007). As western larch is a deciduous conifer, I propose it begins and ceases radial growth after Douglas-fir. Therefore, the scar position resulting from a single fire could be found in the earlywood of western larch and in the dormant-season position of Douglas-fir.  Temporal Variation in Fire Frequency Anthropogenic changes in land-use significantly impacted the fire regime and this was evident in the fire record. The period of documented European settlement had the shortest mean fire intervals (Figure 3.7) and a significantly higher percent of scarred trees than only the earliest pre-European settlement period. Fire intervals were shortest and the percent scarred trees were lowest during the fire suppression period. These results are consistent with other parts of western North America where both human land-use and climatic trends were linked to increases in fire  69  frequency during settlement and subsequent decreases in fire frequency during the 20 century th  (Veblen et al. 2000). A separate research effort has shown climate to be a strong factor in determining when historic fires burned for these plots and 10 additional plots in southeastern B.C. (Daniels et al. 2007). Unusually long fire-free intervals during the 20 century is one criterion used to document th  changes to fire regimes (Swetnam et al. 1999). In my study area, half of the stands currently have fire-free intervals that were shorter than the maximum fire interval recorded in the plot and were within the historic range of variability. This suggests that the fire regime has been altered on only part of the landscape, making it difficult to determine where and how ecological restoration and/or fuels mitigation are needed. At the same time, it is important to consider the spatial context of individual stands. For example, an individual stand may have a relatively short fire-free interval, but still be susceptible to severe fires because it is surrounded by stands with long fire-free intervals in which fuels have accumulated. In this case, prescriptions for restoration would need to consider the entire landscape and begin with those stands that have deviated the most from their range of historic variability. In some forests, there is growing evidence that fire severity is increasing as a result of prolonged fire-free intervals (Covington and Moore 1994; Moore et al. 1999a). In my study, I documented six fires during the fire suppression period, including two very recent fires in 2001 and 2003, which is rare in fire history studies conducted in western North America. For various reasons, recently burned stands are often not sampled for fire history. By including these two stands in my study, I can assess the impacts of two modern fires. Both stands experienced high-severity fire with mortality of all trees within the sample plot, including numerous western larch that were >230 years old and several trees that were nearly 300 years old. When these fires started, the fire hazard was "extreme" in both areas, meaning that predicted rate of fire spread and total fuel  70  available to burn were at their highest (Van Wagner 1987). Fire suppression is highly effective when fire hazard is low to moderate; however, the probability of containment is greatly reduced as the initial fire size and initial rates of spread increase (Arienti et al. 2006). This is supported by other research which has found that the average time between fire discovery and control has increased by as much as 37 days since 1970 (Westerling et al. 2006). The fires in 2001 and 2003 are inconsistent with the historic fire-scar records and may be outside the historic range of variability for some parts of the landscape. Our ability to effectively suppress low and moderateseverity fires and not high-severity fires could be resulting in a shift in the fire regime of these stands as fire is only occurring when conditions are extreme. Further research could test if recent fires in this area have reduced fire behaviour heterogeneity.  Conclusions Montane spruce forests of southeastern B.C. were influenced by a mixed-severity fire regime. The MSdk subzone consists of a mixture of both even-aged and structurally complex stands, with complex stands constituting 25% of the area. The majority of sampled plots experienced multiple, low-severity fires, with considerable variability in fire frequency and the length of the fire recording period. The spatial scale of examination influenced the calculated frequency of fires, with the shortest fire intervals at the regional scale and the longest fire intervals at the scale of individual trees. Local topography significantly influenced the frequency of fires at each plot. Plot elevation was more important than slope angle and solar radiation, while aspect and latitude were not significant. Fire frequency was influenced by changes in human land-use, increasing during the period of European settlement and decreasing during the fire suppression period.  71  If management based on natural'disturbance is used to maintain ecological resilience and complex stand structure, management strategies need to incorporate observed variability, the drivers of variability and the scale at which these are influential. Focus should be placed on lower elevation areas where frequencies were historically higher and in areas where the time since the last fire has deviated the most from the historic range of intervals. Management guidelines on the frequency of disturbance within these stands must be re-considered. It is apparent from this research that fire was disturbing these at a frequency much less than every 150 to 200years as suggested by current forest management guidelines provided by the provincial government (BCMOFR 1995).  72  Chapter 4. Looking Forward: Suggestions for Sustainable Forest Management in the MSdk Subzone of Southeastern British Columbia  Introduction This research was a collaborative effort between the University of British Columbia, Tembec Industries Inc., Canfor Corporation, Parks Canada and the BC Ministry of Forests and Range. Contributing agencies share a common interest in practicing Sustainable Forest Management (SFM) in southeastern British Columbia, requiring an adequate ecological foundation on which to base management practices. In particular, the need for an increased understanding of the historic fire regime in the Dry Cool Montane Spruce (MSdk) subzone was identified by collaborating agencies. Concerns had arisen that the ecological resilience of structurally complex forests in the MSdk subzone is threatened due to a lack of disturbance or the occurrence disturbances which are outside of the range of historic variability. Ecological resilience, as summarized in Peterson et al. (1998), is a measure of how much change is required to transform a system from being maintained by one set of processes and structures to another set of processes and structures. Historic fire regimes in the MSdk subzone have resulted in structurally complex stands, but current forest management strategies have changed disturbance patterns by increasing the proportion of stand-replacing disturbances on the landscape through forest harvesting and decreasing the proportion of stand-maintaining disturbances through fire suppression. The net effect of these changes may be the transformation of these forest ecosystems to a different state. Natural disturbance based management (NDBM) has been suggested as a way to maintain ecological resilience using silvicultural strategies, which  73  retain the structures and processes that perpetuate desired states while reducing those that enhance resilience of undesirable states (Bergeron et al. 2002; Drever et al. 2006). The objective of this chapter is to address the requirements of forest management in achieving SFM and promoting ecological resilience in the complex stands of the MSdk. NDBM will be referred to when details applying to the implementation of particular silvicultural strategies or planning approaches mimic the fire regime characteristics. Key fire regime characteristics presented in Chapter 3 are examined to address the following questions: (1) What is the range of' historic variability in stands with complex structure and how does it vary spatially and temporally? (2) How can forest management incorporate the newly quantified fire regime parameters as sustainability thresholds for SFM to promote ecological resilience within complex stands? and (3) Is restoration required in the complex stands of the MSdk subzone? These questions are addressed as they apply to five critical attributes of the fire regime for the complex stands in the MSdk of southeastern B.C.  Critical Fire Regime Attributes Quantity on the Landscape Throughout the province, the BC Ministry of Forests and Range has classified vegetation cover using aerial photographs and a classification scheme that identifies homogenous forest stands. These stands have been delineated, classified by features such as species composition, age class and height class, and digitized. The resulting forest cover maps provide a database of "polygons" (discrete patches of forest) that can be searched for compositional and structural attributes of interest. As detailed in Chapter 3,1 searched all forest cover polygons in the Invermere TSA and a portion of the Cranbrook TSA (grey area in Figure 3.1) for the following attributes:  74  v. Dry Cool Montane Spruce (MSdk) biogeoclimatic subzone; and vi. Documented presence of at least two separate cohorts; or vii. Oldest tree cohort established before 1860; and viii. Not logged between 1950 - 1999. The entire study area contained 83,535 polygons and 5812 of these met the criteria of my sample population, a total area of 62,573 hectares (ha). This sample population constituted 25% of the area in the MSdk subzone, with a mean polygon size of 13ha. Low- and moderate-severity fires were very evident in the current forest structure and fire history of sampled stands, suggesting that currently 25% of the MSdk resulted from such fires. Fire intervals were variable within and among plots (Figure 3.4); however, the longest recorded fire interval was 137 years. The Natural Disturbance Type (NDT) classification for this entire subzone is NDT 3, indicating a fire-dominated disturbance regime with return intervals of 150-200 years for stand-replacing fires (BCMOF 1995). A discrepancy between my findings and the NDT classification clearly is evident and suggests a misclassification of the disturbance regime for at least 25% of the MSdk subzone. I suggest that a mixed-severity fire regime was the dominant disturbance for the forests of MSdk subzone in southeastern B.C. Mixed-severity regimes are an intermediate between the infrequent, high-severity regimes of the high elevation Engelmann-Spruce-Subalpine fir (ESSF) forests and the frequent, low-severity fire regimes of the low-elevation Ponderosa Pine (PP) and Interior Douglas-fir (IDF) forests. However this regime is not simply an intermediate state between low- and high-severity fire regimes, rather it exists as part of a continuum within the range of fire severity with individual fires exhibiting a dynamic combination of low-, moderate- and high-severity fire behaviour (Agee 2004). For the MSdk subzone, low- and moderate-severity fires were  75  observed in the fire history of the complex stands, while high-severity fire can be attributed to 24% of the landscape which is dominated by lodgepole pine (relative composition of pine >50%).  Change in Quantity Forest harvesting practices do not appear to be having a negative effect on the abundance of complex, mixed conifer forests on the landscape. Complex stands, for the purpose of sampling, were only included in the sample population if they had not been logged between 1950 and 1999. To assess the influence of logging on these stands for the Invermere TSA, the total area of complex stands, regardless of logging history, was calculated as 49,273ha. When the Timber Supply Review (2003) data on logging history is considered, this area is reduced by 8,116ha. This represents a decrease in the quantity of complex stands within the MSdk subzone of 4% over the period of time for which spatial data on logging history exists (-1950 to 2003). Shifts in disturbance regimes prior to modern forest management may have had a negative effect on the abundance of complex stands. Fire frequency was found to be at its highest during the period of European settlement. Increased fire frequency and fire severity across the landscape as a result of settlement and mining exploration may have reduced the percentage of complex stands on the landscape. This hypothesis could be tested by reconstructing past stand dynamics using the CWD characteristics of stands that burned with a high-severity in the settlement era. Further knowledge on the landscape patterns of complex stands is needed to support the statement that forest management is not negatively affecting these stands. Patch size distribution is important to the NDTs, as patch size targets are often used in forest management. Small to medium patches will constitute the majority of the complex stands within the MSdk subzone. Established patch size targets are similar to the recent history of patch size distribution, as the  76  targets are developed from a landscape level GIS analysis which based on forest structure within the last 20years. Maintaining the abundance of these stands is important, but this should be linked to their size and distribution across the landscape. Fragmentation and a reduction in area of individual patches containing complex stands may reduce ecological resilience in many ways, including increased susceptibility to high-severity fire. Analysis of neighbouring stands may provide an indication  Variability of Fire and Drivers Inherent in the definition of mixed-severity fire regimes is a dynamic combination of both spatial and temporal variation; this was apparent in the fire history of complex mixed-conifer stands of the MSdk subzone. Eighteen of 20 plots had visible fire scars and varied considerably in their history of fire frequency and length of the fire recording period. Within-plot variability in the location and type of fire-scarred trees suggested that fire intensity and/or spread was variable as not all trees recorded the same fire. Between plots, the fire recording period varied considerably, with the shortest fire records providing indirect evidence of high-severity fire at some plots that burned relatively recently, killing recorder trees and consuming fire-scarred coarse woody debris (Ehle and Baker 2003). Determining the set of physical and ecological conditions and processes that drive the observed spatial and temporal variability in fire is critical to the successful implementation of NDBM. I found that elevation was the strongest influence on fire frequency; lower elevation plots had a higher fire frequency than higher elevation plots. Slope angle and solar radiation were also significant in influencing fire frequency, although their.effects were not as strong as elevation and the trends were not intuitive. Surprisingly, aspect and latitude of the plots did not have a significant influence fire frequency. Temporally, the fire frequency of fire did change in association with  77  changes in land-use (Figure 3.8). Notably, the current fire suppression era has resulted in an increased mean number of years between fires at the plot and landscape scales. What does this mean to forest management? The major take-home message from these results is variability.  Variability in fire occurrence and frequency, and consequently, in resultant  stand structure, was evident at the plot, sub-regional and regional scales. A principal focus of NDBM should be the retention of variability at all three scales. A second message relates to the. frequency of fire within these stands; modeled plot-level fire intervals ranged from 18 to 111 years. Fire was disturbing these stands more frequently than previously classified by NDT. The structural characteristics of these complex stands in the MSdk subzone resulted from frequent, low- to moderate-severity disturbance. The use of patch size targets to manage these stands may retain their distribution on the landscape; however, the characteristics of these stands may be shifting as modern disturbances do not occur as frequently or with low- to moderate-severity.  Past Severity of Fire Understanding the role of fire severity in creating and maintaining the structure, age distribution and coarse woody debris (CWD) characteristics of these stands is critical to the implementation of NDBM. Retention of ecological structures as described in NDBM fits well with managing for resilience (Drever et al. 2006) and fire severity appears to have played key role in creating this structure. Other studies in western North America have found tree death and subsequent regeneration occur within 20 years of low-severity fire, illustrating the strong link between fire and forest dynamics (Ehle and Baker 2003). In this study, low- to moderate-severity fires likely resulted in some tree mortality and the creation of growing space for regeneration. This assertion is supported by the result that in 16 of 18 stands in which fire was observed, at least one sampled fire-scarred tree regenerated within 20 years of an earlier fire. Further research is ongoing  78  to elaborate on the specific role of fire mortality, creation of CWD, and regeneration. In conjunction with fire history data, complete age cohort data will be used to make the link between structure and process.  Restoration Ecosystem restoration and hazard mitigation as a response to the build-up of forest fuel after decades of fire suppression is currently a hot topic in BC forest ecology. In an audit following the unprecedented 2003 forest fire season, the southern Rocky Mountain Trench was singled out as an area with particularly severe fuel build-up (Filmon 2004). An important question is whether or not the complex forests of the MSdk subzone are in need of forest restoration. At a stand level, comparing time since last fire to the range of fire intervals in the historic record has been suggested as one way to document changes in fire regimes (Swetnam et al. 1999). In my study area, half of the stands currently have fire free intervals that are shorter than the maximum fire interval recorded in the plot and are within the historic range of variability (Figure 4.1). This suggests that the fire regime has been altered on only part of the landscape, making it difficult to determine  where  and  how ecological  restoration  and/or fuels  mitigation  are needed.  79  160 140  *  1 2 0  S  100  <D  60  c  H  +  +  P14  P18  20 40  4-  oH i P04  r  P05  * P07  P09  P11  P13  P20  P21  I  |  |  P01  P02  P06  "T—:  P10  1  P12  T  P15  P17  P19  Figure 4.1 Historic range of fire intervals (gray bars) for 18 sampled plots in the montane forests of southeastern British Columbia. Crosses indicate the time since last fire at each plot  Fuel build-up resulting from fire suppression threatens ecological resilience the most in areas where historic fires burned frequently (Schoennagel et al. 2004). Stands with the greatest need for restoration treatments would have a small range of fire intervals with a time since last fire that exceeds this range. At the same time, it is important to consider individual stands in the context of adjacent stands in the landscape.  For example, an individual stand may have a  relatively short fire-free interval, but still be susceptible to severe fires because it is surrounded by stands with long fire-free intervals in which fuels have accumulated. In mixed-severity regimes, heterogeneity of stands was maintained through variability in fire behaviour. If fire suppression is resulting in increased homogeneity of fuels, then future fires will display homogeneity in fire behaviour. Ecological resilience will likely fail in this situation as forests transition from stands with complex structures to even-aged stands with simpler structures. I suggest that effort to restore ecological resilience be focused on stands at lower elevations and those located in the valley bottom of the Rocky Mountain Trench. In mid-elevations in the montane forests, I believe that fire should be returned to landscape either by prescribing fire, allowing natural fires to burn or through active management of natural fires. Fire needs to occur at times of the year when fire hazard is less than extreme and fires can burn at low- to moderate severities. Currently, the majority of fires are burning under extreme conditions, when fire suppression is failing. The effect of these severe stand-replacing fires is to convert complex forests into much simpler ones.  Conclusions Complex stand structure composes a significant portion of the MSdk subzone in southeastern BC and forest fires were a much more frequent disturbance within these stands than  81  previously thought. Variability in fire behaviour has created these stands as part of a historic mixedseverity fire regime. Forest managers should be aware of the characteristics of this fire regime when considering how to manage for the future of this structure on the landscape. Awareness of the frequency and severity of the fires is important, but so is the variability in fire between and within stands. There is no doubt that fire played a significant role in creating this complex structure and it can be concluded that this disturbance regime was critical in creating the valuable habitat that is associated with this structure. Management strategies that do not consider protecting, restoring and creating this structure will result in a reduced ecological resiliency and a high likelihood of habitat loss.  82  Literature Cited  Agee, J.K. 1993. Fire Ecology of the Pacific Northwest Forests. Island Press, Washington, D. C.  Agee, J.K. 2004. The complex nature of mixed severity fire regimes. In Mixed Severity Fire Regimes: Ecology and Management. Edited by I Taylor, J. Zelnik, S. Cadwallader, and B. Hughes, Spokane, WA.  Albini, F.A. 1992. Dynamics and Modeling of Vegetation Fires: Observations. In Fire in the Environment: The Ecological, Atmospheric, and Climatic Importance of Vegetation Fires. Edited by P.J. Crutzen, and J. G. Goldammer. John Wiley & Sons, New York.  Arienti, M.C, Cumming, S.G., and Boutin, S. 2006. Empirical models of forest fire initial attack success probabilities: the effects of fuels, anthropogenic linear features, fire weather, and management. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 36(12): 3155-3166.  Arno, F.S., and Sneck, K.M. 1977. A method for determining fire history in coniferous forests of the mountain west. USDA Forest Service General Technical Report INT-42.  Arno, S.F., and Fiedler, CE. 2005. Mimicking nature's fire: restoring fire-prone forests in the West. Island Press, Washington. >  Arno, S.F., Parsons, D.J., and Keane, R.E. 2000. Mixed-severity fire regimes in the northern Rocky Mountains: consequences of fire exclusion and options for the future. In Wilderness science in a time of change conference - Volume 5 - Wilderness ecosystems, threats, and management. Edited by D. N. Cole, S. F. McCool, W. T. Borrie, and J. O'Loughlin, Missoula, MT. pp. 225-232.  Baisan, C.H., and Swetnam, T.W. 1990. Fire history on a desert mountain range: Rincon Mountain Wilderness, Arizona, U.S.A. Canadian Journal of Forest Research 20:1559-1569.  Baker, W.L., and Ehle, D. 2001. Uncertainty in surface-fire history: the case of ponderosa pine forests in the western United States. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 31(7): 1205-1226.  Baker, W.L., Veblen, T.T., and Sherriff, R.L. 2007. Fire, fuels and restoration of ponderosa pineDouglas fir forests in the Rocky Mountains, USA. Journal of Biogeography 34(2): 251-269.  83  Bancroft, H. 1890. History of British Columbia, 1792-1887. History Co., San Francisco.  Barrett, S.W., and Arno, S.F. 1982. Indian Fires as an Ecological Influence in the Northern Rockies. Journal of Forestry 80(10): 647-651.  BCMOF (British Columbia Ministries of Forests). 1995. Biodiversity Guidebook, Forest Practices Code of British Columbia. BC Ministry of Forests and Water, Land and Air Protection, Victoria, BC. www.for.gov.bc.ca/tasb/legsregs/fpc/fpcguide/biodiv/biotoc.htm  BCMOFR (British Columbia Ministry of Forests and Range). 2007. Historical Fire Database. BC Ministry of Forests and Range, Protection Branch, Invermere, British Columbia.  Beaty, R.M., and Taylor, A.H. 2001. Spatial and temporal variation of fire regimes in a mixed conifer forest landscape, Southern Cascades, California, USA. Journal of Biogeography 28(8): 955-966.  Bergeron, Y., Leduc, A., Harvey, B.D., and Gauthier, S. 2002. Natural fire regime: A guide for sustainable management of the Canadian boreal forest. Silva Fennica 36(1): 81-95.  Braumandl, T.F. (ed). 1992. A Field Guide for Site Identification and Interpretation for the Nelson Forest Region. B.C. Ministry of Forests Research Branch, Victoria, B.C.  Brown, A.A., and Davis, K.P. 1973. Forest Fire: control and use. 2nd ed. McGraw Hill Book Co, New York NY.  Brown, P.M., Kaufmann, M.R., and Shepperd, W.D. 1999. Long-term, landscape patterns of past fire events in a montane ponderosa pine forest of central Colorado. Landscape Ecology 14(6): 513-532.  Brown, P.M., Ryan, M.G., and Andrews, T.G. 2000. Historical surface fire frequency in ponderosa pine stands in research natural areas, Central Rocky Mountains and Black Hills, USA. Natural Areas Journal 20(2): 133-139.  Brown, P.M., and Shepperd, W.D. 2001. Fire history and fire climatology along a 5° gradient in < latitude in Colorado and Wyoming, USA. Palaeobotanist 50:133-140.  84  Buechling, A., and Baker, W.L. 2004. A fire history from tree rings in a high-elevation forest of Rocky Mountain National Park. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 34(6): 1259-1273.  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.  Conover, W.J. 1999. Practical nonparametric statistics. Third ed. John Wiley & Sons, Inc., New York.  Covington, W.W., and Moore, M.M. 1994. Southwestern ponderosa forest structure: changes since Euro-American settlement. Journal of Forestry 92(1): 39-47.  Daniels, L.D. in review. A mixed-severity fire regime influenced by ENSO in the Cariboo region of British Columbia.. Canadian Journal of Forest Research.  Daniels, L.D., Cochrane, J.D., and Gray, R.W. 2006. Refining Mixed-Severity Fire Regimes in the Rocky Mountain Forest District.. Report to Jembec Inc., BC Division, Canadian Forest Products Ltd., Radium Hot Springs, and the Forest Investment Account of British Columbia.  Daniels, L.D., Cochrane, J.D., and Gray, R.W. 2007. Mixed-severity fire regimes: regional analysis of the impacts of climate on fire frequency in the Rocky Mountain Forest District. Report to Tembec Inc., BC Division, Canadian Forest Products Ltd., Radium Hot Springs, and the Forest Investment Account of British Columbia.  Daniels, L.D., and Watson, E. 2003. Climate-fire-vegetation interactions in the Cariboo Forests: A dendrochronological analysis. Forest Innovation and Investment Forest Research Program Final Report.  Delong, S.C, Daniels, L.D., Heemskerk, B., and Storaunet, K.0.2005. Temporal development of decaying log habitats in wet spruce-fir stands in east-central British Columbia. Canadian Journal of Forest Research 35(12): 2841-2850.  Dieterich, J.H., and Swetnam, T.W. 1984. Dendrochronology of a Fire-Scarred Ponderosa Pine. Forest Science 30(1): 238-247.  85  Drever, R.C., Peterson, G., Messier, C , Bergeron, Y., and Flannigan, M. 2006. Can forest management based on natural disturbances maintain ecological resilience? Canadian Journal of Forest Research 36:2285-2299.  Dubayah, R., and Rich, P.M. 1995. Topographic Solar-Radiation Models for Gis. International Journal of Geographical Information Systems 9(4): 405-419.  Ehle, D.S., and Baker, W.L. 2003. Disturbance and stand dynamics in ponderosa pine forests in Rocky Mountain National Park, USA. Ecological Monographs 73(4): 543-566.  Everett, R., Schellhaas, R., Ohlson, P., Spurbeck, D., and Keenum, D. 2003. Continuity in fire disturbance between riparian and adjacent sideslope Douglas-Fir forests. Forest Ecology and Management 175(1-3): 31-47.  Fall, J. 1998. Reconstructing the historical frequency of fire: A modeling approach to developing, and testing methods. In Master of Natural Resource Management thesis. Simon Fraser University, Burnaby, British Columbia, p. 93.  Filmon, G. 2004. Firestorm 2003: Provincial Review. Available from www.2003firestorm.qov.bc.ca/  Franklin, J.F., Shugart, H.H., and Harmon , M.E. 1987. Tree death as an ecological process. Bioscience 37(8): 550-556.  Fule, P.Z., Covington, W.W., and Moore., M.M. 1997. Determining reference conditions for ecosystem management of southwestern ponderosa pine forests. Ecological Applications 7(3): 895-908.  Gray, R.W., Riccius, E., and Wong, C. 1998. Comparison of current and historic stand structure in 2 IDFdm2 sites in the Rocky Mountain Trench. Report to BC Ministry of Forests.  Grissino-Mayer, H.D. 2001a. Evaluating crossdating accuracy: a manual for the computer program COFECHA. Tree-Ring Research 57(2): 205-221.  Grissino-Mayer, H.D. 2001b. FHX2 - Software for analyzing temporal and spatial patterns in fire regimes from tree rings. Tree-Ring Research 57(1): 115-124.  Gutsell, S.L., and Johnson, E.A. 1996. How fire scars are formed: coupling a disturbance process to its ecological effect. Canadian Journal of Forest Research 26:166-174.  86  Hansen, A.J., Spies, T.A., Swanson, F.J., and Ohmann, J.L 1991. Conserving biodiversity in managed forests. Bioscience 41(6): 382-392.  Hare, F.K., and Thomas, M.K. 1979. Climate Canada. 2nd ed. John Wiley & Sons Toronto, O.N.  Heyerdahl, E.K., Brubaker, L.B., and Agee, J.K. 2001. Spatial controls of historical fire regimes: A multiscale example from the interior west, USA. Ecology 82(3): 660-678. Heyerdahl, E.K., Brubaker, L.B., and Agee, J.K. 2002. Annual and decadal climate forcing of historical fire regimes in the interior Pacific Northwest, USA. Holocene 12(5): 597-604.  Heyerdahl, E.K., Lertzman, K., and Karpuk, S. 2007. Local-scale controls of a low-severity fire regime (1750-1950), southern British Columbia, Canada. Ecoscience 14(1): 40-47.  Heyerdahl, E.K., and McKay, S.J. 2001. Condition of live fire-scarred ponderosa pine trees six years after removing partial cross sections. Tree-Ring Research 57(2): 131-139.  Johnson, E.A. 1979. Fire recurrence in the subartic and its implications for vegetation composition. Canadian Journal of Botany 57:1374-1379.  Johnson, E.A., and Gutsell, S.L. 1994. Fire frequency models, methods and interpretations. Advances in Ecological Research 25:239-285.  Kauffman, J.B., and Martin, R.E. 1990. Sprouting Shrub Response to Different Seasons and Fuel Consumption Levels of Prescribed Fire in Sierra-Nevada Mixed Conifer Ecosystems. Forest Science 36(3): 748-764.  Kobziar, L, Moghaddas, J., and Stephens, S.L. 2006. Tree mortality patterns following prescribed fires in a mixed conifer forest. Canadian Journal of Forest Research 36(12): 3222-3238.  Kou, X.J., and Baker, W.L. 2006. Accurate estimation of mean fire interval for managing fire. International Journal of Wildland Fire 15(4): 489-495.  Landres, P.B., Morgan, P., and Swanson, F.J. 1999. Overview of the use of natural variability concepts in managing ecological systems. Ecological Applications 9(4): 1179-1188.  Lertzman, K., Fall, J., and Dorner, B. 1998. Three kinds of heterogeneity in fire regimes: At the crossroads of fire history and landscape ecology. Northwest Science 72:4-23. (  87  Lucas, B. 2007. Personal Communication on Historic First Nations Land Use. Edited by J.D. Cochrane.  Manning, T., Bradford, P., White, C , Rowe, D, Densmore, N. & Guy, S. 2002. British Columbia dangerous tree assessment process. In Proceedings of the symposium on the ecology and management of dead wood in western forests, W.F. Laudenslayer, P.J. Shea, B.E. Valentine, CP. Weatherspoon and T.E. Lisle (editors). U.S. Department of Agriculture Forest-Service, Albany, California. General Technical Report. PSW-GTR-181:863-868. http://www.fs.fed.us./psw/publications/documents/gtr-181/064_Manning.pdf  Mcalpine, R.S., Lawson, B.D., and Taylor, E. 1991. Fire spread across a slope. In 11th Conference on Fire and Forest Meteorology Missoula, MT.  McBride, J.R., and Laven, R.D. 1976. Scars as an indicator of fire frequency in San Bernadino Mountains, California. Journal of Forestry 7:439-442.  Meidinger, D., and Pojar, J. 1991. Ecosystems of British Columbia. B.C. Ministry of Forests Research Branch, Victoria, B.C.  Moore, M.M., Covington, W.W., and Fule, P.Z. 1999a. Reference conditions and ecological restoration: a soutwestern ponderosa pine perspective. Ecological Application 9(4): 12661277.  Moore, T.R., Trofymow, J.A., Taylor, B., Prescott, C , Camire, C , Duschene, L, Fyles, J., Kozak, L, Kranabetter, M., Morrison, I., Siltanen, M., Smith, S., Titus, B., Visser, S., Wein, R., and Zoltai, S. 1999b. Litter decomposition rates in Canadian forests. Global Change Biology 5(1): 75-82.  Peterson, G., Allen, C.R., and Holling, C.S. 1998. Ecological resilience, biodiversity and scale. Ecosystems 1998(1): 6-18.  Pickett, S.T.A., and White, P.S. 1985. The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, Orlando, FL.  Pojar, J., Klinka, K., and Meidinger, D.V. 1987. Biogeoclimatic ecosystem classification in British Columbia. Forest Ecology and Management 22:119-154.  Pyne, S.J., Andrews, P.L., and Laven, R.D. 1996. Introduction to Wildland Fire, second ed. John Wiley & Sons, New York.  88  Reed, W.J. 1994. Estimating the historic probability of stand-replacement fire using the age-class distribution of undisturbed forest. Forest Science 40(1): 104-119.  Schoennagel, T., Veblen, T.T., and Romme, W.H. 2004. The interaction of fire, fuels, and climate across rocky mountain forests. Bioscience 54(7): 661-676.  Smith, K.T., and Sutherland, E.K. 2001. Terminology and biology of fire scars in selected central hardwoods. Tree-Ring Research 57(2): 141-147.  Society, G.D.H. 1972. Kinbasket country: the story of Golden and the Columbia Valley. Golden & District Historical Society, Golden, B.C.  Spurr, S.H., and Barnes, B.V. 1980. Forest Ecology. 3rd ed. John Wiley and Sons Inc., New York, NY.  Stokes, M.A., and Smiley, T.L. 1968. An introduction to tree ring dating, University of Chicago Press, Chicago, Illinois, USA.  Swetnam, T.W., Allen, CD., and Betancourt, J.L. 1999. Applied historical ecology: Using the past to manage for the future. Ecological Applications 9(4): 1189-1206.  Swetnam, T.W., and Betancourt, J.L. 1990. Fire - Southern Oscillation Relations in the Southwestern United-States. Science 249(4972): 1017-1020.  Taylor, A.H. 2000. Fire regimes and forest changes in mid and upper montane forests of the southern Cascades, Lassen Volcanic National Park, California, USA. Journal of Biogeography 27(1): 87-104.  Taylor, A.H., and Skinner, C.N. 1998. Fire history and landscape dynamics in a late-successional reserve, Klamath Mountains, California, USA. Forest Ecology and Management 111(2-3): 285-301.  Taylor, A.H., and Skinner, C.N. 2003. Spatial patterns and controls on historical fire regimes and forest structure in the Klamath Mountains. Ecological Applications 13(3): 704-719.  Trofymow, J.A., Moore, T.R., Titus, B., Prescott, O, Morrison, I., Siltanen, M., Smith, S., Fyles, J., Wein, R., CamirT, O, Duschene, L, Kozak, L., Kranabetter, M., and Visser, S. 2002.  89  Rates of litter decomposition over 6 years in Canadian forests: influence of litter quality and climate. Canadian Journal of Forest Research 32(5): 789-804.  USDA. 1960. The True Story of Smokey Bear. Edited by U.S. Department of Agriculture. Western Publishing Company Inc.  Valentine, K.W.G., Sprout, P.N, Baker, T.E, and Lavkulich, LM. (eds). 1978. The Soil Landscapes of British Columbia. B.C. Ministry of Environment, Victoria, B.C.  Van Home, M l , and Fule, P.Z. 2006. Comparing methods of reconstructing fire history using fire scars in a southwestern United States ponderosa pine forest. Canadian Journal of Forest Research 36(4): 855-867.  Van Wagner, CE. 1978. Age-class distribution and the forest fire cycle. Canadian Journal of Forest Research 8: 220-227.  Van Wagner, CE. 1987. Development and structure of the Canadian Forest Fire Weather Index System. Canadian Forestry Service.  Veblen, T.T, Kitzberger, T , and Dunnegan, J. 2000. Climatic and human influences on fire regimes in ponderosa pine forests in the Colorado Front Range. Ecological Applications 10(4): 1178-1195.  Westerling, A.L, Hidalgo, H.G, Cayan, D.R, and Swetnam, T.W. 2006. Warming and earlier spring increase western US forest wildfire activity. Science 313(5789): 940-943.  White, D. 1988. Fort Steele. Heritage House Pub. Co, Surrey, B.C.  Wildlife Tree Committee of British Columbia. 2005. Wildlife/danger tree assessor's course workbook - forest harvesting and silviculture course module. Victoria B.C.: British Columbia Ministry of Forests and Range, WorkSafeBC, Ministry of Environment. Revised November 2005. http://www.for.qov.bc.ca/hfp/values/wildlifeAA/LT/index.htm  Wolf, J.J, and Mast, J.N. 1998. Fire history of mixed-conifer forests on the North Rim, Grand Canyon National Park, Arizona. Physical Geography 19(1): 1-14.  90  Wong, C.W, Dorner, B, and Sandmann, H. 2003. Determining disturbance variability: a review, a catalogue of methods and a demonstration. Report to BC Ministry of Forests Research Branch.  WorkSafe BC. 2006. Occupational Health and Safety Regulations. http://www2.worksafebc.com/Portals/Forestry/Home.asp http://www2.worksafebc.com/Publications/OHSRequlation/Home.asp  Wright, C S , and Agee, J.K. 2004. Fire and vegetation history in the eastern Cascade Mountains, Washington. Ecological Applications 14(2): 443-459.  Yamaguchi,,D.K. 1991. A simple method for cross-dating increment cores from living trees. Canadian Journal of Forest Research 21:414-416.  91  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0100776/manifest

Comment

Related Items