Influences of an Active Spruce Beetle Outbreak on Fire Severity in Spruce-Fir Forests in Southern Colorado

Transcription

1 University of Colorado, Boulder CU Scholar Geography Graduate Theses & Dissertations Geography Spring Influences of an Active Spruce Beetle Outbreak on Fire Severity in Spruce-Fir Forests in Southern Colorado Robert Anton Andrus University of Colorado Boulder, Follow this and additional works at: Part of the Forest Biology Commons, Forest Management Commons, Physical and Environmental Geography Commons, Remote Sensing Commons, and the Terrestrial and Aquatic Ecology Commons Recommended Citation Andrus, Robert Anton, "Influences of an Active Spruce Beetle Outbreak on Fire Severity in Spruce-Fir Forests in Southern Colorado" (2015). Geography Graduate Theses & Dissertations This Thesis is brought to you for free and open access by Geography at CU Scholar. It has been accepted for inclusion in Geography Graduate Theses & Dissertations by an authorized administrator of CU Scholar. For more information, please contact

2 INFLUENCES OF AN ACTIVE SPRUCE BEETLE OUTBREAK ON FIRE SEVERITY IN SPRUCE-FIR FORESTS IN SOUTHERN COLORADO By Robert Anton Andrus B.A., Willamette University, 2007 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirement for the degree of Masters of Arts Department of Geography 2015

3 This thesis entitled: Influences of an active spruce beetle outbreak on fire severity in spruce-fir forests in southern Colorado Written by Robert Anton Andrus has been approved for the Department of Geography by Thomas T. Veblen Jennifer Balch Date The final copy of this thesis has been examined by the signatories, and we Find that both the content and the form meet acceptable presentation standards Of scholarly work in the above mentioned discipline.

4 iii Andrus, Robert Anton (M.A. Geography) Influences of an active spruce beetle outbreak on fire severity in spruce-fir forests in southern Colorado Thesis directed by Professor Thomas T. Veblen Recent large and severe outbreaks of native bark beetles has raised concern among the general public and land managers about potential risks of amplified fire activity in western North America. However, there are no field studies of the effects of the severity of spruce beetle (Dendroctonus rufipennis Kirby) infestation on subsequent fire severity in subalpine Engelmann spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa) forests. We collected extensive field data in subalpine forests in the eastern San Juan Mountains, southwestern Colorado, to investigate whether fire severity changed across a range of severity of early, gray-stage spruce beetle infestations. Using correlation and multivariate generalized linear regression analyses, we found that pre-fire spruce beetle severity was not correlated with or a significant predictor of most field measurements of fire severity or a remotely sensed estimate of fire severity under moderate and extreme burning conditions. However, under moderate burning conditions only, the number of fire-killed stems and fire-killed basal area were weakly but positively correlated with the severity of beetle infestation. In comparison to severity of the pre-fire beetle outbreak, we found that topography, live and dead basal area at time of fire, and short-term weather conditions exerted a stronger effect on fire severity. This suggests the beetle outbreak and fire were either independent of one another or factors unrelated to beetle outbreak overrode any effects of the outbreak on fire activity. This finding of no effect agrees with retrospective studies using remotely sensed products to examine fire activity following other spruce beetle outbreaks, which similar to this study, concluded that the overriding influence of climate is essential for creating conditions conducive to large, high-severity fires in the subalpine zone of Colorado.

5 iv ACKNOWLEDGEMENTS Research was funded by the Collaborative Research: Spruce Beetle and Wildfire Interactions under Varying Climate in the Colorado Rockies award (# ) from the National Science Foundation. Funding support was also received from the Colorado Mountain Club s Kurt Gerstle Fellowship and a Team Grant from the University of Colorado s Undergraduate Research Opportunities Program. Special thanks to Thomas Veblen, Brian Harvey, and the Biogeography Lab for their guidance in project development, analysis, and comments on this thesis. For field assistance and data processing, I would like to recognize Carah Bordner and Steven Fiske for their patience, flexibility, and positive attitude in the field. I also thank my committee members Jennifer Balch and Stefan Leyk for their support.

6 v TABLE OF CONTENTS CHAPTER 1: FIRE AND SPRUCE BEETLE INTERACTIONS: CURRENT STATE OF KNOWLEDGE... 1 INTRODUCTION... 1 BACKGROUND... 2 Spruce-fir forests... 2 Disturbance ecology in subalpine forests... 2 RESEARCH GOALS... 4 CHAPTER 2: INFLUENCES OF AN ACTIVE SPRUCE BEETLE OUTBREAK ON FIRE SEVERITY IN SPRUCE-FIR FORESTS IN SOUTHERN COLORADO... 6 INTRODUCTION... 6 METHODS Study Area Sampling design Determining the effects of spruce beetle and fire on stand structure Pre-fire beetle outbreak severity assessment Fire severity assessment Topography and solar radiation Assessing burning conditions Burning condition and RdNBR overlay analysis Determining the effects of pre-fire spruce beetle outbreak severity on fire severity RESULTS Effects of spruce beetle and fire on stand structure Burning Condition and RdNBR overlay analysis Effects of pre-fire spruce beetle outbreak severity on fire severity DISCUSSION Burning conditions and RdNBR overlay analysis Effects of pre-fire spruce beetle outbreak severity on fire severity Contribution and limitations CONCLUSION REFERENCES APPENDIX Appendix A: Time series of spruce beetle caused tree mortality in Colorado Appendix B: Evidence and criteria for tree classification Appendix C: Moderate and extreme burning conditions classification Appendix D: Spearman correlation analysis of fire severity metrics... 45

7 vi LIST OF TABLES Table 1: Assessment of landscape scale fire severity (RdNBR) and hectares burned by burning condition from the burning condition and RdNBR overlay analysis Table 2: Spearman Correlation (rho) and p-values ((*): P <0.05, (.): P <0.10) for each fire severity response variable and percent of basal area killed by spruce beetle in each plot across all fires. The number of plots in each burning condition and the range of percent beetle killed basal area sampled are listed below Table 3: Generalized linear model results from the fire severity models used to compare the beetle outbreak severity coefficient and significance to the following non-beetle related predictor variables: pre-fire total basal area (live and dead basal area of all species at time of fire) and slope position (relative position on slope). The fine fuel percentage is the percent of the bole with fuel less than a quarter inch remaining. A positive beta increases fire severity and negative beta decreases fire severity

8 vii LIST OF FIGURES Figure 1: Conceptual framework of (a) fuels characteristics and (b) fire behavior relative to preoutbreak conditions for red, gray, and old (snagfall and regrowth) phases. Surface fire properties include reaction intensity, rate of spread, and flame length. For postoutbreak phases, solid lines indicate higher confidence in responses based on Fig. 3, and dashed lines indicate lower confidence (more disagreement, fewer studies, or knowledge gaps). (Fig. 2 from Hicke et al. 2012) Figure 2: Location of the five subalpine fires studied and the two weather stations used to assess burning conditions in the southern Colorado Rocky Mountains. All fires burned in the summer of 2013 with the exception of the Little Sands Fire that burned in The inset displays the location of the study area within Colorado Figure 3 (A-D): Stand structure characteristics prior to spruce beetle outbreak and fire, after spruce beetle outbreak, and after spruce beetle outbreak and fire for all plots (n=143). Preoutbreak and pre-fire characteristics were reconstructed with the criteria presented in Appendix B. All values are mean averages and 95% confidence intervals Figure 4 (A-G): Fire severity measures plotted against spruce beetle killed basal area under moderate and extreme burning conditions (moderate = circle, extreme = triangle) in an early, gray-stage spruce beetle outbreak in subalpine, spruce-fir forests. RdNBR stands for Relative differenced Normalized Burn Ratio. Percent fine fuel measures the remaining fine fuel (< 1cm) attached to the tree bole. BA stands for basal area

9 1 CHAPTER 1: FIRE AND SPRUCE BEETLE INTERACTIONS: CURRENT STATE OF KNOWLEDGE INTRODUCTION From a local to a global scale, Earth s patterns and processes are changing. Documented shifts in climate, land-cover patterns, and resource depletion have been attributed to direct and indirect influences of human consumption and population. The complex interaction between humans and the environment has hampered the ability to predict the scale and timing of feedbacks on and changes to ecosystem structure and function (National Research Council 2013). Disturbances are a component of ecosystem function and an essential process that is linked to climate, land use patterns, and human activity, and thus will likely shift in magnitude and timing with trends in atmospheric warming. Human-caused and natural disturbance events (e.g. logging and wildfire) quickly alter ecosystems, which may expedite shifts in ecosystem state and function if climate change sufficiently affects ecosystem resilience or if human activity continually limits an ecosystem s ability to rebound from change. Thus, research is needed to explore how and when disturbances significantly alter ecological systems, which will help inform management strategies and address potential feedbacks on social systems (Turner 2010). In this thesis, I examine a disturbance interaction to inform the public, land managers, and scientists about the relative influence of a spruce beetle outbreak severity and non-beetle related factors on wildfire severity in the subalpine zone in Colorado.

10 2 BACKGROUND Spruce-fir forests Engelmann Spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa) (spruce-fir) forests in Colorado inhabit the subalpine zone ( m) of the Rocky Mountains (Romme et al. 2009). In the absence of disturbance, the more shade-tolerant and long-lived Engelmann spruce will dominate stands of spruce-fir. Other species known to inhabit the subalpine zone are aspen (Populus tremuloides), lodgepole pine (Pinus contorta) in the north and central Colorado Rocky Mountains, and in lower elevation sites, white fir (Abies concolor), blue spruce (Picea pungens), and limber pine (Pinus flexilis) are present (Peet 2000). In these subalpine forests, temperatures are moderated by elevation and moisture. Snow during the winter commonly persists late into the spring and monsoon rains during the summer help maintain a cool and wet climate year-round. Tree growth rates are slow because long and harsh winters result in a short growing season and soils are poorly developed. Disturbance ecology in subalpine forests Avalanches, blowdown from windstorms, bark beetle outbreaks, wildfire, and human influences (e.g. logging) are the primary disturbances that effect subalpine forests. These disturbances operate at different spatial and temporal scales to alter forest composition and structure, spatial heterogeneity in forest ecosystems, and future disturbance patterns. Avalanches operate on an annual basis to disrupt vegetation successional pathways and remove biomass, which maintains a relatively young stage age and discontinuous fuel bed in avalanche chutes (Veblen et al. 1994). Blowdown events range in size from the individual tree to the landscape (15 hectares) scale (Veblen et al. 1989) and create large quantities of dead fuel, which has been shown to elevate fire severity (Kulakowski and Veblen 2007).

11 3 Large wildfire events and epidemic-level spruce beetle outbreaks are infrequent (i.e. much greater than 100 years, Veblen et al. 1994), but affect forest ecosystems on broad scales in the subalpine forests of the southern Rocky Mountains (Veblen et al. 1994, Buechling and Baker 2004). A cool and wet climate in the subalpine zone reduces the threat of frequent fire and generates large quantities of biomass, which produce large, high severity fire events when annual drought and short-term weather conducive to large fire development coincide (Sibold et al. 2006, Schoennagel et al. 2007). Though fire suppression is practiced in the subalpine zone, the longintervals between historic fires and dependence on infrequent, extreme drought imply that the 60 to 80 years of active fire suppression have not greatly altered the natural fire regime of the subalpine zone (Sherriff et al. 2001, Sibold et al. 2006). Further, the remote location and steep terrain of spruce-fir forests and the extreme fire behavior produced by the weather conditions present during large fire events (e.g. West Fork complex) has probably limited the effectiveness of suppression activities. While subalpine fire events commonly kill most trees within the fire perimeter, spruce beetle outbreaks preferentially attack larger diameter spruce, which alters forest composition and structure (Veblen et al. 1994). In Colorado, spruce beetles are a native forest insect that primarily infest Engelmann spruce and blue spruce by burrowing into the inner bark to feed and reproduce in the phloem tissues (Schmid and Frye 1977). Spruce beetles intercept the flow of essential nutrients within a tree, which can kill a tree if a sufficient number of beetles are present to girdle the tree. Localized (endemic) beetle populations are primarily supported by weakened or fallen trees (Schmid 1981), while widespread (epidemic) spruce beetle outbreaks are known to be initiated by a 1) large population of mature host trees (Schmid and Frye 1977), 2) blowdown or logging activity (Schmid 1981), and/or 3) climate (drought) (Christiansen et al. 1987, Hart et al.

12 4 2013). Once initiated, localized populations successfully attack more trees and accelerate reproduction, causing widespread tree mortality (Veblen et al. 1994). Since spruce beetles preferentially attack large diameter spruce, severe outbreaks kill most of the dominant canopy trees and transition canopy dominance to suppressed subalpine fir, spruce, and sometimes aspen (Schmid and Frye 1977, Veblen et al. 1994, DeRose and Long 2007). Both large wildland fire events and epidemic-level spruce beetle outbreaks are linked to climate, particularly warmer and drier periods (Bentz et al. 2010, Dennison et al. 2014). Thus, predicted trends in atmospheric warming across the western US (Cook et al. 2015) will likely increase the frequency of fire (Peterson and Littell 2012) and beetle infestation as well as increase the possibility of these two events overlapping in time and space. RESEARCH GOALS While several studies have examined the influence of spruce beetle severity on fire severity using remotely-derived datasets (Hart et al. in prep, Bigler et al. 2005, Kulakowski and Veblen 2007), this is the first study (co-authored with Thomas T. Veblen, Brian J. Harvey, and Sarah J. Hart) to assess this relationship with field measures of spruce beetle severity and fire severity. Capturing a range of both beetle severities and fire severities, we compare the relative importance of spruce beetle severity to non-beetle related factors in determining fire severity during a critical stage of a spruce beetle outbreak. This research builds upon methodologically similar work in other forest ecosystems (e.g. Harvey et al. 2013) and addresses uncertainty about how disturbance interactions and forest attributes affect fire severity in beetle-affected forest ecosystems. As fire suppression costs continue to rise and increasing financial resources are allocated to treat beetle-infested forests (USDA 2014), decision-making support for forest

13 5 ecosystems is needed to inform forest policy, land managers, citizens of the wildland-urban interface (WUI), and the greater research community.

14 6 CHAPTER 2: INFLUENCES OF AN ACTIVE SPRUCE BEETLE OUTBREAK ON FIRE SEVERITY IN SPRUCE-FIR FORESTS IN SOUTHERN COLORADO INTRODUCTION The coincidence of two historically infrequent subalpine disturbance events, an early, gray-stage spruce beetle (Dendroctonus rufipennis Kirby) outbreak and widespread fires, in subalpine forests of Engelmann spruce (Picea engelmannii) and subalpine fir (Abies lasiocarpa) in the San Juan Mountains of southwestern Colorado in provided a rare opportunity to examine the nature of interactions between these two disturbance types. In this context, it is useful to distinguish two types of disturbance interactions. Disturbances acting in quick succession may act synergistically to produce nonlinear or unpredictable changes in ecosystem structure and function, a concept termed compound disturbances (sensu Paine et al. 1998). As compound disturbances, spruce beetle outbreaks soon followed by fire can have important consequences for post-fire regeneration and resilience of subalpine forests of the Rocky Mountains (Kulakowski et al. 2013). The concept of linked disturbances (sensu Simard et al. 2011) focuses on whether a past disturbance alters the probability, extent, or severity of a subsequent disturbance. Thus, bark beetle outbreaks and wildfire may be linked disturbances if one disturbance type either enhances or dampens the effects, such as severity, of a subsequent occurrence of the other disturbance type. Alternatively, bark beetle outbreaks and wildfire events may act independently of one another or the effects of other biophysical factors may be so overriding that they mask any spatial or temporal connection between the two disturbances (Harvey et al. 2014b). In the context of adapting to increased bark beetle outbreaks and more extensive areas burned from warming temperatures and increased drought severity (Bentz et al.

15 7 2010, Dennison et al. 2014), management and policy decisions need to be informed by a better understanding of the nature and magnitude of any effects of prior spruce beetle outbreaks on subsequent wildfire severity. Widespread tree mortality from recent bark beetle outbreaks across western North American (Raffa et al. 2008, Meddens et al. 2012) has raised concern among the general public, land managers, and scientists about periods of amplified (positive feedback) fire activity (Hicke et al. 2012, Jenkins et al. 2012). To date, the majority of studies of potential links between bark beetle outbreaks and subsequent fire activity in the U.S. Rocky Mountains has focused on outbreaks of mountain pine beetle (Dendroctonus ponderosae) in lodgepole pine (Pinus contorta) (Hicke et al. 2012, Jenkins et al. 2014). The research protocols in these studies have included the application of fire behavior models using fuel inputs from chronosequences of bark beetle outbreaks (Jenkins et al. 2008, Schoennagel et al. 2012), overlay analyses of fire severity classes from satellite imagery with aerial detection survey data on previous bark beetle activity (Kulakowski and Veblen 2007), and rarely the incorporation of field-based measures of severities of both bark beetle outbreak and of fire (Harvey et al. 2013, 2014a). In Colorado the mountain pine beetle epidemic that initiated in c and affected 1.4 million hectares peaked in 2008 and by 2014 declined to only 1200 new hectares infested (USDA 2015). In contrast, the spruce beetle outbreak that initiated in c. 2001, affecting half a million hectares, has increased sharply since 2008 with over a 100,000 new hectares affected in 2014 (Appendix A). Thus, the potential linkage of fire severity to previous spruce beetle outbreak is becoming an increasing concern among the general public and policy makers in Colorado (Colorado State Forest Service 2012, Finley 2013). The overlap of the current outbreak of spruce beetle with several large wildfires in the San Juan Mountains of southwestern Colorado provides an important opportunity

16 8 to examine potential linkages between the severity of the spruce beetle outbreak and the severity of the subsequent fires. Large, severe fires are the characteristic fire type in subalpine forests of Colorado but they occur at relatively long intervals (i.e. much greater than 100 years, Veblen et al. 1994, Sibold et al. 2006) and few such fires have occurred in the past c. 100 years (Kulakowski and Veblen 2002, Sibold et al. 2006). Thus, there have been few opportunities to closely examine the spatial heterogeneity of fire severity in relation to pre-fire vegetation structure and other conditions. In the subalpine zone, climate, topography, forest composition and structure, and ecological legacies from previous disturbance interact in a complex manner to alter the spatial pattern in fire severity (Veblen et al. 1994, Kulakowski et al. 2003). Initiation and development of large (e.g. > 5000 ha) fire events in the mesic subalpine zone is rare because fire spread in these relatively cool, moist forests is dependent on extreme annual-scale drought and firepromoting short-term weather conditions (hot, dry, and windy weather) (Buechling and Baker 2004, Sibold et al. 2006, Schoennagel et al. 2007). Thus, lengthy fire-free periods and slow rates of decomposition allows ample time for high-levels of biomass to accumulate, which promotes extensive, stand-replacing fires under extreme fire weather conditions. Under non-drought conditions, fuel moisture typically remains high and fire activity is minimal. While climatic factors are an essential prerequisite for fire activity and a dominant driver of fire behavior and effects in subalpine ecosystems, differences in forest structural classes and composition created by disturbance legacies and topographic microclimates also alter fire severity. For example, aspen stands and younger structural classes of lodgepole pine and spruce-fir, which commonly regenerate after fire are less susceptible to high severity fire than mature stands of spruce-fir (Bigler et al. 2005). Thus, the impact of previous fire on vegetation can limit the spatial pattern

17 9 of subsequent fire severity (Kulakowski and Veblen 2007) and loss of large diameter trees from fire is known to reduce the likelihood of spruce beetle outbreak for decades (Schmid and Frye 1977, Veblen et al. 1994). However, only sparse evidence is available for a connection between widespread spruce beetle outbreak and fire severity, and such evidence is derived from both fire behavior modeling and from retrospective studies of actual fire events. Fire-behavior modeling studies based on fuel inputs from spruce beetle-infested forests suggest complex influences on subsequent fire intensity, which are dependent on the stage of beetle attack, the selected fire behavior parameter, weather scenarios, and other factors (Jenkins et al. 2008, Page et al. 2014). A spruce beetle outbreak transitions canopy fuel to the forest floor and changes forest composition and structure, but the magnitude of change to the fuel profile varies with time since outbreak and outbreak severity (Fig. 1A, Jenkins et al. 2008, Hicke et al. 2012). In the initial stage of attack ( green-stage ), foliar moisture and flammability remain highly variable, but a decrease in foliar moisture and changes to foliar chemistry in the red/yellow-stage promote needle flammability for a brief period prior to needle drop (Fig. 1B, Page et al. 2014). The pulse of needles (1-2 years post-attack) and increasingly coarser fuels to the forest floor initiates the gray-stage (3-10 years post-outbreak), a period in which the risk of canopy fire diminishes in severely-attacked stands and surface fire intensity is hypothesized to increase due to the increase in forest floor fuels (Jenkins et al. 2008, DeRose and Long 2009). Canopy fire risk may still be elevated in less severely attacked stands because canopy bulk density (CBD) remains high (DeRose and Long 2009). During the old-stage (1-4 decades post-outbreak), the transition of standing snags to coarse surface fuels is hypothesized to further elevate the potential for increased fire intensity and flame length (Jenkins et al. 2008), which is hypothesized to increase fire severity. These predictions are based on stand-scale fire behavior

18 Figure 1: Conceptual framework of (a) fuels characteristics and (b) fire behavior relative to preoutbreak conditions for red, gray, and old (snagfall and regrowth) phases. Surface fire properties include reaction intensity, rate of spread, and flame length. For postoutbreak phases, solid lines indicate higher confidence in responses based on Fig. 3, and dashed lines indicate lower confidence (more disagreement, fewer studies, or knowledge gaps). (Fig. 2 from Hicke et al. 2012) 10

19 11 models, which offer insight for fire suppression operations and scientific inquiry about the timing and severity of beetle mortality that may elevate flame length, increase fire line intensity, and promote fire type (Jenkins et al. 2008, DeRose and Long 2009). However, these models are limited by numerous factors including insufficient data on pre-outbreak fuel profiles, lack of flammability data on needles in different stages of an outbreak, the inability to consider finescale heterogeneity in fuels, lack of consideration of meso-scale fire-atmospheric feedbacks, and a scarcity of empirical field validation of actual fire behavior (Coen 2005, Jolly et al. 2012, Alexander and Cruz 2013). These limitations and the fundamentally different questions addressed by stand-level fire behavior models implies that these models are not sufficient for predicting fire severity across topographically complex landscapes with estimated fuel profiles resulting from bark beetle infestation. Landscape-scale retrospective studies in spruce-fir forests of Colorado have not supported the general expectation that spruce beetle-killed forests will inevitably increase fire frequency (Bebi et al. 2003), extent (Bigler et al. 2005), or severity (Hart et al. in prep, Kulakowski et al. 2003, Bigler et al. 2005). In the subalpine zone of Colorado, no detectable influence of spruce beetle on fire severity was found in studies conducted 2-3 years (Kulakowski and Veblen 2007), less than 5 years (Hart et al. in prep), and 10 years (Kulakowski et al. 2003) after a spruce beetle outbreak, and only a slight increase in the probability of burning at high severity was noted 50 years after a spruce beetle outbreak (Bigler et al. 2005). This may be partially explained by high-levels of preexisting fuel within mature stands of spruce, the area preferentially infested by spruce beetle, that commonly leads to high-severity fire irrespective of beetle activity (Veblen et al. 1994, Bebi et al. 2003, Kulakowski and Veblen 2007). However, these studies were limited by the accuracy and spatial resolution of the remotely derived beetle

20 12 and fire severity datasets, which lacked field validation. These limitations highlight the need for a research approach that includes quantification of the severity of both the pre-fire bark beetle outbreak and of the fire (e.g. Harvey et al. 2013). Beginning in the early 2000s, a widespread spruce beetle outbreak was detected across the eastern San Juan Mountains, southwestern Colorado (Appendix A, Colorado State Forest Service 2012). Following an extensive drought in 2012, the lightning-ignited West Fork Complex fires (West Fork, Papoose, and Windy Pass Fires; 2013), East Fork fire (2013), and Little Sands fire (2012) burned almost 60,000 ha through a range of early, gray-stage spruce beetle outbreak severities. Thus, this provides a rare opportunity to conduct an empirical, fieldbased study examining whether an active spruce beetle outbreak had a detectable influence on fire severity. Specifically, we seek to address the following question: How did biotic factors (forest attributes, including effects of beetle infestation) and abiotic variables (topography and weather) influence the severity of the fire events in the spruce-fir forests of southern Colorado? In accordance with the findings of previous studies based on fire behavior models (Jenkins et al. 2008, DeRose and Long 2009, Hicke et al. 2012), we hypothesize that fires burning though an early, gray-stage pre-fire beetle infestation (<5 years) would decrease the severity of crown fire in severely attacked stands (DeRose and Long 2009), elevate surface fire severity, or show no detectable effect on either crown or surface fire severity (Jenkins et al. 2008, Hicke et al. 2012). If the two disturbances are unrelated, then topography, extreme weather, and pre-existing high levels of fuel in mature stands of spruce-fir will be the overriding predictors of spatial variability of fire severity.

21 13 METHODS Study Area We sampled five lightning ignited fires that burned 60,000 ha in three National Forests of the southern Colorado Rocky Mountains (Fig. 2). The fires primarily burned through subalpine forests of Engelmann spruce, subalpine fir, and aspen, but also spread into upper montane mixed conifer forests of Douglas fir (Psuedotsuga menziesii) and white fir (Abies concolor). Lodgepole pine is uncommon in this region of the Rocky Mountains (Peet 2000). The study area ranges from the lower extent of the spruce-fir forest (2500m) to treeline (3500m) across steep terrain with plot slope ranging from 0 to 62 (mean 25 ). The mountains and basins surrounding the fires are composed of Precambrian rocks and more recent alluvial sediments, which were built and eroded by a variety of geologic processes (Romme et al. 2009). A late spring snowpack and summer monsoon rains maintain a cool and moist climate (Romme et al. 2009). From 1981 to 2010, annual mean precipitation was 85 cm, and mean annual temperatures ranged from - 7 C to 13 C (PRISM 2014).

22 Figure 2: Location of the five subalpine fires studied and the two weather stations used to assess burning conditions in the southern Colorado Rocky Mountains. All fires burned in the summer of 2013 with the exception of the Little Sands Fire that burned in The inset displays the location of the study area within Colorado. 14

23 15 The epidemic-level spruce beetle outbreak in the eastern San Juan Mountains, southwestern Colorado, resulted in high-levels of Engelmann spruce mortality (up to 95% of basal area as measured in stands of 0.04 ha). At time of fire, the area around the West Fork Complex fires was severely infested by spruce beetle (mean plot basal area infested, 58%), while beetle infestation in the Little Sands and East Fork fire was minimal (mean plot basal area infested, 8%). Trees within the study area were primarily in the early, gray-stage of attack at time of fire (<5 years post-outbreak, Appendix A), but many trees in earlier-stages of attack were observed in the field. All fires burned through remote areas of National Forest, primarily during extreme fire weather conditions. Consequently, suppression efforts were limited. We restricted sampling to areas without evidence of active suppression, which was determined by the evidence of fireline and recent chainsaw activity. Sampling design Plots (n= 143) were located across a range of spruce beetle and fire severities within spruce-fir dominated stands. Starting from a random location >100m from the fire perimeter, plots were systematically spaced by at least 400m to reduce the influence of spatial autocorrelation, and randomly located in a suitable site (e.g. > 80 % spruce-fir BA and no rock outcrops or meadow). In smaller fires (Windy Pass and East Fork Fire), a 400m grid was aligned within the fire perimeter to maximize potential sampling locations. Within each 20 x 20 meter plot (0.04ha), data were collected on pre- and post-fire stand structure and composition, severity of pre-fire beetle effects on tree mortality (percent of total basal area killed), and fire severity. Aspect ( ), slope ( ), elevation (m), and geographic position (UTMs) were recorded at plot center. Sampling was conducted from June to August 2014 (1 year after the West Fork Complex and East Fork Fire, and 2 years after the Little Sands Fire).

24 16 Determining the effects of spruce beetle and fire on stand structure We assessed stand structure and composition by recording tree species, tree status (live or dead), diameter at breast height (dbh) to the nearest centimeter, canopy position, and height (m) of the five tallest trees. We computed basal area, tree stems per hectare (stems/ha), and quadratic mean diameter from the plot data (Curtis and Marshall 2000) to examine the effect of spruce beetle and fire on stand structure. Pre-fire beetle outbreak severity assessment Following the methods outlined in Harvey et al. (2014a), pre-fire spruce beetle outbreak severity was measured by removing the bark on dead trees > 4cm (7,186 trees) to examine the cambium for evidence of Dendroctonous activity. Each tree was classified as (1) pre-disturbance snag, (2) killed by bark beetles prior to fire, (3) green attack at time of fire, (4) live at the time of fire, or (5) unknown (Appendix B). Trees classified as killed by spruce beetle prior to fire were divided into two categories: 1) visible cambium (20.9%), and 2) no visible cambium (10.3%). For those trees lacking cambium, surrounding stands with visible cambium were assessed for severity of beetle outbreak and minimum size class attacked (10-15cm). The spruce beetle was responsible for a majority of the tree mortality, but other potential mortality agents included: other bark beetles (e.g. Ips, Dryocoetes), climate, and tree competition (Romme et al. 2009). Fire severity assessment Field based and remotely sensed assessment of fire severity measured fire effects from the forest floor to the canopy. For each tree >4cm at dbh we recorded: char height to the nearest 0.5m, percentage of the tree charred (10% classes), maximum percentage of charring around the circumference of the main bole (10% classes), and the percentage of the bole with fine fuel (<1cm) and needles still attached (10% classes). Trees were classified as killed by fire if they were alive at time of fire, dead at time of sampling, and had no evidence of forest pathogens or

25 17 wounds (Appendix B). Because fire-killed trees contained significantly more moisture at time of fire than the beetle-killed trees, the fire-killed trees retained more fine fuel and bark post-fire. The beetle-killed trees charred deeper into the bole and commonly lacked fine-fuel, which helped distinguish them from the fire-killed trees. We computed fire-caused tree mortality (basal area and stems per hectare) from only those trees alive at time of fire for the Spearman s correlation and generalized linear model (described below). Trees retaining a significant portion of their green needles post-fire and without post-fire beetle activity were classified as live. We measured forest floor fire severity by dividing each plot into quadrants and estimating the percentage of forest floor charred within each quadrant to the nearest 10%. We did not sample other measures of surface fire severity (e.g. litter and duff loss), because precipitation events had significantly altered immediate post-fire characteristics at time of sampling (one to two years post-fire). To compare field-based measures to remote measures of fire severity, we extracted the Relative Differenced Normalized Burn Ratio (RdNBR) value, downloaded from the Monitoring Trends in Burn Severity website (MTBS 2014), for each plot in the West Fork Complex and Little Sands fires. RdNBR data were not available for the East Fork fire. RdNBR is a measure of pre- and post-fire-caused change to live green biomass, moisture content, and soil characteristics. Topography and solar radiation Previous research, both retrospective studies and fire behavior models (e.g. BehavePlus, FARSITE), suggest that differences in solar radiation and available moisture created by topographic variables (e.g. aspect, slope, and curvature) influence fuel characteristics and fire activity (Holden et al. 2009). We derived and extracted the following topographic and radiation variables at plot center from a 30m digital elevation model in ArcGIS 10.2 for the multivariate linear regression models (described below) predicting fire severity: slope ( ), aspect (Northeast Index, Beers et al. 1966), topographic curvature (the second derivative of elevation

26 18 (Zevenbergen and Thorne 1987), slope position, annual solar radiation (McCune and Keon 2002), and heat load index (McCune and Keon 2002). Slope position is a relative measure of elevation from 0 (toe of slope) to 1 (ridge top) that is preferable to raw elevation because it accounts for the relative position of a location on the slope. Both annual solar radiation and heat load index rescale aspect, to measure incident radiation along a north-south axis and relative temperature along a northeast (coolest slope)-southwest axis (warmest slope), respectively. Assessing burning conditions Real-time weather data was not available in close enough proximity to our plots to account for variation in climatic conditions (microclimates) created by the mountainous terrain. Instead, we used daily weather conditions to categorize each day as burned under moderate (55 plots) or extreme (85 plots) burning conditions (Collins et al. 2006, Thompson and Spies 2009) for use in the burning condition and RdNBR overlay analysis and Spearman s correlation (see below). Specifically, we used maps of daily fire progression from the Geospatial Multi-Agency Coordination Group (GeoMAC, USGS 2013) and hourly weather data from the nearest Remote Automated Weather Station (RAWS) during the burning period (1000 to 1700 hours), to assign a burning condition to each day of growth. For the West Fork Complex fires and East fork fire, extreme burning conditions were assigned to days with temperatures >17 C, relative humidity <17.5%, average wind speed >4.5m/s, and rapid-fire growth (Appendix C). Additionally, days with windspeeds >5.4 m/s and relative humidity <22% were also considered extreme burning days. For the Little Sands Fire, extreme burning conditions were assigned to days with temperatures >22 C, relative humidity <14.4%, average wind speed >1.9m/s, and rapid-fire growth (Appendix C). Days that did not have weather conditions that met the criteria for extreme conditions were classified as moderate burning conditions. The thresholds applied here are

27 19 similar to other studies, which have found significant differences in fire severity with short-term weather conditions (e.g. Harvey et al. 2014a). Burning condition and RdNBR overlay analysis To assess the influence of burning conditions on fire severity at the landscape scale, we produced an overlay analysis examining the hectares burned in three levels of RdNBR fire severity (low, moderate, high from Miller et al. 2009) for the West Fork Complex and Little Sands Fire under moderate and extreme burning conditions. First, we extracted all areas within the fires that were dominated by the spruce-fir forest cover type (USDA 2013) and intersected this spatial dataset with the burning conditions spatial dataset to delineate spruce-fir areas that burned under moderate and extreme conditions. From within these two categories, we extracted the RdNBR values. Then, we calculated the mean average RdNBR value, standard error, and percentage that burned at low, moderate, and high severity by burning condition. Determining the effects of pre-fire spruce beetle outbreak severity on fire severity We used two analytical approaches to assess the influence of severity of beetle-caused tree mortality (% of total basal area killed by beetle) on fire severity: 1) Spearman s rank correlations and 2) multivariate linear regression modeling. All statistical analysis was performed using the R statistical software. Using a Spearman s rank correlation, we tested beetle-killed basal area against each quantitative measure of fire severity (char height (m), fire-killed stems (%), fire-killed basal area (%), surface charred (%), fine fuel remaining (%)) for all plots. A significant and positive Spearman s rho implies that increasing beetle outbreak severity may also increase the measure of fire severity, while a significant and negative Spearman s rho implies that increasing beetle outbreak severity may decrease the measure of fire severity. To minimize the chances of not detecting an ecologically significant effect of the severity of beetle outbreak (measured as

28 20 percent of basal area killed by beetles) on fire severity that was actually present (Type II error), we set the α to Spearman correlations used the function cor in the stats package in R. The univariate Spearman s correlation for statistical dependence did not consider other variables (e.g. slope position or stand density) known to alter fire severity, so we also examined beetle severity in the context of other biophysical factors with multivariate generalized linear models. A model constructed for each fire severity variable compared the influence of beetle outbreak severity to topography (slope position, topographic curvature, aspect, slope, slope position), solar radiation (incoming solar radiation, heat load index), and stand structure (combined live and dead basal area and stem density) at time of fire. All fire severity variables represented by a percent were logit transformed, and predictor variables were z-score transformed, so beta coefficients from the models could be compared. We selected the subset of biophysical variables that best predicted fire severity using forward, backward, and both stepwise model selection and the Bayesian Information Criterion (BIC). Variable selection removed beetle-killed basal area from all models, but we forced this variable into the final model to show its relative influence. Though it was a significant predictor, burning condition was excluded from all models. Our sampling design captured a range of fire severity levels under both moderate and extreme conditions, which did not accurately represent the disproportionately high amount of high severity fire. Thus, mean average and range of each fire severity metric was falsely similar under both burning conditions. Instead, the overlay analysis mentioned above was used to examine the influence of burning conditions on fire severity. Final models included basal area (combined live and dead at time of fire), slope position, and beetle-killed basal area. The multivariate generalized linear models used the function glm with a Gaussian distribution in the stats package in R.

29 21 RESULTS Effects of spruce beetle and fire on stand structure The spruce beetle outbreak reduced the live basal area by greater than 50% (Fig. 3A), the stem density by 33% (Fig 3B), and the mean live tree size (Fig. 3C). The severity of beetlecaused tree mortality ranged from 0 to 95% (mean 46%, Fig. 3) of total basal area at the plot level, and was almost entirely composed of spruce beetle killed Engelmann spruce (>99%). The remainder was spruce beetle attacked subalpine fir. The fires further reduced live basal area by 66% and stem density by 83%, while increasing the density of snags (Fig. 3D), mean live tree size, and variance in mean live tree size (Fig. 3C).

30 Figure 3 (A-D): Stand structure characteristics prior to spruce beetle outbreak and fire, after spruce beetle outbreak, and after spruce beetle outbreak and fire for all plots (n=143). Preoutbreak and pre-fire characteristics were reconstructed with the criteria presented in Appendix B. All values are mean averages and 95% confidence intervals 22

31 23 Burning Condition and RdNBR overlay analysis Moderate and extreme burning conditions had similar average temperature, but extreme burning conditions had considerably lower average relative humidity, higher average wind speeds, and higher average wind gusts for all fires (Appendix C). A majority of the acreage burned by all fires occurred during extreme burning conditions (91% of hectares burned, Table 1, Appendix C) and burned at a high severity (78% of hectares burned, Table 1). Examining only spruce-fir dominated stands (Table 1), moderate burning conditions resulted in lower fire severity (mean RdNBR = 785, se = 2.33) than extreme burning conditions (mean RdNBR = 929, se = 0.70), and both burning conditions produced a disproportionately high-level of high severity fire. Low fire severity was evenly distributed under both burning conditions, but a higher percentage of moderate fire severity was represented under moderate burning conditions. Table 1: Assessment of landscape scale fire severity (RdNBR) and hectares burned by burning condition from the burning condition and RdNBR overlay analysis Burning Conditions Average RdNBR RdNBR SE Low Severity (ha) Moderate Severity (ha) High Severity (ha) Hectares burned Moderate (21%) 367 (14%) 1,726 (65%) 2,666 9 Percent of fire burned Extreme ,251 (21%) 2,180 (8%) 20,299 (79%) 25, * West Fork Complex and Little Sands fires

32 24 Effects of pre-fire spruce beetle outbreak severity on fire severity Despite the disproportionally small area that burned as low to moderate severity, we sampled a range of each fire severity metric to account for any circumstances where fire effects may be affected by a range of beetle outbreak severities. Under moderate burning conditions, when examining univariate relationships (Table 2), the percentage of beetle-killed basal area was positively correlated with the percent of fire-killed stems (rs =.231, P < 0.09, Fig 4G) and firekilled basal area (rs =.275, P < 0.04, Fig. 4H). All other univariate relationships were not significant (P < 0.10) under either moderate or extreme burning conditions (Table 2; Fig. 4A-E). Table 2: Spearman Correlation (rho) and p-values ((*): P <0.05, (.): P <0.10) for each fire severity response variable and percent of basal area killed by spruce beetle in each plot across all fires. The number of plots in each burning condition and the range of percent beetle killed basal area sampled are listed below. Extreme Moderate Total Fire Effect rho P rho P rho P Char height (m) Bole scorch (%) Fine fuel remaining (%) Surface charred (%) Fire-killed stems/ha (%) Fire-killed basal area (%) * RdNBR (*) P < 0.05, (.) P < 0.1 Range of Beetle kill = 0-95% of total basal area

33 Figure 4 (A-G): Fire severity measures plotted against spruce beetle killed basal area under moderate and extreme burning conditions (moderate = circle, extreme = triangle) in an early, gray-stage spruce beetle outbreak in subalpine, spruce-fir forests. RdNBR stands for Relative differenced Normalized Burn Ratio. Percent fine fuel measures the remaining fine fuel (< 1cm) attached to the tree bole. BA stands for basal area. 25

34 26 In multivariate generalized linear regression analysis, basal area (combined live and dead at time of fire) and slope position were the only variables that were significant across multiple fire severity metrics and consistently improved BIC. Beetle-killed basal area was not a significant predictor of any fire severity metrics (Table 3) and consistently had a lower magnitude beta coefficient, except for the percent of fine fuel remaining severity metric. As slope position (positive beta) increased so did fire-killed stems/ha (%) and fire-killed basal area (%), while basal area (combined live and dead, negative beta) decreased fire-killed stems/ha, bole scorch (%), surface charred (%), and RdNBR (Table 3). All other variables (e.g. combined live and dead stems/ha and annual solar radiation) considered in the models were consistently poor predictors. Burning conditions were also excluded from the models, because our sampling protocol captured more low and moderate fire severity than occurred in each fire (see methods). Though some variables were significant predictors of fire severity, all models had low predictive power (pseudo r 2 range = ).

35 27 Table 3: Generalized linear model results from the fire severity models used to compare the beetle outbreak severity coefficient and significance to the following non-beetle related predictor variables: pre-fire total basal area (live and dead basal area of all species at time of fire) and slope position (relative position on slope). The fine fuel percentage is the percent of the bole with fuel less than a quarter inch remaining. A positive beta increases fire severity and negative beta decreases fire severity. Response and Predictor Beta SE t P Pr(> t ) Fire-killed stems/ha (%) Pre-fire total basal area <0.01 ** Slope position * Beetle-killed basal area Fire-killed basal area (%) Slope position * Beetle-killed basal area Char height (m) Pre-fire total basal area Slope position Beetle-killed basal area Fine fuel remaining (%) Pre-fire total basal area Slope position Beetle-killed basal area Bole scorch (%) Pre-fire total basal area <0.01 ** Slope position Beetle-killed basal area Surface charred (%) Pre-fire total basal area <0.01 ** Slope position Beetle-killed basal area RdNBR Pre-fire total basal area Slope position Beetle-killed basal area (**) P < 0.01, (*) P < 0.05, (.) P < 0.1