EFFECTS OF DISTURBANCE ON ANIMA L COMMUNITIES: FIRE EFFECTS ON BIRDS IN MIXED-CONIFER FOREST By NATHANIEL E. SEAVY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Nathaniel E. Seavy
iii ACKNOWLEDGMENTS This work could not have been completed without the help of many people. Throughout the process, my advisor, Colin Chapman, and my committee members, Mike Binford, Benjamin Bolker, Scott Robinson, a nd Katie Sieving, have offered support and guidance. John Alexander was gracious and ge nerous in his invitati on to work with the Klamath Bird Observatory. The opportunity to work with the dive rse group of biologists and educators at the Klamath Bird Observator y has been one of the highlights of this experience. Their input improved this work at all stages. A number of people deserve appreciation for the role they played in specif ic parts of this projec t. During the five-year study, vegetation and bird surveys in the Lit tle Applegate Valley were conducted by Dan van den Broek, Robert Chapman, Michael Cl egg, Elizabeth Crosson, Amanda Darlak, Kevin Glueckert, Sherri Kies, James Lawrence, Frank Lospalluto, and Laurel Sutherlin. Bill Hogoboom assisted with GIS analyses a nd generating the maps. Laurel Genzoli and Leisa Glass assisted with the collection of arthropod samples, and Seabird McKeon assisted with arthropod identification. Convers ations with Joe Font aine, Mark Huff, and Dennis Odion helped to develop my understand ing of fire ecology. Conceptually, this work has benefited from my interactions with both the students and faculty at the Department of Zoology. These friends have made my graduate education challenging, rewarding, and fun. I would also like to thank my parents an d family for their steadfast support. Finally, I thank Steven Herman and David Whitacre for the formative opportunities and education that they provided.
iv Funding for this project was provided by th e Joint Fire Sciences Program project 01B-3-2-10, Rogue River-Siskiyou National Forest, Bureau of Land Management Medford District, the Klamath Bird Observ atory, and the University of Florida, Department of Zoology.
v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... ..x CHAPTER 1 GENERAL INTRODUCTION....................................................................................1 2 INTERACTIVE EFFECTS OF FLORISTICS AND PHYSIOGNOMICS ON PASSERINE BIRD DISTRIBUTION.........................................................................5 Introduction...................................................................................................................5 Methods........................................................................................................................8 Study Area.............................................................................................................8 Vegetation Protocol...............................................................................................9 Bird Surveys........................................................................................................10 Statistical Analyses..............................................................................................11 Model selection............................................................................................11 Model performance......................................................................................12 Visual presentation of models......................................................................13 Results........................................................................................................................ .14 Discussion...................................................................................................................16 Conclusions.................................................................................................................19 3 WILDFIRE EFFECTS ON VEGETATI ON AND PASSERINE BIRDS IN A MIXED-CONIFER FOREST: COMBIN ING HABITAT MODELS WITH A NATURAL EXPERIMENT.......................................................................................41 Introduction.................................................................................................................41 Methods......................................................................................................................43 Study Area...........................................................................................................43 Fire Effects..........................................................................................................45 Vegetation Protocol.............................................................................................45 Bird Surveys........................................................................................................47
vi Statistical Analyses..............................................................................................47 Vegetation data.............................................................................................47 Bird data.......................................................................................................48 Results........................................................................................................................ .50 Vegetation Structur e and Composition...............................................................50 Bird Abundance...................................................................................................51 Are Observed Differences Consistent With Habitat Model Predictions?...........53 Discussion...................................................................................................................54 Changes in Vegetation.........................................................................................54 Changes in Bird Abundance................................................................................55 Conclusions.................................................................................................................58 4 POST-FIRE HABITAT QUALITY FOR PASSERINE BIRDS: PREDATOR ABUNDANCE, ARTHROPOD ABUND ANCE, AND FORAGING BEHAVIOR.72 Introduction.................................................................................................................72 Methods......................................................................................................................74 Study Area and Fire Effects................................................................................74 Study Design.......................................................................................................75 Predator Abundance............................................................................................76 Arthropod Abundance.........................................................................................77 Avian Foraging Behavior....................................................................................77 Statistical Analyses..............................................................................................78 Predator abundance......................................................................................78 Arthropod abundance...................................................................................78 Avian foraging behavior...............................................................................79 Results........................................................................................................................ .79 Predator Abundance............................................................................................79 Arthropod Abundance.........................................................................................80 Avian Foraging Behavior....................................................................................81 Discussion...................................................................................................................81 Predator Abundance............................................................................................81 Arthropod Abundance.........................................................................................83 Avian Foraging Behavior....................................................................................85 Conclusions.................................................................................................................86 5 CONCLUSION...........................................................................................................94 APPENDIX AVIAN FORAGING BEHAVIOR DATA COLLECTED IN AND ADJACENT TO THE QUARTZ FIRE IN SOUTHERN OREGON..............................................98 LIST OF REFERENCES.................................................................................................102 BIOGRAPHICAL SKETCH...........................................................................................112
vii LIST OF TABLES Table page 2-1. Frequency of occurrence for vegetation taxa at 979 point count stations in the Little Applegate watershed of southern Oregon......................................................21 2-2. Vegetation characteristics measured in eight structural and compositional strata in the Little Applegate watershed of s outhern Oregon. Measurements are means with standard errors..................................................................................................23 2-3. Frequency of occurrence for 81 bird sp ecies at 979 point count stations in the Little Applegate watershed of southern Oregon......................................................24 2-4. Parameter values for logistic regr ession models identified using stepwise regression and their Area Under the Curv e (AUC) values derived from receiver operating characteristic plot s for predictive models of bird distribution in the Little Applegate watershed in southern Oregon.......................................................27 2-5. Habitat associations for passerine bi rds in the Little A pplegate watershed of southern Oregon based on the vegetation st rata they were predicted to occupy from logistic regression............................................................................................29 3-1. Likelihood ratio test comparing nested models that describe the proportion of sites occupied as a function of year (Y), area (B; burned and unburned), and the interaction of these predictors..................................................................................60 4-1. The number of foraging observations r ecorded for 16 species of passerine birds in burned and unburned areas of the Little Applegate Valley, Oregon....................89 A-1. Summary of data collected during avia n foraging observations in and adjacent to the Quartz Fire in southern Oregon during 2004 and 2005....................................100
viii LIST OF FIGURES Figure page 2-1. The study area in the L ittle Applegate valley, Ore gon, USA, showing the spatial distribution of stations where inform ation on bird abundance and vegetation characteristics were collected...................................................................................31 2-2. Sampling locations across a physiognomic gradient of vegetation volume (0 = low volume, 1 = high volume) and a flor istic gradient of broadleaf-conifer composition (0 = all broadleaf vegeta tion, 1 = all conifer vegetation)....................32 2-3. Receiver-operating characteristic curves for logistic models predicting probability of occurrence for f our species of passerines..........................................33 2-4. Frequency of occurrence of six vegetati on taxa across the eight vegetation strata...34 2-5. Predicted probability of occurren ce as a function of vegetation volume and vegetation composition passerine birds in the Little Applegate Valley of southern Oregon.......................................................................................................35 2-6. A figure illustrating that treatments designed to reduce ve getation volume by the same amount are predicted to have very different effects on the probability of occurrence of Black-headed Grosbeaks...................................................................40 3-1. A map of the study area in the Little Applegate Valley, Oregon, USA, that shows the area burned by the Quartz fire a nd the locations of burned and unburned stations where data on bird abundance and vegetation were collected from 2001 thru 2005..................................................................................................................62 3-2. Percent cover of six vegetation taxa at burned and unburned areas in the year before and 4 years after fire in the Lit tle Applegate Valley of southern Oregon.....63 3-3. Sampling locations across a physiognomic gradient of vegetation volume (0 = low volume, 1 = high volume) and a flor istic gradient of broadleaf-conifer composition (0 = all broadleaf vegetation, 1 = all conifer vege tation) in burned (panel A) and unburned (panel B) areas in the year before and fourth year after the Quartz Fire..........................................................................................................64 3-4. Proportion of stations occupied by 30 passerine species in burned and unburned areas in the year before and 4 years after the Quartz Fire........................................65
ix 3-5. Example of the observed proportion of occupied stations in burned and unburned areas and the predicted propor tion of occupied stations for the Hermit Warbler....69 3-6. The observed difference in the proporti on of occupied stations in burned and unburned areas versus the predicted di fference based on habitat occupancy models (see chapter 2) for the 5 year study period..................................................70 3-7. Observed difference in the proporti on of occupied stations in burned and unburned areas for eight species that d ecreased as a result of the fire.....................71 4-1. The proportion of total predator detec tions (N = 112) composed of eight avian and mammalian taxa in burned and unbur ned areas of the Little Applegate Valley, Oregon, in 2004 and 2005...........................................................................90 4-2. The proportion of total biomass compos ed of eight arthropod groups from burned and unburned areas of the Little Applegate Valley in 2004 and 2005.....................91 4-3. Spatial (burned and unburned) and te mporal (2004 and 2005) variation in arthropod biomass samples collected by sw eep netting in the Little Applegate valley of southern Oregon........................................................................................92 4-4. Foraging rates compared between obs ervation in burned and unburned areas and in conifer and broadleaf vegetation..........................................................................93
x Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EFFECTS OF DISTURBANCE ON ANIMA L COMMUNITIES: FIRE EFFECTS ON BIRDS IN MIXED-CONIFER FOREST By Nathaniel E. Seavy August 2006 Chair: Colin A. Chapman Major Department: Zoology In the mixed-conifer forests of western No rth America, fire is an important force that changes the structure and composition of forest vegetation. I investigated the association of forest structur e and composition with patterns of bird distribution in the Little Applegate Valley of s outhern Oregon. Using extensiv e bird and vegetation surveys in this watershed, I demonstrated that vegetation composition (broadleaf-conifer composition) and structure (total vegetation vol ume) can be used to predict what forest types are occupied by passerine bird species. For some species, the effect of vegetation volume varied depending on compositional characteristics; for example, species associated with conifer vegetation often pe rsisted in relatively low volume vegetation when it was dominated by conifers, but were ab sent from stands with a similar vegetation volume that had a mix of conifer and broadlea f vegetation. These models can be used to understand how and why wildfires change bird communities. In the summer of 2001, a wildfire burned 2,500 ha of the study area in so uthern Oregon. The burned stations and a
xi subset of the unburned stations were surveyed fo r the next four years. I used these before and after data to evaluate how the proporti on of stations occupied by 29 passerine bird species changed as a result of the fire. Th e ability of the habitat models to predict differences between burned and unburned areas was poor immediately after the fire, but steadily improved over the 4 year study period. These results suggest that bird-habitat models are useful tools for predicting eff ects of large-scale di sturbances, but their predictive power may vary as a function of time since disturbance. Although other studies have discussed the po ssibility that post-fire cha nges in food availability and predator activity may be an important factor driving changes in post-fire bird abundance, I found little evidence that these mechanisms acted independently of fire-induced changes in vegetation structure and compositi on. Together, this information links postfire changes in bird abunda nce to changes in vegetation composition and structure that are created by fire. Specifica lly, they suggest that the role of fire in maintaining a broadleaf component of vegetation in mixed-co nifer forests is an important factor for understanding fire effects on bird communities.
1 CHAPTER 1 GENERAL INTRODUCTION Fire is a pervasive force throughout many landscapes in North America and globally (Pyne 1982, Agee 1993). For much of the 20th century, fire suppression has been the major policy for most federal manage ment agencies in the United States (i.e., Bureau of Land Management, Bureau of Indi an Affairs, Fish and Wildlife Service, National Park Service, and USDA Forest Serv ice). In each year from 2000 to 2004, these agencies combined spent between 890 milli on and 1.6 billion dollars on fire suppression (National Interagency Fire Center 2004). In addition to carrying a large economic cost, policies of fire suppression are often blamed for allowing flammable vegetation to accumulate, thus creating larger, more severe fires than would be expected under natural fire regimes (Agee 1993, Dombeck et al. 2004). In an attempt to avoid the economic and ecological costs of fire suppression, managers are increasingly turni ng to prescribed-fire and mechanical fuels reduction to decrease the risk of catastrophic fire (Spi es et al. 2006). Ideally, these activities would not only reduce the economic risks of wildfire, but would also restore ecological characteri stics that have been altered by decades of fire exclusion (Brown et al. 2004). However, because our unde rstanding of the role of natural wildfire in these systems is still limite d, the ability of alternative me thods of fire management to create the desired conditions for forest eco systems is untested (Tie demann et al. 2000). Understanding how and why fire changes animal communities is an important component of managing forested lands, both when ma king decisions about managing post-wildfire
2 landscapes and when designing management plan s that are meant to mimic the effects of natural fire regimes. The role of fire in structuring bi rd community composition in the Pacific Northwest was recently reviewed by Huff et al. (2005). Fire regimes vary across this region. In the wettest areas fires occur infr equently, conifers do minate, and broadleaf species are generally restricted to the understo ry or riparian areas (Huff et al. 2005). In contrast, drier areas burn more frequently a nd are often dominated by broadleaf trees and shrubs. In these areas, conifer trees must r each a large size before they become resistant to the effects of frequent fires. Intermedia te conditions foster a diverse mix of broadleaf and conifer species that is maintained by t opography, soils, and fire disturbance (Huff et al. 2005). However, despite a well-develope d body of knowledge about how fire creates and maintains spatial heterogeneity in fore st structure and composition (Whittaker 1960, Skinner 1995, Taylor and Skinner 1998), ther e is almost no information on how these dynamics influence bird communities (Huff et al. 2005). The goal of my dissertation is to adva nce the understanding of how and why fire changes bird communities in mixed-conifer forests of the Pacific Northwest by (1) building models that predict patterns of bi rd occurrence as a function of vegetation characteristics that are changed by fire, (2) us ing before and after data to describe the post-fire changes in bird abundance, and (3 ) evaluating the degree to which post-fire changes in bird communities are associat ed with vegetation ch aracteristics, food availability, and predation pressure. In Chapter 2, I investigate the ability of simple measures of vegetation volume and the proportion of vegetation comprised of conife r species to explain spatial patterns of
3 bird distribution in a watershe d in southern Oregon. Ecologist s have long recognized that vegetation characteristics are often strong predictors of sp atial distributio n patterns of terrestrial birds (Grinnell 1917). Subsequent ly, many studies of bird communities have emphasized structural characteristics of vegetation and the species composition of vegetation to explain patterns of bird di versity (MacArthur and MacArthur 1961, Verner and Larson 1989), community composition(Rotenberry 1985, Bersier and Meyer 1994), and the abundance of individual species (Saab 1999). Rather than us ing a large number of variables to generate complex models for each species, I have limited my models to the broadleaf-conifer composition and vegetation volume, predictive variables that are directly influenced by fire. These mode ls can be applied to local and regional management decisions because they predic t how changes in vegetation structure and composition will affect patterns of bird distribution. In Chapter 3, I use data on bird abundance and vegetation, collected one year before and four years after a w ildfire, to evaluate fire eff ects on bird abundance. These data come from the Siskiyou Mountains of southern Oregon, where the Quartz fire burned 2,500 ha of mixed-conifer forest in 2001. In the past, the ability to make inferences about the effects of large-scale changes in ha bitat conditions, as created by fire, on animal populations has been limited be cause traditional expe rimental designs are logistically or ethically impossible to conduct (van Ma ntgem et al. 2001, Parker and Wiens 2005). In lieu of traditional experime ntal designs, I use two approaches to infer the degree to which changes in bird abundan ce were caused by the fire. First, I used before and after data to evalua te amount of variation in bi rd abundance in the absence of fire, and then compared this to areas wher e the fire burned. Second, I used the models
4 presented in Chapter 2 to evaluate whethe r post-fire changes in bird abundance were consistent with the predictions based on ch anges in vegetation composition and structure that occurred over the same time period. In Chapter 4, I use data on predator abundance, arthropod abundance, and bird foraging behavior to evaluate mechanisms beyond changes in vegetation structure and composition that may affect bird abundance. Although these factors may be an important aspect of post-fire changes in bird abundance, they have received very little study (Saab et al. 2004, Huff et al. 2005). This chapte r addresses this gap in our knowledge by (1) comparing the abundance of avian and mammalia n predators that might influence nesting success of birds in burned and unburned ar eas, (2) comparing the abundance of arthropods in burned and unburned areas to evaluate the degree to which food availability may be affected by fire, and (3) using observati ons of bird foraging behavior to evaluate the degree to which birds may respond to post-fire environments by changing their foraging behavior. Together, these chapters link the post-fire changes in bird abundance to the changes in vegetation composition and structure that are created by fire. Specifically, they investigate the degree to which fire may ch ange bird communities by maintaining the broadleaf component of vegetati on in mixed-conifer forests.
5 CHAPTER 2 INTERACTIVE EFFECTS OF FLORISTICS AND PHYSIOGNOMICS ON PASSERINE BIRD DISTRIBUTION Introduction Patterns of community composition are an aggr egation of patterns of many species, each with a unique spatial pattern of dist ribution and abundance. These patterns are driven by environmental varia tion and ecological interacti ons, the importance of which likely varies across spatial scales. At the largest spatial scales, distribution patterns are associated with climatic patterns of temp erature and rainfall (Root 1988), whereas at local scales, variation in factors such as nut rient availability and topography (Clark et al. 1999), microclimate (Seavy In press), and ha bitat structure (Saab 1999) are important, either intrinsically, or becau se they provide such important resources as food and protection from predators. At the scale of th e individual, these patterns may be masked, or enhanced by interactions among organisms, such as competition or facilitation (Bruno 2000, Cheplick 2005). Ecologists have long recognized that vege tation characteristics are often strong predictors of spatial distribution patter ns of terrestrial birds (Grinell 1917). Subsequently, many studies of bird co mmunities have emphasized structural characteristics of vegetation (physiognomy) and the species composition of vegetation (floristics) to explain patte rns of bird diversity (MacArt hur and MacArthur 1961, Verner and Larson 1989), community composition (Rot enberry 1985, Bersier and Meyer 1995), and the abundance of individual species (Saab 1999). These models are useful tools for
6 predicting how changes in environmental conditions will influence patterns of biodiversity (Guisan and Thuiller 2005). These patterns suggest that temporal and spatial variability in vegetation structure and composition are important factors for understanding regional patterns of bird community composition. In the Pacific Northw est, vegetation structure and composition can be highly variable in space and time. Hist orically, much of this variation is created and maintained by complex interactions of disturbance history and abiotic conditions (Agee 1993, Huff et al. 2005). In the wettest areas where fires occur infrequently, conifer species dominate and broadleaf species are ge nerally restricted to the understory or riparian areas (Huff et al. 2005). In contrast , drier areas that burn more frequently are often dominated by broadleaf trees and shrubs , and conifer trees must reach a large size before they become resistant to the effect s of frequent fires (Huff et al. 2005). Intermediate conditions foster a diverse mix of broadleaf and conifer species with highly variable structural characterist ics (Huff et al. 2005). To date, most investigations of bird distribution have focused on ve getation structure, either in response to stand-level silvicultural treatments (Hay es et al. 2003, Hagar et al. 2004 ) or landscape patterns of stands with different stru ctural characteristics (McGar igal and McComb 1995). In contrast, there are fewer studies that have directly addre ssed the importance of forest composition (Easton and Martin 199 8, Cushman and McGarigal 2004). A pervasive challenge of modeling bird distribution with vegetation data is selecting which of the many predictive vari ables should be used. One approach to solving this problem is to include multiple predictor variables and then use model selection tools, such as Akaikeâ€™s Information Criteria, to identify the best model (Rushton
7 et al. 2004). However, the â€œbest modelâ€ may not always be the most useful model (Stephens et al. 2005). For example, the si mplicity of explaining the distribution of multiple species within a community using the same variables may outweigh the advantages of using different predictor variables for each sp ecies. Alternatively, data reduction techniques, such as principal com ponents analysis, can be used to reduce a large number of correlated variables to a smalle r set of variables that capture much of the original variation (James and Wamer 1982, Gu enette and Villard 2005). Unfortunately, these new variables may not always be easy to interpret, especially within the context of forest management. Perhaps in part because of this complexity, few species distribution models have incorporated interactive effects of multiple predictors (Guisan and Thuiller 2005). Here, I investigate the ability of simple measures of vegetation volume (a major physiognomic variable) and the proportion of ve getation comprising of conifer species (a major floristic variable) to explain spatial pa tterns of bird distribution among stands in a watershed in southern Oregon. Rather than us ing a large number of va riables to generate complex models for each species, I limited my models to the broadleaf-conifer composition and vegetation volume in order to (1) build models that could easily be compared across a wide variety of bird species , (2) generate information that could easily be applied to local and regional management decisions, and (3) explicitly test for interactive effects of vege tation structure and compositi on. I use contour plots to graphically present how plot-lev el characteristics of vegeta tion and forest structure are associated with the probability of occurrence of individual bi rd species. This approach
8 allows for both the statistical detection and graphical presentation of complex interactive effects of vegetation composition a nd structure on bird distribution. Methods Study Area This study was conducted in the Little Applegate Valley in Jackson County, approximately 17.5 km southwest of As hland, Oregon (524000E, 4671000N, UTM zone 10, NAD27; Fig. 2.1). Elevation of the sampling locations ranged from 400 to 2250 m above sea level. The forest vegetation of the Klamath-Siskiyou Region is diverse and includes both conifer and hardw ood species (Whittaker 1960). At lower elevations and on drier sites, vegetation is dominated by broa dleaf species, in the form of oak woodlands with both Oregon white oak ( Quercus garryana ) and California black oak ( Quercus kellogii ) or sclerophyllous shrublands (chaparral) in which broadleaf shrubs of the genera Ceanothus, Arcostaphyllus, and Baccharis are abundant (Huff et al. 2005). On wetter sites or middle elevations, the vegetation be comes a mix of conifers, including Douglasfir ( Pseudotsuga menziesii ), ponderosa pine ( Pinus ponderosa ), incense-cedar ( Calocedrus decurrens ), and white fir ( Abies concolor ), and hardwoods include tanoak ( Lithocarpus densiflorus ), Pacific madrone ( Arbutus menziesii ), Canyon Live Oak ( Quercus chrysolepis ), California black oak, Oregon wh ite oak, and big-leaf maple ( Acer macrophyllum ). The relative composition of these sp ecies varies with elevation, aspect, and soils. Generally, these forest types corr espond to the Douglas-fir, Mixed Evergreen Hardwood, or White Fir Types (Franklin a nd Dyrness 1973, Huff et al. 2005). Firerelated studies in these vegeta tion types show a mix of fire severities, frequencies, and sizes, typically characteristic of low a nd moderate-severity fire regimes (Agee 1991, Taylor and Skinner 1998). Over time, such mixed-severity fires can create forests with
9 multiple age classes, often w ith Douglas-fir or ponderosa pine as an emergent canopy above various hardwoods. Within the Litt le Applegate watershed, transects of 10-18 survey stations spaced 250 m apart were es tablished along secondary and tertiary roads and trails (Fig. 2.1). Although locating point count stations on roads may increase the number of detections of bird s associated with shrubs and edges, this method generally does not have large effects on the ability to detect patterns of habitat association with stand-level vegetation charac teristics (Hanowski and Niem i 1995). In the field, GPS units (Garmin GPS 12 XL) were used to record the location (UTM coordinates) of every station. Vegetation Protocol To generate measurements of vege tation volume and broadleaf-conifer composition, I combined several variables th at were available from the vegetation surveys conducted at all point count stat ions in 2001. These vegetation data were collected using variable radius (20-50 m) relev plots (R alph et al. 1993). Within the circular plot defined by the ra dius, eight structural variab les were recorded: total tree cover, total shrub cover, maximum tree di ameter, minimum tree diameter, and average canopy top height. Total cover of the tree stratum (all vegetation 5 m) and shrub stratum (all vegetation 0.5 m and < 5 m) were estimated and assigned the mid-point one of 5 cover classes (0-2.5, 2.5-25, 25-50, 50-75, a nd 75-100%). Maximum and minimum diameter at breast height (DBH) of trees in the tree layer was visually estimated, as was the average height of the canopy top (m). For the analysis of bird diversity and individual species abundance, I used an i ndex of total vegetation volume as a single measure of structure. Measures of vegetati on volume have often been used to explain spatial variation in bird de nsity (Mills et al. 1991) and ri chness (Fleishman et al. 2003,
10 Keller et al. 2003). The measure was cal culated by multiplying the canopy height measurement by the cover measurement for the tree layer. I divide d all values by the maximum observed vegetation volume to create a standardized measure that varied from 0 to 1. The percent cover of indivi dual tree and shrub taxa was also recorded using the 5 cover classes described above. These cover sc ores were used to develop an index of vegetation composition that expressed the proportion of the vegetation comprised of conifer (versus broadleaf) species (Table 2-1). This measure was calculated as: Ci/( Ci+ Bi) where Ci are the cover values of all coniferous tree and shrub species at a station and Bi are the cover values of all broadleaf trees and shrubs at a station. Thus, the conifer proportion ranges from 0 (all tree cover was broa dleaf species) to 1 (all cover was conifer species). Bird Surveys Bird abundance was measured at all st ations between 14 May and 1 July 2001 using standardized point count methodologies (Ralph et al. 1993). Five-minute bird counts were conducted between sunrise a nd 1000 PDT on each station. During each count the distance at the time of first detec tion was estimated (to the nearest meter) for all birds seen or heard. Counts were conduc ted only on days when the wind was < 20 kph and it was not raining. All observers were e xperienced and had been trained for distance estimation and species identification. For the analysis, I excluded non-pass erine birds, flyover detections, and unidentified passerines. I chose to exclude non-passerines both because they are often not detected well by point count methods a nd because many of them (e.g., woodpeckers)
11 are associated with vegetation characteristics (e.g, snags) that were not captured in these measurements of vegetation. Additionally, I only used bird detections within 100 m of the point, and assumed that the ability of an observer to detect birds within 100 m of the station was equivalent among ve getation types (Smucker et al . 2005). For all analyses, I used presence/absence (0,1) during a 5-minut e survey as a measure of occurrence for each species. Statistical Analyses I began by plotting vegetation volume agai nst the conifer proportion to visualize the extent of variation across these two dime nsions of vegetation. Predictably, maximum vegetation volume was less in broadleaf stands than it was in conife r stands (Fig. 2-2) because hardwood species in this area do not grow as tall as the conifer species. To facilitate the discussion of this vegetation space, I divided vegetation volume into three strata (low, moderate, and high) and vegetati on composition into three strata (broadleaf, mixed, and conifer). Because there were no stations in the high volume broadleaf stratum, there were a total of eight strata . Because I recognized that other important vegetation characteristics may covary with vegetation volume and broadleaf-conifer composition, I described the vegetation characte ristics of these stra ta. For each of these strata I calculated average maximum DBH, ca nopy height, plant species richness, and the frequency of six vegetation taxa ( Ceanothus spp., Arctostaphylos spp., Quercus spp., Arbutus menziesii , Pseudotsuga menziesii , Abies spp.) that are indi cative of dominant vegetation types in the region (Wh ittaker 1960, Huff et al. 2005). Model selection To evaluate the ability of my measures of physiognomy and floristics to predict the distribution of each bird species across this two dimensional vegetation space, I used
12 logistic regression. These generalized lin ear models with a logit link and binomial distribution included parameters for an inter cept and the effects of vegetation volume and composition; logit[ Pi] = 0 + cC + vV + cvCV + ccC2 + vvV2 ; where P is the probability of a bird being detected at the i th station, and C and V are the values for the broadleaf-conifer and vegeta tion volume index at that station. The parameter 0 is the model intercept, c is a parameter describing the main effect of composition, v describes the main effect of vegetation volume, cv is the interactive effect of these variables, and cc and vv are the quadratic effects of these variables. For species that occurred on at le ast 5% of the stations, I used a step-wise model selection procedure (Venables and Ripley 2002) that started with this most complex model and removed and added terms until the model with the best Akaike Information Criterion (AIC) was found. Generalized linear mode ls were fit using the software R (R Development Core Team 2005) Model performance Logistic regression models the probabili ty (between 0 and 1) of an event occurring given a set of pred ictor variables. With a th reshold probability, below which an event is predicted to be absent, and above which the event is pr edicted to present, these models can be used to generate predictio ns that can be compared to observed data. These predictions can be (1) true positives, (2) true negatives, (3) false positives, or (4) false negatives. Summarizing the frequency of the four outcomes provides information about model performance. Two frequently used indices of m odel performance are sensitivity and specificity. Sensitivity is the proportion of observed positives correctly predicted; specificity is the proportion of observed negatives correctly predicted (Liu et
13 al. 2005). However, these indices can only be calculated after a threshold value has been chosen. To assess the performance of these models independently of threshold values, I used Receiver Operating Characteristic (ROC ) curves (Cantor et al. 1999, Liu et al. 2005). ROC curves express the performan ce of a classification model by plotting sensitivity versus 1-specificity across threshold values ranging from 0 to 1. Good models maximize sensitivity and minimize 1-specificit y, such that the curve approaches the upper left corner of the plot. For models with no predictive power, the curve is simply a diagonal line from the lower left to upper right. The Area Under the Curve (AUC) represents model performance ranging from 1.0 (a perfect model) to 0.5 (a diagonal line; Fig. 2-3). Generally, the predic tive ability of models can be roughly ranked as fail (AUC = 0.5 to 0.6), poor (AUC = 0.6 to 0.7), fair (AUC = 0.7 to 0.8), good (AUC = 0.8 to 0.9), and excellent (AUC = 0.9 to 1.0) (Swets 1988). After selecting the best model, I then limited further analyses to models with an AUC value > 0.65. All statistics were conducted in R (R Development Core Team 2005). Visual presentation of models For models with AUC > 0.65 I used cont our plots to present the predicted probability of occurrence as a function of vegetation volume, vegetation composition, and their interactive and quadratic terms. I then used the ROC curves to identify a threshold value that maximized the sum of se nsitivity and specificity. I considered all areas of vegetation space above this threshol d to be â€œoccupiedâ€ habitat and the vegetation area below this threshold to be â€œunoccupiedâ€. When > 50% of a stratum was above the threshold, I classified that stratum as occupi ed by the species. I then grouped species together into habitat associati ons on the basis of the vegetatio n strata they were predicted to occupy. To compare the model results to a smoothed representation of the data, I used
14 binomial kriging (McNeill 1991, Gotway and Wo lfinger 2003) to calculate a probability surface that was based entirely upon the â€œspatia lâ€ pattern of occu rrences. Unlike the logistic regression models, this method does not force the prediction to be unimodal, and thus can characterize more complex patterns in the data set. The degree to which the contour lines produced by kriging match the contour lines of the logistic regression models provides a visual repres entation of the models ability to describe the patterns in the data. Results Vegetation characteristics varied pred ictably across the vegetation volume and composition gradients characterized by the eight strata. Max imum tree DBH and canopy height increased with vegetation volume (Tab le 2-2). Vegetation species richness was greatest in the mixed strata and lower in both broadleaf and conifer st rata (Table 2-2). The frequency of species indicative of chapa rral and oak woodlands were greatest in the broadleaf strata and lowest in the conifer strata. Similarly, major conifer taxa were most frequent in the conifer strata . Both sets of species were well represented in the mixed strata (Fig. 2-4). During the study, I detected 81 species of passe rine birds (Table 2-3). Of these, 44 were sufficiently abundant (occurring at > 5% of stations) to model (Table 2-4). The performance of these models as measured by the area under the ROC curve ranged from 0.55 to 0.90. Of the species for which I built models, 17 were below the AUC threshold of 0.65; thus, I consider these models to ha ve poor predictive power and do not consider them in subsequent analyses. For the 27 species with an AUC > 0 .65, step-wise model selection process identified 4 species that were best descri bed by the main and/or quadratic effects of
15 vegetation composition alone, 14 species that were describe d by main and/or quadratic effects of vegetation structur e and composition, and 9 species with models that included an interactive term of struct ure and composition (Table 2-4). Thus, for most (23/27) of these bird species both vegetation structure and composition were impor tant predictors of occurrence, either through a significant interaction term, or through significant main or quadratic effects of both stru cture and composition. The contour plots of predic ted probabilities (Fig. 2-5) demonstrated seven very different patterns of habitat association by th e 27 species (Table 2-5). Chaparral and oak woodland associates were those species, such as the Wrentit ( Chamaea fasciata ) and Bullockâ€™s Oriole ( Icterus bullockii ; Fig. 2-5), that were predicte d to occur primarily in the low and moderate volume broadleaf strata. Br oadleaf and mixed-coni fer associates were those predicted to occur in both the broadleaf strata and at le ast two of the mixed strata. The Green-tailed Towhee ( Pipilo chlorurus ) and Chipping Sparrow ( Spizella passerina ) were the only species that were define d only by vegetation volume; these species occurred only in low volume vegetation strata across a range of compositions (Fig 2-5). I classified these species as shrub associates (Table 2-5). Mature mixed-conifer species, such as the Brown Creeper ( Certhia americana , Fig. 2-5), were those that were associated with high volume mixed strata and moderate and high volume conifer. In contrast, the Red-breasted Nuthatch ( Sitta canadensis ), Yellow-rumped Warbler ( Dendroica coronata ), and Townsendâ€™s Solitaire ( Myadestes townsendii ; Fig. 2-5) were conifer associates that occurred in all conifer strata, regardless of the volume classification. Only one sp ecies, Mountain Chickadee ( Poecile gambeli ), was associated with low and moderate volume conifer strata without occurring in high volume conifer
16 strata. Similarly, only th e Golden-crowned Kinglet ( Regulus satrapa ; Fig. 2-5) occurred predominantly in moderate and high volume conife r strata; thus I classified it as a mature conifer species (Table 2-5). Discussion I described the vegetation of the Little Applegate watershed using two simple variables: vegetation volume and the propor tion of vegetation composed of conifer species. My analyses suggest that these va riables capture the diverse compositional and structural characteristics of vegetation in the Klamath Siskiyou ecoregion. At the broadleaf end of the composition gradient, chaparral habitats (low volume) and oak woodland (moderate volume) vegetation type s occur. These ve getation types are a unique aspect of vegetation in southern Oregon and northern Ca lifornia (Chappell and Kagan 2000, Huff et al. 2005). Many of the vegetation species in oak woodland/chaparral are able to persist in dry conditions and resprout quickly after fire. Where moisture and the absence of fire permit, conifer trees are able to colonize. At lower elevations this includes species such as Pinus ponderosa and Pseudotsuga menziesii , whereas Abies spp. is more frequent at higher elevations. Over time, forest stands may become dominated by conifers, especially on mesic sites and at higher elevation. Thus, the structural and compositi onal gradients I describe represent complex interactions of edaphics, disturbance hist ory, and elevation (Ski nner 1995, Odion et al. 2004). I used these gradients to predict spatial vari ation in the occurrence of bird species in the Little Applegate watershed. Four species were best modeled by vegetation composition alone (Table 2-4). These speci es occurred across a range of vegetation volumes, but were associated either with broadleaf-domi nated vegetation (e.g., Western
17 Wood-Pewee and Western Scrub Jay), coni fer dominated vegetation (Towensendâ€™s Solitaire), or intermediate sites (Wilsonâ€™s Warble r; Fig. 2-5). In contrast, there were no species that were best mode led by vegetation volume alone. The majority of the species (14 of 29) were best modeled by a co mbination of the vegetation volume and composition variables, either as main or quadra tic effects (Table 2-4). In some cases the combination of main and quadratic effects pr oduced complex distributional patterns (e.g., Western Tanager, Fig. 2-5). This composition by structure interaction wa s a significant parameter in the logistic regression models for 9 of 27 species I mode led (Table 2-4). This interaction has important management implications because it indicates that the effect of changing structure may vary depending on the compositiona l characteristics of a stand. The Brown Creeper demonstrates the importance of thes e interactive effects; decreasing vegetation volume in mixed stands is predicted to m ove the stand below the occupancy threshold much more quickly than in conifer-dominated st ands (Fig. 2-5). Even in the absence of a significant interaction, simply considering th e joint effects of structure and composition may still lead to inferences that would not be apparent if these effects were considered alone. Although the interaction term was not significant for the Hermit Warbler, the composition and structure variables act in concert because both main effects were significant. As a result, the effect of reducing vegetation volume relative to the occupancy threshold would be much different in mixed versus conifer-dominated stands (Fig. 2-5). These models make two major contributi ons to our understanding of the habitat associations of passerine birds. First, they demonstrate that complex patterns of
18 distribution can be modeled with relatively simple measures of forest structure and composition. The habitat associations generate d by these models (Tab le 2-5) effectively capture general knowledge of habitat requirement s for these birds. For example, my list of chaparral/oak woodland associates includes six species that were identified as oak woodland focal species by the Partners in F light bird conservation plan (Altman 2000). Similarly, these models capture the well -known associations of Golden-crowned Kinglets, Brown Creepers, and Red-breasted Nu thatches with conifer vegetation (Adams and Morrison 1993, Hansen et al. 1995, Sallabank s et al. 2002). These models can also be used to identify the degree to which sp ecies are associated with structural and compositional characteristics of vegetation. For example, California Towhees ( Pipilo crissalis ) occurred across a range of vegetati on volumes but were restricted to predominantly broadleaf stand composition (Fig. 2-5), whereas Green-tailed Towhees occur only in low volume stands but across a wi der range of composition (Fig. 2-5). This difference is consistent with distributional patterns at larg er spatial scales; Green-tailed Towhees occur in a wide variety of shrub environments, including shrub-steppe, montane shrub-fields, and chaparral, whereas California Towhees are restricted to oak woodland and chaparral (Dobbs et al. 1998, Kunzmann et al. 2002). Second, the approach I have used builds upon previous efforts to describe birdhabitat associations by explicitly modeling the interaction between structural and compositional variables. Considering such in teractive effects may explain the apparently inconsistent responses of some bird species to habitat modification. For example, in a literature review, Kotliar et al. (2002) listed a number of species that have been observed to both increase and decrease after wildfire . Based on the models presented here, I
19 propose that such variation may arrive in two ways. First, if sites differ in vegetation composition then the effect of reducing ve getation volume may vary among sites. For example, reducing vegetation volume increases the predicted probability of occurrence for Black-headed Grosbeaks ( Pheucticus melanocephalus ) in broadleaf and mixedconifer stands, but increases the probability of occurrence in conifer dominated stands (Fig. 2-5). Second, even when sites are id entical, the effects of changing vegetation structure may differ if they vary in the de gree to which they change composition. For example, the effect of reducing vegetation volume on the probability of occurrence of Black-headed Grosbeaks would be positive if mostly conifers were removed, but negative if mostly broadleaf species were removed (Fig. 2-6). For approximately 45% of the species I investigated, there were no significant terms in the model, or the pr edictive performance of the m odel was poor (AUC < 0.65). I propose two possibilities that w ould explain these modelsâ€™ failu re to predict occurrence accurately. For some species, vegetation volume and proportion of conifer vegetation probably fail to capture the stand-level habita t characteristics that are important to these species. For other species, larger-scale (p atch or landscape) characteristics may be important (Cushman and McGarigal 2004). Thus, although vegetation volume and proportion of conifer vegetation are important, they cannot predict th e distribution of all species. Conclusions These results demonstrate that physiognomy and floris tics can have interactive effects on patterns of bird distribution. Thes e interactions have important implications for the management and conservation of forest bi rds. The ability of forest birds to persist after management or disturbance may de pend not only on the changes in vegetation
20 volume, but also on floristic composition. In fa ct, of the species I i nvestigated, four were associated with vegetation composition alone while none were associated with vegetation structure alone. These models suggest that species associated w ith conifer vegetation will be less affected by the reduction of vegeta tion volume in pure conifer stands than in mixed-conifer stands. More extraordinar ily, reducing vegetation volume may in fact have positive effects on bird abundance in co nifer habitats, but negative effects in broadleaf habitats. I encourage both research ers and managers to consider the potential for interactive effects of physiognomics a nd floristics when designing studies and management plans.
21 Table 2-1. Frequency of occurre nce for vegetation taxa at 979 point count stations in the Little Applegate watershe d of southern Oregon. Taxa Frequency Type Douglas-fir, Pseudotsuga menziesii 0.740 Conifer Fir spp., Abies spp. 0.383 Conifer Incense cedar, Calocerdrus decurrens 0.289 Conifer Yew, Taxus brevifolia 0.062 Conifer Mountain hemlock, Tsuga mertensiana 0.021 Conifer Western white pine, Pinus monticola 0.017 Conifer Sierra juniper, Juniperus occidentalis 0.016 Conifer Sugar pine, Pinus lambertiana 0.006 Conifer Western hemlock, Tsuga heterophylla 0.003 Conifer Ponderosa or Jeffrey pine, Pinus ponderosa/jeffreii 0.583 Conifer Madrone, Arbutus menziessii 0.520 Broadleaf Ceanothus spp., Ceanothus spp. 0.433 Broadleaf Oak spp., Quercus spp. 0.399 Broadleaf Manzanita spp., Arctostaphylos spp. 0.335 Broadleaf Ocean spray, Holodiscus discolor 0.320 Broadleaf Poison oak, Toxicodendron diversiloba 0.232 Broadleaf Willow spp., Salix spp. 0.196 Broadleaf Bigleaf maple, Acer macrophyllum 0.168 Broadleaf Gooseberry/currant spp., Ribes spp. 0.139 Broadleaf Hazelnut, Corylus cornuta 0.115 Broadleaf Alder spp., Alnus spp. 0.096 Broadleaf Barberry spp., Berberis spp. 0.096 Broadleaf Common snowberry, Symphoricarpos albus 0.093 Broadleaf Rose spp., Rosa spp. 0.092 Broadleaf Serviceberry. Amelanchier alnifolia 0.085 Broadleaf Elderberry spp., Sambucus spp. 0.066 Broadleaf Cherry or plum spp., Prunus spp. 0.064 Broadleaf Mountain mahogany, Cercocarpus betuloides 0.053 Broadleaf Giant chinquapin, Chrysolepis chrysophylla 0.052 Broadleaf Canyon live Oak, Quercus chrysolepis 0.044 Broadleaf Ash sp., Fraxinus spp. 0.041 Broadleaf Mock orange, Philadelphus lewisii 0.031 Broadleaf Mountain maple, Acer glabrum 0.020 Broadleaf Common rabbit-brush, Chrysothamnus nauseosus 0.015 Broadleaf Sage spp., Artemesia spp. 0.014 Broadleaf Quaking aspen, Populus tremuloides 0.013 Broadleaf
22 Table 2-1. Continued Taxa Frequency Type Dogwood spp., Cornus spp. 0.012 Broadleaf Poplar or cottonwood, Populus spp. 0.008 Broadleaf Oregon boxwood, Paxistima myrsinites 0.007 Broadleaf Cascara spp., Rhamnus spp. 0.006 Broadleaf Nine-bark, Physocarpus capitatus 0.005 Broadleaf Mountain ash, Sorbus scopulina 0.005 Broadleaf Mountain mahogany, 0.004 Broadleaf Bitterbrush, Purshia tridentata 0.003 Broadleaf Red huckleberry, Vaccinium parvifolium 0.003 Broadleaf Creeping snowberry, Symphoricarpos mollis 0.002 Broadleaf Water birch, Betula occidentalis 0.001 Broadleaf Oso berry, Oemlaria cerisiformis 0.001 Broadleaf Rhododendron/azalea spp, Rhododendron spp. 0.001 Broadleaf
23 Table 2-2. Vegetation characte ristics measured in eight structural and compositional strata in the Little Appl egate watershed of southern Oregon. Measurements are means with standard errors. Vegetation strata N Maximum tree DBH (cm) Canopy height (m) Vegetation richness Low-volume broadleaf 90 35 2.2 10 0.6 4.7 0.2 Moderate-volume broadleaf68 55 2.6 20 0.7 6.2 0.2 Low-volume mixed 80 44 2.2 15 0.7 6.8 0.3 Moderate-volume mixed 234 60 1.6 23 0.4 6.9 0.2 High-volume mixed 80 77 2.6 30 0.6 7.5 0.2 Low-volume conifer 86 52 3.2 13 0.8 4.8 0.2 Moderate-volume conifer 229 68 1.9 22 0.4 5.4 0.2 High-volume conifer 112 90 2.8 31 0.6 4.7 0.2
24 Table 2-3. Frequency of occurrence for 81 bird species at 979 point c ount stations in the Little Applegate watershed of southern Oregon. Species Frequency American Crow, Corvus brachyrhynchos 0.004 American Dipper, Cinclus mexicanus 0.013 American Goldfinch, Carduelis tristis 0.006 American Robin, Turdus migratorius 0.299 Ash-throated Flycatcher, Myiarchus cinerascens 0.010 Barn Swallow, Hirundo rustica 0.002 Black-capped Chickadee, Poecile atricapilla 0.030 Bewick's Wren, Thryomanes bewickii 0.036 Blue-gray Gnatcatcher, Polioptila caerulea 0.010 Brown-headed Cowbird, Molothrus ater 0.061 Black-headed Grosbeak, Pheucticus melanocephalus 0.476 Brewer's Blackbird, Euphagus cyanocephalus 0.006 Brown Creeper, Certhia Americana 0.062 Black-throated Gray Warbler, Dendroica nigrescens 0.113 Bullock's Oriole, Icterus bullockii 0.033 Bushtit, Psaltriparus minimus 0.077 Cassin's Finch, Carpodacus cassinii 0.033 California Towhee, Pipilo crissalis 0.040 Cassin's Vireo, Vireo cassinii 0.199 Chestnut-backed Chickadee, Poecile rufescens 0.106 Cedar Waxwing, Bombycilla cedrorum 0.003 Chipping Sparrow, Spizella passerine 0.065 Clark's Nutcracker, Nucifraga Columbiana 0.002 Common Raven, Corvus corax 0.050 Dark-eyed Junco, Junco hyemalis 0.478 Dusky Flycatcher, Empidonax oberholseri 0.212 European Starling, Sturnus vulgaris 0.010 Evening Grosbeak, Coccothraustes vespertinus 0.007 Fox Sparrow, Passerella iliaca 0.028 Golden-crowned Kinglet, Regulus satrapa 0.202 Golden-crowned Sparrow, Zonotrichia atricapilla 0.001 Gray Jay, Perisoreus Canadensis 0.019 Green-tailed Towhee, Pipilo chlorurus 0.035 Hammond's Flycatcher, Empidonax hammondii 0.008 Hermit Thrush, Catharus guttatus 0.135 Hermit Warbler, Dendroica occidentalis 0.463 House Finch, Carpodacus mexicanus 0.004 House Wren, Troglodytes aedon 0.110
25 Table 2-2. Continued Species Frequency Hutton's Vireo, Vireo huttoni 0.049 Lazuli Bunting, Passerina amoena 0.220 Lesser Goldfinch, Carduelis psaltria 0.080 Lincoln's Sparrow, Melospiza lincolnii 0.026 MacGillivray's Warbler, Oporornis tolmiei 0.208 Mountain Bluebird, Sialia currucoides 0.003 Mountain Chickadee, Poecile gambeli 0.136 Nashville Warbler, Vermivora ruficapilla 0.503 Northern Rough-winged Swallow, Stelgidopteryx serripennis 0.002 Oak Titmouse, Baeolophus inornatus 0.011 Orange-crowned Warbler, Vermivora celata 0.012 Olive-sided Flycatcher, Contopus cooperi 0.128 Pine Siskin, Carduelis pinus 0.126 Pacific-slope Flycatcher, Empidonax difficilis 0.158 Purple Finch, Carpodacus purpureus 0.123 Purple Martin, Progne subis 0.001 Red-breasted Nuthatch, Sitta Canadensis 0.424 Red Crossbill. Loxia curvirostra 0.002 Rock Wren, Salpinctes obsoletus 0.005 Red-winged Blackbird, Agelaius phoeniceus 0.008 Song Sparrow, Melospiza melodia 0.030 Spotted Towhee, Pipilo maculates 0.318 Steller's Jay, Cyanocitta stelleri 0.327 Swainson's Thrush, Catharus ustulatus 0.007 Townsend's Solitaire, Myadestes townsendi 0.084 Townsend's Warbler, Dendroica townsendi 0.019 Tree Swallow, Tachycineta bicolor 0.014 Varied Thrush, Ixoreus naevius 0.028 Violet-green Swallow, Tachycineta thalassina 0.001 Warbling Vireo, Vireo gilvus 0.136 White-breasted Nuthatch, Sitta carolinensis 0.032 White-crowned Sparrow, Zonotrichia leucophrys 0.001 Western Bluebird, Sialia mexicana 0.017 Western Kingbird, Tyrannus verticalis 0.002 Western Scrub-Jay, Aphelocoma californica 0.076 Western Tanager, Piranga ludoviciana 0.359 Western Wood-pewee, Contopus sordidulus 0.139
26 Table 2-2. Continued Species Frequency Wilson's Warbler, Wilsonia pusilla 0.052 Winter Wren, Troglodytes troglodytes 0.073 Wrentit, Chamaea fasciata 0.088 Yellow-breasted Chat, Icteria virens 0.011 Yellow-rumped Warbler, Dendroica coronata 0.345 Yellow Warbler, Dendroica petechia 0.002
27 Table 2-4. Parameter values for logistic regression models identified using stepwise regression and their Area Under the Curve (AUC) values derived from receiver operating characteristic plot s for predictive models of bird distribution in the Little Applegat e watershed in southern Oregon. Species AUC Best model Composition only models Western Wood-Pewee 0.685 -0.69 2.44 C Townsendâ€™s Solitaire 0.672 -3.27 + 1.94 C 2 Western Scrub-jay 0.881 0.1 11.26 C + 6.45 C 2 Wilsonâ€™s Warbler 0.655 -3.43 + 5.46 C 7.02 C 2 Composition and structure effects models Bullockâ€™s Oriole 0.865 -1.13 4.79 C 2.2 V Wrentit 0.784 -0.47 3.12 C 1.79 V Lesser Goldfinch 0.77 -0.67 2.76 C 1.93 V Chipping Sparrow 0.714 -1.25 0.65 C 3.63 V Lazuli Bunting 0.713 0.22 -2.35 C â€“ 1.04 V Red-breasted Nuthatch 0.709 -2.12 + 2.67 C + 0.84 V White-breasted Nuthatch 0.806 -2.62 4.79 C + 6.21 V 7.14 V 2 Black-throated Gray Warbler 0.773 -2.09 3.94 C + 9.74 V 10.33 V 2 Hermit Warbler 0.75 -3.98 + 9.79 C + 2.17 V 6.54 C 2 Winter Wren 0.699 -5.12 + 5.54 C + 2.53 V -4.4 C 2 Bushtit 0.748 -2.05 + 4.48 x2 -3.34 C 2-7.13 V 2 Yellow-rumped Warbler 0.698 -1.66 + 2.72 C 2.54 V + 2.26 V 2 Pacific Slope Flycatcher 0.678 -3.69 + 6.34 C 5.24 C 2 + 2.03 V 2 Western Tanager 0.667 -1.44 + 1.33 C + 5.73 V â€“ 3.09 C 2 3.09 V 2 Composition*structure interaction models Spotted Towhee 0.738 0.44 1.92 C + 1.16 V 3.43 CV Mountain Chickadee 0.693 -1.71 + 0.87 C 5.49 V + 5.51 CV Brown Creeper 0.683 -6.11 + 4.4 C + 5.54 V -6.19 CV California Towhee 0.909 -1.11 -13.81 C + 5.75 V -15.5 V 2 -15.5 CV Bewickâ€™s Wren 0.904 -0.91 -10.2 C + 2.49 V -15.35 V 2 -15.35 CV Green-tailed Towhee 0.802 -1.49 + 5.34 C -13.62 V -8.68 C 2 -8.68 CV Golden-crowned Kinglet 0.782 -4.26 + 5.49 C 0.42 V -3.48 C 2 -3.48 CV Black-headed Grosbeak 0.723 -0.72 + 1.64 C + 7.52 V -2.73 C 2 -2.73 V 2 -2.73 CV Nashville Warbler 0.701 -2.09 + 8.53 C + 4.28 V -7.97 C 2 -7.97 V 2 -7.97 CV Models with AUC < 0.65 Warbling Vireo MacGillivrayâ€™s Warbler Olive-sided Flycatcher Brown-headed Cowbird
28 Table 2-4. Continued Species AUC Best model Dusky Flycatcher Hermit Thrush Stellerâ€™s Jay Huttonâ€™s Vireo Cassinâ€™s Finch House Wren Dark-eyed Junco Cassinâ€™s Vireo Chestnut-backed Chickadee American Robin Purple Finch Pine Siskin Common Raven
29 Table 2-5. Habitat associations for passerine birds in the Little Applegate watershed of southern Oregon based on the vegetation st rata they were predicted to occupy from logistic regression. Species Vegetation strata occupied Chaparral/oak woodland associates Bewick's Wren low and moderate volume broadleaf Bullock's Oriole low and moderate volume broadleaf California Towhee low and moderate volume broadleaf Western Scrub-Jay low and moderate volume broadleaf Western Wood-Pewee low and moderate volume broadleaf White-breasted Nuthatch low and moderate volume broadleaf Wrentit low and moderate volume broadleaf, low volume conifer Broadleaf and mixed-conifer associates Black-headed Grosbeak low and moderate volume broadleaf, low, moderate, and high mixed Black-throated Gray Warbler low and moderate volume boadleaf, moderate volume mixed Bushtit low and moderate volume broadleaf, moderate volume mixed Lazuli Bunting low and moderate volume broadleaf, low and moderate volume mixed Lesser Goldfinch low and moderate volume broadleaf, low and moderate volume mixed Spotted Towhee low and moderate volume broadleaf, low, moderate, and high volume mixed Wilsonâ€™s Warbler Low and mode rate volume broadleaf, low, moderate, and high volume mixed Western Tanager moderate volume broadleaf, moderate and high volume mixed Nashville Warbler low, moderate, and high volume mixed, moderate volume conifer Shrub associates Green-tailed Towhee low vol ume broadleaf and mixed Chipping Sparrow low volume br oadleaf, mixed, and conifer Mature mixed-conifer and conifer associates Brown Creeper high volume mixed, moderate and high volume conifer Hermit Warbler high volume mixed, low, moderate and high volume conifer Pacific-slope Flycatcher hi gh volume mixed and conifer
30 Table 2-5. Continued. Species Vegetation strata occupied Winter Wren high volume mixed and conifer Conifer associates Red-breasted Nuthatch low, moderate, and high volume conifer Yellow-rumped Warbler low, moderate, and high volume conifer Townsendâ€™s Solitaire low, moderate, and high volume conifer Young conifer associate Mountain Chickadee low and moderate volume conifer Mature conifer associates Golden-crowned Kinglet high volume conifer
31 Figure 2-1. The study area in the Little A pplegate valley, Oregon, USA, showing the spatial distribution of stations wher e information on bird abundance and vegetation characteristics were collected.
32 Figure 2-2. Sampling locations across a physiogn omic gradient of vegetation volume (0 = low volume, 1 = high volume) and a fl oristic gradient of broadleaf-conifer composition (0 = all broadleaf vege tation, 1 = all conifer vegetation).
33 Figure 2-3. Receiver-operating characteristic curves for logistic models predicting probability of occurrence for four species of passerines. The line describes the change in sensitivity and 1-specific ity as a function of possible threshold values (ranging from 0-1). The threshold probability that maximizes the sum of sensitivity and specificity is marked at the upper left hand corner of each curve. The Golden-crowned Kinglet a nd Lesser Goldfinch are examples of models that perform well (area unde r the curve > 0.70), whereas Huttonâ€™s Vireo and MacGillivrayâ€™s Warbler are ex amples of models that have poor predictive ability (area under the curve < 0.65).
34 Figure 2-4. Frequency of occu rrence of six vegetation taxa across the eight vegetation strata. Abbreviations: cean = Ceanothus spp., arct = Arctostaphylos spp.., quer = Quercus spp., arme = Arbutus menziessii , psme = Pseudotsuga menziesii , and abie = Abies spp. Frequency of Plant taxa Frequencyofoccur r ence
35 Vegetation volume index Broadleaf-conifer composition Figure 2-5. Predicted probability of occurr ence as a function of vegetation volume and vegetation composition passerine birds in the Little Applegate Valley of southern Oregon. Dashed red lines are the threshold value, identifying occupied and unoccupied stations, es timated with the receiver-operating characteristic curves. Bold black lines are the contours of predicted probabilities from the logistic regression models (Table 2-4). Thin blue lines are the contours of predicted probabi lities generated by binomial kriging.
36 Vegetation volume index Broadleaf-conifer composition Figure 2-5. Continued
37 Vegetation volume index Broadleaf-conifer composition Figure 2-5. Continued
38 Vegetation volume index Broadleaf-conifer composition Figure 2-5. Continued
39 Broadleaf-conifer composition Vegetation volume index Figure 2-5. Continued
40 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 BeforeAfterProbability of occurence A B C D Figure 2-6. A figure illustrating that treatm ents designed to reduce vegetation volume by the same amount are predicted to have ve ry different effects on the probability of occurrence of Black-headed Grosbeaks. The three treatment options would reduce vegetation volume of the stand (A) by removing conife r trees only (B), removing both conifer and broadleaf trees (C), or removing broadleaf trees only (D).
41 CHAPTER 3 WILDFIRE EFFECTS ON VEGETATION AND PASSERINE BIRDS IN A MIXEDCONIFER FOREST: COMBINING HABITAT MODELS WITH A NATURAL EXPERIMENT Introduction Disturbance is a fundamental process in fluencing community properties such as diversity, food-chain lengt h, and stability (Connell 1978, Collins 2000, Post 2002). However, the strength of disturbance e ffects is highly dependent on community composition and the timing and extent of disturbance (Wootton 1998, Collins 2000, Mackey and Currie 2001). As a result, it ha s been difficult to generalize when and how much disturbance will cha nge ecological communities. Large, infrequent disturbances are respons ible for long-lasting changes in forest structure and composition (Foster et al. 1998). These changes are recognized as a critical element of bird community dynamics because many bird species are associated with postdisturbance vegetation characte ristics (Brawn et al. 2001). However, in many regions of western North America, fires burn with consid erable spatial (within fire) and temporal (among fire) variability (Agee 1993), creating co mplex mosaics of vegetation patches. In these systems, changes in bird abundance ar e likely to be driven by a combination of post-fire patterns of vegeta tion characteristics (Huff et al. 2005), food availability (Johnson and Sherry 2001), and predati on (Martin 1988, Saab et al. 2004). Understanding the importance of these mechan isms is difficult because the relative strengths may change over time and vary with the spatial scale of disturbance. As a result, natural fire effects are best studied over long temporal scales (years or decades)
42 and at large spatial scales (thousands of hectares). These spatial and temporal characteristics of large, infrequent distur bances make it difficult to apply traditional experimental designs to quan tify disturbance effects on natu ral systems (van Mantgem et al. 2001). Quantifying post-disturbance cha nges is further complicated because many animal communities are spatially and tem porally dynamic even in the absence of disturbance; thus, the detection of distur bance effects is diffi cult (Parker and Wiens 2005). Given the infeasibility of traditional experimental designs, which are often logistically or ethically im possible, several approaches can be used to strengthen inferences about the effects of large disturbances. One a pproach is the Before-After Control-Impact (BACI) approach to intervention analysis (Osenberg et al. 1994, StewartOaten and Bence 2001, Parker and Wiens 2005). This approach compares sequential observations of a system before and after an intervention event to observations over the same period in which the intervention even t did not occur. Increasingly, these BACI approaches are being used to quantify the e ffects of wildfire (Smu cker et al. 2005) and large-scale forest management (Crome et al. 1996, Hagar et al. 2004, Camprodon and Brotons 2006) on bird populations . Another approach is to use models to evaluate the degree to which observed differences are c onsistent with the changes in predictor variables that occurred over the same time pe riod. This approach has been applied to large-scale patterns that are difficult or impo ssible to manipulate experimentally, such as the effects of recent climate change on the di stribution of butterflies (Hill et al. 1999). Here, I use five years of data (1 pre-fi re and 4 post-fire) on bird abundance and vegetation to evaluate the effects of this fi re on bird abundance. The objectives were to
43 (1) identify species that increased or decreased as a result of this mixed-severity fire, (2) investigate how these changes in abundance ch anged as a function of time since fire, and (3) evaluate the degree to which changes in bird abundance agreed with predicted changes based on models that used vegetati on structure and composition to predict bird abundance. Methods Study Area This study was conducted in the Little Applegate Valley in Jackson County, approximately 17.5 km southwest of As hland, Oregon (524000E, 4671000N, UTM zone 10, NAD27; Fig. 3-1), in the Klamath-Siski you region. Elevation of the watershed ranges from approximately 400 to 2250 m above sea level. The forest vegetation of the Klamath-Siskiyou region is diverse (Whitta ker 1960) and includes a variety of both conifer and hardwood species. At lower elev ations and on drier sites, vegetation is dominated by broadleaf species, in the form of oak woodlands with both Oregon white oak ( Quercus garryana ) and California black oak ( Quercus kellogii ) or sclerophyllous shrublands (chaparral), in which broadleaf shrubs of the general Ceanothus, Arctostaphyllus, and Baccharis are abundant (Huff et al. 2005). On wetter sites or middle elevations the vegetation becomes a mix of conifers, including Douglas-fir ( Pseudotsuga menziesii ), ponderosa pine ( Pinus ponderosa ), incense-cedar ( Calocedrus decurrens ), and white fir ( Abies concolor ); and hardwoods, including tanoak ( Lithocarpus densiflorus ), Pacific madrone ( Arbutus menziesii ), Canyon Live Oak ( Quercus chrysolepis ), California black oak, Oregon wh ite oak, and big-leaf maple ( Acer macrophyllum ). The relative composition of these sp ecies varies with elevation, aspect, and soils. Generally, these forest types corr espond to the Douglas-fir, Mixed Evergreen
44 Hardwood, or White Fir Types (Franklin a nd Dyrness 1973, Huff et al. 2005). Firerelated studies in these vegeta tion types show a mix of fire severities, frequencies, and sizes, typically characteristic of low and mixed-severity fire regimes (Agee 1991, Taylor and Skinner 1998). Over time, such mixed-seve rity fires can create forests with multiple age classes, often with Dougl as-fir or ponderosa pine as an emergent canopy above various hardwoods. In combin ation with diverse topography a nd soil characteristics, fire creates a complex mosaic of forest patches that vary in size, age, and species composition (Whittaker 1960). This study design capitalizes on an unplanned wildfire that burned in an area where surveys of bird abundance and vegetation structure and composition had been conducted before the fire burned. In the spring of 2001, 979 stations were surveyed in the Little Applegate Valley as part of an investigation into the effect s of vegetation structure and composition on bird distribution (see previous chapter). These stati ons were spaced 250 m apart on transects of 10-18 stations along s econdary and tertiary roads and trails. Although locating point c ount stations on roads may increas e the number of detections of birds associated with shrubs and edges, this method generally does not have large effects on the ability to detect patterns of habita t association with stand-level vegetation characteristics (Hanowski and Niemi 1995). La ter in the summer of 2001, the Quartz fire burned 2,500 ha near the center of this study area across an elevation range from 7701920 m. The area burned included 52 stations. As controls, I selected a subset of 77 stations that were > 500 m outsi de the fire perimeter, matched with respect to elevation (700-2000 m), and were not outside of the ra nge of variation for vegetation volume or
45 broadleaf-conifer composition (see below) of the burned stations (Fig. 3-2). These stations were re-surveyed in 2002, 2003, 2004, and 2005. Fire Effects The Quartz Fire burned between the 9th and 31st of August 2001 in a valley that faces to the north with slopes in the burned area ranging from 0 to 63 degrees. As is typical of fires in this region, fire severity was highly variable acr oss this area (Alexander et al. In press). Fire se verity was mapped by the Burned Area Emergency Rehabilitation team (BAER) of the United States Forest Se rvice (USFS). These maps identified severity as low, medium, or high, based on post-burn aerial photography and/or observations from the field. In areas categorized as low sever ity, fires usually spread rapidly and residence time is short, or these areas may be skipped over by fire. In either case, vegetation is only lightly scorched vegetation and there is mi nimal tree. The effects of the fire in areas categorized as moderate severity are more extr eme; soil structure remains intact but fine fuels near the ground are mostly consumed a nd 40 to 80% of trees are killed, but retain brown needles that later fall to the forest floor. In areas of high fire severity, fire and heat residence times are longest and ground cover may be completely consumed. In these high severity areas, soil structure may be al tered if soil organic matter is consumed. Because tree crowns are completely consume d, few leaves or needles remain on trees and mortality is often 100%. Of the 2,500 ha Quartz Fire, the BAER map classified it as 23% low severity, 36% moderate severity, and 41% hi gh severity (Alexander et al. In press). Vegetation Protocol To generate measurements of vege tation volume and broadleaf-conifer composition, I combined several variables th at were available from the vegetation surveys conducted at all point count stat ions in 2001. These vegetation data were
46 collected using variable radius (20-50 m) relev plots (R alph et al. 1993). Within the circular plot defined by the ra dius, eight structural variab les were recorded: total tree cover, total shrub cover, maximum tree di ameter, minimum tree diameter, and average canopy height. Total cover of the tree stratum (all vegetation 5 m) and shrub stratum (all vegetation 0.5 m and < 5 m) were estimated and assigned the mid-point one of 5 cover classes (0-2.5, 2.5-25, 25-50, 50-75, and 75-100%). Maximum and minimum diameter at breast height (DBH) of trees in the tree layer was visually estimated, as was the average height of the canopy (m). To quan tify vegetation structure, I used an index of total vegetation volume as a single measure. The measure was calculated by multiplying the canopy height measurement by the cover measurement for the tree layer. I divided all values by the maximum observed vegetation vol ume to create a standardized measure that varied from 0 to 1. The percent cover of indivi dual tree and shrub taxa was also recorded using the 5 cover classes described above. These cover sc ores were used to develop an index of vegetation composition that expressed the proportion of the vegetation comprised of conifer (versus broadleaf) species. This measure was calculated as: Ci/( Ci+ Bi) where Ci are the cover values of all coniferous tree and shrub species at a station and Bi are the cover values of all broadleaf trees and shrubs at a station. Thus, the conifer proportion ranges from 0 (all tree cover was broa dleaf species) to 1 (all cover was conifer species). When all live vegetation was re moved from a station by fire, I used the vegetation composition score from the previous year as the scor e for that year rather than
47 setting it to 0 (implying that all vegetation wa s broadleaf). This occurred at fewer than 10 stations. Bird Surveys Each year, bird abundance was measured at all stations between 11 May and 2 July using standardized point count methodologies (Ralph et al. 1993). Five-minute bird counts were conducted between sunrise a nd 1000 PDT on each station. During each count the distance at the time of first detecti on was recorded for all birds seen and heard. Counts were conducted only on days when th e wind was < 20 kph and it was not raining. All observers were experienced and had been trained for distance estimation and species identification. For the analysis, I excluded non-pass erine birds, flyover detections, and unidentified passerines. Add itionally, I only used bird det ections within 100 m of the point, and assume that the ability of an obs erver to detect birds within 100 m of the station was equivalent among ve getation types (Smucker et al . 2005). To calculate total bird abundance per point, I su mmed the number of individuals of all species at each, but for all other analyses, I used presence/absence (0,1) as a measure of occurrence for each species. Statistical Analyses Vegetation data I used the vegetation data to describe changes in vegetation structure and composition that occurred after the fire . For six important vegetation taxa ( Ceanothus spp., Arctostaphylos spp., Quercus spp., Arbutus menziesii , Pseudotsuga menziesii , Abies spp.), I calculated the mean percent cover in each year of the study for burned and unburned areas. Additionally, I calculated the mean vegetation volume and composition
48 index for each year. I plotted vegetati on volume against the c onifer proportion to visualize the extent to which the stru cture and composition of burned and unburned stations changed over the course of the st udy within the context of the bird-habitat models that were developed from the exte nsive survey effort conducted in 2001 (see previous chapter). Bird data I limited this analysis to bird species that occurred at > 15% of the stations in the burned area in at least one year of the study. To evalua te changes in the proportion of stations occupied over the course of the st udy, I used generalized li near models with a logit link and binomial distribution with area (b urned versus unburned) and year (0-4) as predictor variables. Using likelihood ratio tests, I test ed three nested models of decreasing complexity. The mo st complex model allows the proportion of occupied sites to vary differently across years in burned and unburned areas. This model is consistent with a fire effect that decreases or increases the probability of detection in burned areas while detection probabilities do not occur in the same way at unburned stations. Without the interaction term, the simpler model allows for annual variation a nd consistent spatial variation between burned or unbur ned areas in all years (i.e ., a site difference between burned and unburned areas that is not created by the fire). The simplest model describes a situation in which the proportion of occupied sites varies among years, but not between burned and unburned areas. Thes e models were written as: 1) logit( Pi) = 0 + yY + bB + ybYB 2) logit( Pi) = 0 + yY + bB 3) logit( Pi) = 0 + yY,
49 where Pi is the probability of a bird occurring at the i th station, Y is a dummy variable identifying the year of the study (0-4), and B identifies the point as either burned or unburned. The parameter 0 is the model intercept, y is a parameter describing annual variation, b describes spatial varia tion between burned and unburned areas, and yb is an interaction effect that accounts for different pa tterns of annual variation depending on whether or not a point was burned. Because I had only a single year of prefire data, I could not quantify annual variation at burned and unburned before the fire. The close spatial proximity of the burned and unburned stations and th e selection of stations with relatively similar habitat features minimized the possibility that the stations would exhibit radically different patterns of annual fluctua tion in the absence of an effect of the fire. Thus, I assumed that year to year variability was consistent betw een burned and unburned areas in the absence of a fire and inferred fire effects on the proba bility of detection by first using the results of the likelihood ratio tests to identify sp ecies for which the interaction model was significantly better than the si mpler models. Then I plotted the proportion of occupied stations for burned and unburned areas in each year of the study. These plots could be separated into interactive effects as either â€œdirectionalâ€, when the probability of detection at burned stations was either gr eater or less than that at unburned stations in at least 3 of the four years after the fire, or â€œmixed,â€ wh en the probability of detection was greater at burned stations in one year, but greater at unburned stations the following year. In addition to evaluating which birds were a ffected by the fire, I wished to evaluate the degree to which the post-fire changes in the proportion of stations occupied were consistent with the degree to which the fire had changed the structural and compositional
50 characteristics of the vegetation at the points. In the absence of true replication (multiple fires), this approach can increase the st rength of the inference that the observed differences are consistent with the changes in predictor variables that occurred over the same time period. In the previous chapter, I used logistic regression to model the probability of detection for 27 passerine bi rds as a function of vegetation volume and composition with data collected during an exte nsive (979 stations) su rvey of the Little Applegate Valley. This project included the st ations where pre-fire data were collected for this analysis. Because the measures of vegetation volume and the proportion of vegetation composed of conifer species presente d here were identical to those used in the previous paper, I used the yearly vegeta tion measurements to generate a predicted probability of occurrence of a bird species at each station. I then averaged the probabilities of all stations in burned a nd unburned areas to generate a predicted proportion of occupied stations in each year . Because these models were based on a single year of data, they failed to capture the y ear to year fluctuations in the proportion of stations occupied. Thus, I focused on th e agreement between the observed difference between burned and unburned areas and the predicted difference between burned and unburned areas, rather than the absolute propor tion of occupancy. I treated each species as an independent replicate a nd used Pearsonâ€™s correlation as a measure of the strength with which predicted differences were correlat ed with observed differences in each year of the study. Results Vegetation Structure and Composition Vegetation composition changed after the fire . The most dramatic change was the decrease in Douglas-fir cover th at occurred in the first year after the fire (Fig. 3-2). In
51 contrast, the change in cover of broadleaf taxa was much less pronounced, and often was not more extreme than a decline in cover obs erved at the control stations (Fig. 3-2). However, in the years after the fire, the cover of several broa dleaf taxa, including Ceanothus spp., Quercus spp., and Arbutus menziesii (Pacific madrone), increased relative to cover at control sta tions (Fig. 3-2). As a result of the decrease in conifer cover and the maintenance, or increase, of broa dleaf cover, the vegetation at many burned stations became nearly completely dominate d by broadleaf species, and the overall mean for all burned points shifted toward greater br oadleaf composition in the four years after the fire (Fig. 3-3). In the burned area, fire had the expect ed effect of reducing total vegetation volume dramatically in the first year. In the following three year s, there was a slow increase in vegetation volume. Although a similar pattern was observed in the mean values of the control stations, the magnitude and directionality of this change was much less (Fig. 3-3). Furthermore, by the end of the study the control stations almost all fell within the range of variation of vegetation characteristics record ed within the first year of the study, whereas the burned stations had clearl y shifted outside of the original range of variation (Fig. 3-3). Bird Abundance At the burned stations, 55 bird species were detected in at least one year of the 5year study period. Of these, all but two species were also detected at least once in the unburned area during this same time period. These two exceptions were the Western Bluebird ( Sialia mexicana ) and White-crowned Sparrow ( Zonotrichia leucophrys ), neither of which was detected in burned area in the spring before the fire. Twenty-nine species met the abundance criteria for further analysis. Of thes e, eight species were best
52 described by a model with only annual variati on (Table 3-1). These species, such as the Purple Finch ( Carpodacus purpureus ) and the Warbling Vireo ( Vireo gilvus ), occupied a similar proportion of stations in both unburned and burned areas throughout the study (Fig. 3-4). For nine specie s, the best model included annua l variation in addition to a difference between burned and unburned stations (Table 3-1). For these species, such as the Dusky Flycatcher ( Empidonax oberholseri ), Western Tanager ( Piranga ludocviciana ), and Spotted Towhee ( Pipilo maculatus ), the spatial va riation between burned and unburned sites presumably existed even before the fire (Fig. 3-4). The interaction model was identifie d as the best model for 12 sp ecies (Table 3-1), suggesting that the annual variation differed between burned and unburned areas. The characteristics of this interaction varied among species (Fig. 3-4). The Chestnut-backed Chickadee ( Poecile rufescens ), Red-breasted Nuthatch ( Sitta canadensis ), Hermit Thrush ( Catharus guttatus ), Nashville Warbler ( Vermivora ruficapilla ), Black-throated Gray Warbler ( Dendroica nigrescens ), and Hermit Warbler ( Dendroica occidentalis ) all occurred at consistently fewe r of the burned stations after the fire compared to unburned stations. Howeve r, the difference between the proportion of burned and unburned stations occupied was ge nerally much greater in the second to fourth year after the fire than in the year immediately after the fire (e.g., Red-breasted Nuthatch, Hermit Thrush, and Hermit Warbler; Fig. 3-4). Two sp ecies, Cassinâ€™s Vireo ( Vireo cassinii ) and Brown Creeper ( Certhia americana , Fig. 3-4), exhibited patterns suggestive of a negative effect in the years immediately afte r the fire, but had recovered to pre-fire levels by the fourth ye ar of the study. The Pine Siskin ( Carduelis pinus )was the only species with a significant interacti on term that appeared to become more
53 widespread at burned stations after the fire (Fig. 3-4). Fo r the other three species, the patterns of distribution between burned and unbur ned areas varied from year to year and were difficult to interpret. Both American Robins ( Turdus migratorius ) and Dark-eyed Juncos ( Junco hymenalis ; Fig. 3-4) were more widespread at burned stations the first year after the fire, but very similar to the unbur ned stations in the following three years, suggesting a short-term response. The Yellow-rumped Warbler ( Dendroica coronata ; Fig. 3-4) was dramatically less abun dant at burned stations in th e third year after the fire. Are Observed Differences Consistent With Habitat Model Predictions? Although the model predictions were generally consistent with the direction of the observed differences, the ability of the mode ls to predict the absolute proportion of occupied stations was generally limited due to high year to year variation (for example the Hermit Warbler, Fig. 3-5). The ability of the habitat model to predict differences in occupancy of burned and unburned stations varied over the study. In the first year of the study, there was very little va riation predicted by the mode l, and although most species exhibited relatively little difference, th e observed difference for Western Tanager, Mountain Chickadee, Spotted Towhee, and Western Wood-Pewee ( Contopus sordidulus ) were greater than predicted by the model (Fig. 3-6). In the first y ear after the fire the correlation between observed and predicted differences increased, but was still not statistically significant. By the third year, this relationshi p was stronger an d statistically significant, but suggested a slope considerab ly greater than one. Observed negative differences for Hermit Warbler and Red-brea sted Nuthatch were much greater than predicted, as was the observed positiv e difference for the Lazuli Bunting ( Passerina amoena ; Fig. 3-6). In the following year (2004), the slope was st ill greater than one, with many species that responded negatively to th e fire showing greater observed differences
54 than predicted. However, in the final y ear of the study, the observed differences converged with the predicted differences such that the relationship between observed and predicted was highly signifi cant and close to 1:1. Discussion Changes in Vegetation I used the vegetation data to describe th e degree to which this natural wildfire changed vegetation volume and composition in the area immediately surrounding the stations. Predictably, tota l vegetation volume was dramatically reduced at burned stations in the first year after the fire, follo wed by a very gradual recovery. This recovery was driven by the growth of broadleaf vege tation. Unlike conifers, many broadleaf taxa, especially those that are adapted to fire, respr out after fires. As a result of the reduced conifer cover and the maintenance or increase in cover of broadleaf taxa, the vegetation composition became less dominated by conifer vegetation during the course of the study. In contrast, there was very little change in vegetation composition at the unburned stations during this same time period (Fig. 3-3) . There was considerab le variation in fire severity as evidenced by the fact that some stations remained within the range of variation of the pre-fire vege tation characteristics, whereas other stations were moved to much lower vegetation volumes and broadleaf dominated vegetation compositions that did not exist before the fire (Fig. 3-3). Both vegetation volum e and composition are important factors associated with patterns of site occupancy by many passerine bird species in this watershed (previous chapte r), suggesting that bot h factors should be considered when interpreting post-fi re changes in avian abundance.
55 Changes in Bird Abundance I concluded that 6 of 29 species exhib ited a pattern of occupancy that was consistent with decreased abundance in burne d areas and two additi onal species exhibited evidence of a short term (2 y ear) decrease (Fig. 3-4). Of th ese 8 species, nearly all of them have been shown to decrease after ei ther fire (Hutto 1995, Kreisel and Stein 1999, Smucker et al. 2005) or selective logging (Hansen et al. 1995, Sallabanks et al. 2002, Hayes et al. 2003) suggesting th at that these species are sensitive to conifer cover reduction. For the eight species that decreased after the fire, the difference between the burned and unburned areas did not become pronounced un til the second year af ter the fire (Fig. 3-7). Such time lags have been recognized as a complicating fact or in interpreting the response of birds to large-sc ale habitat modification (Wiens et al. 1986) and have been documented in several studies of bird response to fires (Moreira et al. 2003, Pons et al. 2003). This time-lag may result because of hi gh site fidelity. For many birds that have been studied, the decision to return to nest in the same location is greater if they successfully bred there in the previous year (Hoover 2003, Sedgwick 2004). Since the Quartz fire burned in August, well after the nesting season was over, many birds may have made habitat selection decisions in 2002 based on pre-fire habitat conditions, whereas habitat selection decisions after 2003 would be more likely to reflect post-fire habitat conditions. The only species for which I found statistical evidence consistent with increased abundance after the fire was the Pine Siskin. The interaction term was also significant for the Dark-eyed Junco, but this species was more abundant at burned stations only in the first year after the fire (Fig. 3-4). Smuc ker and co-workers (2005) documented increased
56 numbers of Dark-eyed Juncos in areas that burned with moderate severity, and other studies have described this species as mo re abundant in recently burned areas (Hutto 1995). Both Dark-eyed Juncos and Pine Sisk in occur across a wide range of vegetation structure and composition characteristics (previ ous chapter) and may therefore be able to take advantage of increased food availability in burned areas. A lthough not statistically significant, the proportion of stations where House Wrens ( Troglodytes aedon ) were detected increased consistently during the c ourse of the study at bur ned stations, but not at unburned stations (Fig. 3-4) , a pattern that is supporte d by other post-fire studies (Kotliar et al. 2002). Furthermore, this anal ysis did not include woodpeckers, which also often increase immediately after fire (Hutto 1995). None-the-less, there were more species that clearly decreased as a result of the fire than species that increased. This is consistent with other studies of fire effects on birds, for example, of 40 species analyzed by Sm ucker and co-workers (2005) five were found to have decreased significan tly (P < 0.05) but only one sp ecies increased significantly (though three species increase at the P < 0.10 level). One explanation for this pattern is that the colonization of recently burned area s by open-habitat birds is limited by dispersal (Brotons et al. 2005, Pons and Bas 2005). Ho wever, if post-fire increases in bird abundance were limited by recruitment, one would expect the observed difference between burned and unburned areas to be sy stematically lower than the predicted difference. This is not the case (Fig. 3-6), suggesting that in this system recruitment limitation is not a major factor for po st-fire patterns of bird abundance. In the absence of a significan t interaction, a sign ificant area term indicates that the proportion of occupied stations varied between burned and unburned areas, but probably
57 not as a result of the fire. Despite the effort to use control stations that were similar in terms of elevation and vegetation, I found such spatial differences for species including Western Wood-Pewee, Dusky Flycatcher, West ern Tanager, and Spotted Towhee (Table 3-1, Fig. 3-4). If pre-fire data had not b een collected, such differences would have incorrectly been interpreted as a fire effect . These patterns demonstr ate the importance of using pre-impact data to eval uate the effects of disturbanc es. Even in the absence of multiple years that can be used to quantify temporal variation (Stewart-Oaten and Bence 2001), a single year can still dram atically change the inferences. In addition to making inferences based on pa tterns of bird abunda nce at burned and unburned areas, I also compared these results to the predictions generated by habitat models that used characteris tics of vegetation structure a nd composition to predict the occurrence of bird species (pre vious chapter). Based on the characteristics of vegetation structure and composition measured in each year, the model predicted increasing differences between bird abundance at burne d and unburned stations over the study (Fig. 3-6). By the fourth year after the fire, th ese predicted differences matched well with the observed differences. However, in the first year after the fire I found relatively little concordance, probably as a result of the lagtime with which birds responded to habitat changes (Fig. 3-7). In the second and thir d year after the fire there was a strong correlation between the observed and predicte d differences, but the observed differences were more extreme (both positively and negatively) than the predicted differences. This suggests that short-term effects of fire were greater than what would be predicted based on patterns of vegetation in the immediate area of the station. One explanation for these patterns would be neophobic responses to novel habitat characteristics created by the fire
58 (Davis and Stamps 2004, Mettke-Hofmann et al. 2005). Alternatively, such a pattern could suggest that birds avoide d these areas because fire a ffected food availability or predator activity independently of its e ffects on vegetation stru cture and composition. These model results support the inference that the d ecreased abundance of Hermit Warblers and Red-breasted Nuthatches was inde ed a result of the habitat changes caused by the fire. Additionally, the directions of several non-si gnificant differences were consistent with the model predictions, incl uding the greater abundance of the Lazuli Bunting and Spotted Towhee and the decr eased abundance of Western Tanager and Golden-crowned Kinglet ( Regulus satrapa ; Fig. 3-6). Conclusions The ability of many previous studies to ma ke causal inferences about the effects of wildfire on bird abundance has been limited be cause logistical constraints prevent the application of experimental de signs that are randomized and replicated. I have increased the ability to make inferences about the eff ects of fire on bird abundance by using data collected before the fire and comparing the post-fire changes to habitat model predictions. These data show that fire modifies ve getation characteristics by an immediate reduction in vegetation volume, followed by a gradual decrease in the contribution of conifer vegetation relative to broadleaf vegetation. Because fire is spatially heterogeneous, post-fire vegetation characteristics retain characteristics of the pre-fire vegetation where fire severity is low, but in areas where fire severity is high, changes in vegetation structure and composition are more dramatic. I found evidence that eight out of 29 bird spec ies decreased as a result of the fire. These species were generally those associated with mature coniferous forest. Species
59 associated with mixed-conifer forest did not exhibit strong responses to the changes created by the fire. The strength of these pa tterns varied through tim e; most species that declined did not do so until the second year of the fire, but by the fourth year after the fire the difference between burned and unburned areas had decreased. I found fewer examples of species that increased as a re sult of the fire, but comparing the model predictions with the observed differences indicated that this was not a result of recruitment limitation.
60 Table 3-1. Likelihood ratio test comparing nested models that describe the proportion of sites occupied as a function of year (Y), area (B; burned and unburned), and the interaction of these predictors. For species for which the interaction model fit better than the main effect models, I identified the response as negative (if fewer burned stations were o ccupied in at least three of the four years after the fire) or mixed (if the nu mber of occupied burned stations was greater in some years but less in others). B+Y+BxY vs. B+Y B+Y vs. Y Species Deviance P Deviance P Response to fire Olive-sided Flycatcher -1.363 0.851 -19.53 < 0.001 not significant Western WoodPewee -0.74 0.946 -15.288 < 0.001 not significant Dusky Flycatcher -1.831 0.767 -29.932 < 0.001 not significant Pacific-slope Flycatcher -7.136 0.129 -0.211 0.646 not significant Cassin's Vireo -12.047 0.017 -3 .476 0.062 negative short term Warbling Vireo -0.773 0.942 -2.429 0.119 not significant Steller's Jay -7.089 0.131 -0.094 0.759 not significant Mountain Chickadee -2.113 0.715 -13.488 < 0.001 not significant Chestnut-backed Chickadee -25.44 < 0.001 -17.266 < 0.001 negative Red-breasted Nuthatch -33.442 < 0.001 -18.874 < 0.001 negative Brown Creeper -11.41 0.022 -6.04 0.014 negative short term House Wren -7.264 0.123 -118.531 < 0.001 not significant Golden-crowned Kinglet -7.784 0.1 -14.214 < 0.001 not significant Townsend's Solitaire -8.854 0.065 -0.005 0.941 not significant Hermit Thrush -21.723 < 0.001 -5.032 0.025 negative American Robin -11.913 0.018 -0.142 0.706 mixed Wrentit -7.399 0.116 -0.498 0.48 not significant Nashville Warbler -20.002 < 0.001 < 0.001 0.989 negative Yellow-rumped Warbler -21.404 < 0.001 -1.944 0.163 mixed Black-throated Gray Warbler -15.867 0.003 -8.123 0.004 negative Hermit Warbler -21.031 < 0.001 -33.553 < 0.001 negative MacGillivray's Warbler -1.878 0.758 -2.65 0.104 not significant Wilson's Warbler -7.095 0.131 < 0.001 0.991 not significant Western Tanager -3.783 0.436 -13.814 < 0.001 not significant Spotted Towhee -1.009 0.908 -21.26 < 0.001 not significant Oregon Junco -13.304 0.01 -1. 144 0.285 positive short term
61 Table 3-1. Continued B+Y+BxY vs. B+Y B+Y vs. Y Species Deviance P Deviance Species Deviance Black-headed Grosbeak -7.567 0.109 -1.292 0.256 not significant Lazuli Bunting -5.296 0.258 -13.154 < 0.001 not significant Purple Finch -1.241 0.871 -0.069 0.792 not significant Pine Siskin -36.674 < 0.001 -0.515 0.473 positive
62 Figure 3-1. A map of the study area in the Little Applegat e Valley, Oregon, USA, that shows the area burned by the Quartz fi re and the locations of burned and unburned stations where data on bird a bundance and vegetation were collected from 2001 thru 2005.
63 Figure 3-2. Percent cover of six vegetation taxa at burned and unburned areas in the year before and 4 years after fire in the Little Applegate Va lley of southern Oregon.
64 Figure 3-3. Sampling locations across a physiogn omic gradient of vegetation volume (0 = low volume, 1 = high volume) and a fl oristic gradient of broadleaf-conifer composition (0 = all broadleaf vegetation, 1 = all conifer vege tation) in burned (panel A) and unburned (panel B) areas in the year before and fourth year after the Quartz Fire. Arrows connect the mean values of all stations during the five year study period (starting in 2001 and ending in 2005). Unburned Burned stations
65 Proportion of occupied stations Years after fire Figure 3-4. Proportion of sta tions occupied by 30 passeri ne species in burned and unburned areas in the year before an d 4 years after the Quartz Fire.
66 Figure 3-4. Continued Proportion of occupied stations Years after fire
67 Proportion of occupied stations Years after fire Figure 3-4. Continued
68 Proportion of occupied stations Years after fire Figure 3-4. Continued
69 Figure 3-5. Example of the observed proportion of occupied stations in burned and unburned areas and the pred icted proportion of occupi ed stations for the Hermit Warbler.
70 Predicted difference Observed difference Figure 3-6. The observed differe nce in the proportion of occu pied stations in burned and unburned areas versus the predicted di fference based on habitat occupancy models (see chapter 2) fo r the 5 year study period.
71 Figure 3-7. Observed difference in the propor tion of occupied stations in burned and unburned areas for eight species that decreased as a result of the fire.
72 CHAPTER 4 POST-FIRE HABITAT QUALITY FOR PASSERINE BIRDS: PREDATOR ABUNDANCE, ARTHROPOD ABUND ANCE, AND FORAGING BEHAVIOR Introduction Fire and other large disturbances change bird community composition by shifting vegetation structure and composition such that birds associated with late successional vegetation decrease in abundance, and t hose associated with early successional vegetation increase (Brawn et al. 2001). Additionally, fire may affect birds independently of vegetation if it changes food availabili ty or the risk of predation (Huff et al. 2005). These factors may determine whethe r land managers can mimic the effects of natural wildfire using prescr ibed fire and mechanical treatments. Increasingly, these tools are being used to manage fire-pr one vegetation communities and reduce the economic risk of severe wildfire (Stephe ns 1998, O'Laughlin 2005). However, if these treatments do not produce the same effects creat ed by natural wildfire, then they may fail to provide suitable habitat conditions for wildlif e species that they are designed to benefit (Tiedemann et al. 2000). Nest predation (Martin 1988) and food availability (Nagy and Holmes 2004) are widely recognized as factors that influence patterns of bird abundance and community composition. Important nest predators in western North American forests include squirrels (Tamiasciurus spp.), chipmunks ( Tamias spp.), mice ( Peromyscus spp.), corvids, and raptors (Bradley and Marzlu ff 2003, Cain et al. 2003, Vigallon and Marzluff 2005). The effects of these predators on bi rd communities may be important. In a
73 comparison of bird communities in areas of the Rocky Mountains with and without squirrels, Siepielski (2006) found that ca nopy and understory nesting birds were more abundant in areas without squi rrels. Furthermore, differe nces in predation risk can change when and where birds decide to nest (Forstmeier and Weiss 2004). Food availability is also recognized as an impor tant factor that regu lates bird abundance, especially for species that can raise two broods per bree ding season when food allows (Nagy and Holmes 2004). If food limitation caus es young to grow slowly or fledge late, their recruitment into the population the follo wing year may be reduced (Morton et al. 2004). Furthermore, changes in food availabili ty may change the results of competitive interactions among species (Shochat et al. 2004 ). As a result of these effects, food availability may act independently of, or in concert with, vegetation structure to structure bird community composition (Rabenold 1978, Janes 1994). Although post-fire changes in predator activ ity and food availability may influence how birds respond to fire, these topics have received very little study (Saab et al. 2004, Huff et al. 2005). Here, I use data on pred ator abundance, arthropod abundance, and bird foraging behavior, collected in an area burned by a natu ral wildfire and adjacent unburned areas, to evaluate differences in food availability and predator activity that may affect bird abundance. The objectives were to (1) compare the abundance of avian and mammalian predators that might influence nesting success of bi rds in burned and unburned areas, (2) compare the abundance of arthropods in burned and unburned areas to evaluate the degree to which food availabi lity may be affected by fire, and (3) use observations of bird foraging behavior to evaluate the degree to which fire changes foraging behavior of birds.
74 Methods Study Area and Fire Effects This study was conducted in the Little Applegate Valley in Jackson County, approximately 17.5 km southwest of As hland, Oregon (524000E, 4671000N, UTM zone 10, NAD27; Fig. 3.1), in the Klamath-Siskiyou re gion. Elevation of the watershed ranges from approximately 400 to 2250 m above sea leve l. The forest vegetation of the KlamathSiskiyou region is diverse and includes a variety of both conifer and hardwood species(Whittaker 1960). At lower elevations and on drier sites, vegetation is dominated by broadleaf species, in the form of oa k woodlands with both Oregon white oak ( Quercus garryana ) and California black oak ( Quercus kellogii ) or sclerophyllous shrublands (chaparral), in which broadl eaf shrubs of the genera Ceanothus, Arctostaphyllus, and Baccharis are abundant (Huff et al. 2005). On wetter sites or middl e elevations the vegetation becomes a mix of coni fers, including Douglas-fir ( Pseudotsuga menziesii ), ponderosa pine ( Pinus ponderosa ), incense-cedar ( Calocedrus decurrens ), and white fir ( Abies concolor ); and hardwoods including tanoak ( Lithocarpus densiflorus ), Pacific madrone ( Arbutus menziesii ), Canyon Live Oak ( Quercus chrysolepis ), California black oak, Oregon white oak, and big-leaf maple ( Acer macrophyllum ). The relative composition of these species varies with elev ation, aspect, soils, and disturbance history (Whittaker 1960). Generally, these forest types correspond to the Douglas-fir, Mixed Evergreen Hardwood, or White Fir Types (F ranklin and Dyrness 1973, Huff et al. 2005). Fire-related studies in these vegetation types show a mix of fire severities, frequencies, and sizes, typically characteris tic of low and moderate-sever ity fire regimes (Agee 1991, Taylor and Skinner 1998). Over time, such mixed-severity fires can create forests with
75 multiple age classes, often w ith Douglas-fir or ponderosa pine as an emergent canopy above various hardwoods. Between the 9th and 31st of August 2001, the Quartz fi re burned 2,500 ha across an elevation range from 770-1920 m near the center of this study area. This fire burned in a valley that faces to the north with slopes in the burned area ranging from 0 to 63 degrees. As is typical of fires in this region, fire severity was highly vari able across this area (Alexander et al. In press). Fire severity was mapped by the Burned Area Emergency Rehabilitation team (BAER) of the United St ates Forest Service (USFS). These maps identified severity as low, medium, or high, based on postburn aerial photography and/or observations from the field. In areas categor ized as low severity, fires usually spread rapidly and residence time is short, or these areas may be skipped over by fire. In either case, vegetation is only lightly scorched vege tation and there is minimal tree. The effects of the fire in areas categorized as moderate severity are more extreme; soil structure remains intact but fine fuels near the ground are mostly consumed and 40 to 80% of trees are killed, but retain brown needles that later fall to the forest floor. In areas of high fire severity, fire and heat residence times ar e longest and ground cover may be completely consumed. In these high severity areas, so il structure may be altered if soil organic matter is consumed. Because tree crowns are completely consumed, few leaves or needles remain on trees and mortality is of ten 100%. Of the 2,500 ha Quartz Fire, the BAER map classified it as 23% low severity, 36% mode rate severity, and 41% high severity (Alexander et al. In press). Study Design The Quartz Fire burned in an area wh ere extensive bird surveys had been conducted at in the spring of 2001 (see Chapter 2). These surveys were conducted at
76 stations that were spaced 250 m apart on tran sects of 10-18 stations along secondary and tertiary roads and trails. The area burned in cluded 52 stations where information on bird abundance and vegetation had been collected in the spring before the fire. These sites, and an additional 77 stations that were > 500 m outside the fire perimeter, were resurveyed in 2002, 2003, 2004, and 2005 to investigate the effects of this fire on passerine bird abundance (see Chapter 3). In 2004 a nd 2005 data on predator abundance, arthropod abundance, and avian foraging behavior were collected. Predator Abundance Predator abundance was measured at all stations between 12 May and 2 July of 2004 and 2005 using standardized point count methodologies (Ralph et al. 1993). Fiveminute bird counts were conducted between sunrise and 1000 PDT on each station. During each count the distance at the time of fi rst detection was recorded for all birds and mammals seen and heard. Counts were c onducted only on days when the wind was < 20 kph and it was not raining. All observers we re experienced and had been trained for distance estimation and species identification. For the analysis, I grouped predators into avian and mammalian classes. Avian predators were corvids (3 species) and accip itrids (3 species). Mammalian predators detected during surveys were squirrels and ch ipmunks. I only used detections within 100 m of the point, and assumed the ability of an observer to detect bird s within 100 m of the station was equivalent among vegetation types (Smucker et al. 2005). Because multiple avian predators were often detected at a single station, I calculated their abundance by summing the number of individuals of all avia n predator detected at each station. Because mammalian predators were detected much less frequently and more than one species was never detected at single station, I used the pr esence or absence of mammalian
77 predators detected at a stati on and analyzed the difference in the frequency of occurrence between burned and unburned stations. Arthropod Abundance I measured arthropod abundance at a subset of the 129 stations where information on bird abundance was recorded. This subs et included nearly all (49 of 52) of the stations in the burned area and a random sample of 23 (of 77) stations in unburned areas. Each station was sampled twice in 2004 a nd 2005 between 6 May and 12 July. At each station arthropods were collect ed using a sweep net (38 cm diameter). Because point count stations were located along the road, I moved 10-15 m perpendicular from the road before beginning to sample. Each sample c onsisted of 10 sweeps of the net, with each sweep directed at unswept ve getation between 0.5 and 2 m in height. Arthropod samples were preserved in 95% ethanol. I identifie d all arthropods to order and measured each individual to the nearest millimeter. These lengths were then converted to an estimate of dry mass using the equation: M = 0.0305 L2.62 where M is mass (mg) and L is length (mm) (R ogers et al. 1976). Avian Foraging Behavior I observed passerine birds with 10x40 binocul ars and recorded their behavior with a tape recorder (Robinson and Holmes 1982). After identifying a focal individual, I followed that bird for up to two minutes, recording all movements associated with searching for and capturing prey. I recorded the duration (s) of the observation and the total number of three activity types: hops, fli ghts, and attacks on pre y. For all activities, substrate (live foliage, dead foliage, bark /stem, ground, log, air, flower, other) was recorded as well as the species of vegetation in which the bird was foraging. Hops were
78 defined as movements when the birds do not extend their wings, as opposed to flights, where birds change perches by flying. I attemp ted to record whether or not attacks were successful, and information on types of prey captured, but in most cases prey were too small to be observed. Statistical Analyses Predator abundance The count of avian predators detected at each station was analyzed using a generalized linear model with a log link and Poisson erro r distribution. This model included parameters for area (burned and unburned), year (2004 and 2005) and their interaction. Significance was ev aluated using Wald tests of parameter values with alpha = 0.05. Because the interacti on term was non-significant ( P = 0.425), I subsequently refit the model with only the main effects of y ear and area. Because mammalian predators were detected at very few st ations, I analyzed each year separately and used Fisherâ€™s exact test to evaluate if th e proportion of stations with mammalian predators differed between burned and unburned areas. Arthropod abundance For each station, I pooled the two samples collected in each year, such that the biomass estimate was based on the number of insects captured in a total of 20 sweeps. For each station, I calculated total biomass by summing the biomass of all individuals in that sample. This was performed for the following groups: all ar thropods, Arachnida (spiders), Diptera (flies), Hy menoptera (ants, bees, and wasps), Hemiptera (true bugs, aphids, and leafhoppers), Cole optera (beetles), and larvae (primarily Lepidoptera). Spatial and temporal variation in the biom ass of these groups was evaluated using a Kruskal-Wallis rank sum test for diffe rences in biomass among 2004 burned and
79 unburned and 2005 burned and unburned. When the global test of differences was significant, I used a nonparametric Steel procedure for post-hoc multiple comparisons (Munzel and Hothorn 2001). Avian foraging behavior Only observations > 10 s in duration were us ed in the analysis. I also excluded all observations of foraging activity on the gr ound and on completely dead vegetation, but this was a relatively small proportion (< 10 %) of the observations. I pooled observations for all species and considered each observation an independent sample. For each observation, foraging rate was calculated by di viding the total number of prey attacks by the duration of the observation. To evaluate if the proportion of foraging bouts observed in conifer and broadleaf vegetation differe d between burned and unburned areas, I used Fisherâ€™s exact test. I used Wilcoxonâ€™s Signe d-Rank test to compare the rates at which birds attacked prey between burned and unburned areas and betw een broadleaf and conifer vegetation. Results Predator Abundance Six species of avian predators we re detected. Stellerâ€™s Jays ( Cyanocitta stelleri ) accounted for the majority (55%) of all predator detections (Fig. 4-1) . The mean number of avian predators detected per point (pooling across burned and unburned areas) was 0.395 in 2004 and 0.357 in 2005. The mean number of detections (pooling across years) was 0.327 per station in burned areas and 0.410 per station in unburned areas. There was no evidence that the number of avian predators per station varied between years (Wald test, Z = -0.507, P = 0.612) or between burned and unbur ned areas (Wald test, Z = 1.054, P = 0.292). Mammalian predators were detected much less frequently (Fig. 4-1). In
80 2004 mammalian predators were detected at onl y 5 of 72 unburned stations and none of the 52 burned stations. In 2005 they were de tected at 5 unburned stations and 3 burned stations. Thus, although ther e was a trend toward fewer ma mmalian predators in burned areas, in neither year was the difference statis tically significant (2004: Fisherâ€™s exact test, P = 0.081; 2005: Fisherâ€™s exact test, P = 1.000). Arthropod Abundance Pooling across both years, Hemiptera made up the greatest pr oportion (55%) of the total arthropod biomass (Fig. 4-2). La rvae, Coleoptera, Diptera, Neuroptera, Hymenoptera, and Arachnida were also importa nt components of arthropod biomass (Fig. 4-2). Total arthropod biomass varied across sites and years (Kruskal-Wallis X2 = 13.06, P = 0.005). In 2004, the total arthropod biom ass did not differ between burned and unburned stations, but in 2005 there was significa ntly greater biomass at burned stations than there had been in 2004, and greater bi omass than at the unburned stations in 2005 (Fig. 4-3). However, patterns in variation of overall ar thropod biomass were the result of very different responses by different arthropod orde rs. Both hemipteran biomass (KruskalWallis X2 = 21.06, P < 0.001) and Coleopteran biomass (Kruskal-Wallis X2 = 28.27, P < 0.001) varied in a pattern that suggested spa tial variation in biomass was influenced by the fire. Hemipteran biomass was consistently greater at burned st ations than at unburned stations in both 2004 and 2005 (Fig. 4-3) . Similarly, Coleopteran biomass was consistently greater at burned stations than at unburned sta tions in both years, but this difference became more pronounced in 2005. Although dipteran biomass varied among burned and unburned and by year (Kruskal-Wallis X2 = 20.26, P < 0.001), this variation was relatively modest (Fig. 4-3). I fo und no evidence that biomass of Arachnida
81 (Kruskal-Wallis X2 = 0.75, P = 0.860), Hymenoptera (Kruskal-Wallis X2 = 1.81, P = 0.612), Neuroptera (Kruskal-Wallis X2 = 3.97, P < 0.264), nor larvae (Kruskal-Wallis X2 = 7.56, P < 0.056) varied either spatially or temp orally (Fig. 4-3). Thus, the greater biomass of all arthropods at burned stations in 2005 was probably largely due to an increase in the biomass of beetles between 2004 and 2005. Avian Foraging Behavior A total of 55 foraging observations of 16 sp ecies were recorded (Table 4-2). The proportion of observations recorded from c onifer and broadleaf vegetation differed between the burned and unburned areas (Fisherâ€™s exact test, P = 0.015). In burned areas, the majority (15 of 26; 58%) were recorded in broadleaf vegetation, whereas in unburned areas the majority (22 of 29; 76%) occurred in conifer species. Foraging rates were highly variable, but there was no evidence that the rate at which birds attacked prey differed between burned and unburned areas (Fig. 4-4; Wilcoxon, P = 0.899) or when they foraging on broadleaf versus coni fer vegetation species (Fig. 4-4; Wilcoxon, P = 0.795). Discussion Predator Abundance I found no evidence of systematic differe nces in the abundance of avian or mammalian predators between burned and unbur ned areas. The measure of avian predators is believed to be relatively reliable , especially with resp ect to its ability to measure the activity of corvids, which are voc al and visually conspicuous. In contrast, my estimate of mammalian predators is likel y limited, in part because relatively few sciurids were detected, and also because entire groups of mammalian predators (e.g., murids, canids, and mustelids) were not detected during point counts. Peromyscus mice
82 in particular are recognized as an important pr edator of forest bird s and these also could not be detected (Bradley and Marzluff 2003). Several authors have raised the possibility that nest pr edation may differ between burned and unburned areas, based on indir ect evidence of predator activity and abundance (Jones et al. 2004) a nd greater reproductive succe ss of birds in burned areas (Saab et al. 2004, Brawn 2006). De spite these observations, ne st predation studies have not consistently documented differences between burned and unburned areas. For example, based on observation of real and artif icial nests, Stuart-Smith and Hayes (2003), found no consistent evidence that nest preda tion increased with residual tree density of burned and selectively logged stands in Brit ish Columbia. Furthermore, they found no difference between predation ra tes on artificial nests in logged and burned stands. Similarly, Wood Thrushes ( Hylocichla mustelina ) had similar nest success in unburned areas and areas burned by prescribed fire in Ohio, a result which the authors linked to the heterogeneity in fire severity that left pocke ts of vegetation with suitable cover for nest concealment (Artman and Downhower 2003). Ho wever, it is possible that apparently static patterns of reproductiv e success are the resu lt of complex predation dynamics. In a study of artificial nests in unbur ned areas and areas treated with prescribed fire in Georgia, Jones et al. (2002) found no differe nce in overall nest success between burned and unburned areas, but demonstrated avian pr edators were more important in burned areas and mammalian predators were more im portant in unburned areas. Even in the absence of changes in predator abundance, th e risk of predation may change if post-fire vegetation characteristics decrease the ability of adults to detect pr edators or to conceal themselves or their nests from predators (M etcalfe 1984). Thus, although I did not find a
83 substantial difference between predator co mmunities of burned and unburned areas, the literature suggests that in some areas this effect may be important. Arthropod Abundance Total arthropod biomass was comparable at burned and unburned stations in 2004 (3 years post-fire), but greater at burned st ations in 2005 (4 years post-fire). Because there was no significant difference between arthropod biomass at unburned stations in 2004 and 2005, this difference resulted from an increase in arthropod biomass at the burned stations (Fig. 4-3). This pattern appear ed to be driven primarily by an increase in beetle biomass at burned stations between 2004 and 2005, and perhaps to a lesser extent by a similar, though less dramatic, increase in dipterans at burn ed stations. In contrast, biomass of Arachnida, Hymenoptera, and larvae varied relatively little either spatially or temporally. Hemipteran biom ass was greater at burned st ations, and this pattern was consistent during both years of the study (Fig. 4-3). The ability to generalize the degree to which these patterns of arthropod biomass are characteristic of post-fire environments is limited by two factors. First, there is a surprising lack of information on how arthr opod communities in mixed-conifer forests of western North America respond to fire (Apigian et al. 2006). Most st udies that have been published have focused on leaf-litter arthr opods using pit-fall traps (Niwa and Peck 2002, Apigian et al. 2006). The taxa sampled by this technique may respond very differently to fire than the foliage dwelling arthropods sampled in this study. Second, many of the studies that have addressed th is topic have focused on the short-term (1-2 year post-fire) response to fire (Meehan and George 2003, Apig ian et al. 2006). Such studies probably document the degree to which insect commun ities are altered by direct mortality and subsequent recruitment limitation (Knight and Holt 2005) rather than the longer-term
84 changes in vegetation compositi on that are likely important in this study. The post-fire succession of arthropod communities in mixed-c onifer forests needs more comprehensive studies that evaluate both the short-term effects of recruitment limitation and the longerterm response to post-fire vegetation. However, some information is available on foliage-dwelling arthropod responses to fire in forest systems from Australia (A ndersen and Muller 2000, Orgeas and Andersen 2001) and Switzerland (Moretti et al. 2004). Th ese studies have shown that spiders and hymenopterans are generally resilient to fi re, occurring in equivalent (Andersen and Muller 2000) or greater (Moretti et al. 2004) a bundance at burned areas. Furthermore, they have demonstrated an increase in or ders with many phytophagous species, including Coleoptera and Hemiptera (Andersen a nd Muller 2000, Orgeas and Andersen 2001, Moretti et al. 2004). These results are generall y consistent with the patterns of arthropod abundance that I measured three and four years post-fire. One of the few studies that has describe d post-wildfire insect abundance in mixedconifer forest similar to the area burned by the Quartz fire was conducted after the Megram fire burned 20,700 ha in Northern Ca lifornia (Meehan and George 2003). Using suspended malaise traps, th ese authors compared aerial arthropod abundance at Olivesided Flycatcher territories that were unburned to those dur ing the first and second year post-fire. They found that overall arthropod biomass and the biomass of Coleoptera, Arachnida, and Hemiptera (from which they excluded the â€œHomopteraâ€) was greater at unburned territories, but that there was no difference in the bi omass of Homoptera (aphids and leafhoppers), Hymenoptera, and Diptera between burned and unburned
85 territories. Meehan and George (2003) al so documented lower ne sting success of Olivesided Flycatchers at burned territories, whic h they attributed to reduce food availability. The difference between Meehan and George â€™s (2003) results (f ewer arthropods in burned areas) and my results (equal or more arthropods in burned areas) may reflect the timing of these two studies. Meehan and George (2003) compared arthropod abundance in the first two years post-fire, and I comp ared the third and fourth year post-fire. Linking these studies together would suggest that arthropod biomass in burned areas is reduced in the first two years after fire, recovered by the th ird year, and greater in the fourth year. This is consistent with tempor al pattern the Mesa fire (Arizona) where many insect groups took up to two years to recove r (Pippin and Nichols 1996). It is also consistent with the patterns of bird abunda nce observed in Quartz fire; several bird species were less abundant at burned areas in the year s immediately following than would be predicted by post-fire vegetation char acteristics alone, but this difference had disappeared by the fourth year post-fire (see Chapter 3). Avian Foraging Behavior Birds in burned areas foraged more freque ntly in broadleaf vegetation than did birds in unburned areas, reflecti ng the overall pattern of greate r broadleaf availability in the post-fire vegetation (see Chapter 3). The pattern of foraging location likely resulted both because behavioral plasticity allows birds that use both burned and unburned areas (e.g., Black-headed Grosbeaks and Nashville Warb lers, Table 4-2) to use either conifers or broadleaf species, and because species that are associated with conifers (e.g., Hermit Warbler, Red-breasted Nuthatch) were less abu ndant in burned areas (see Chapter 3). Generally, many birds have obvious pref erences for foraging in particular vegetation species (Holmes and Robinson 1981 , Gabbe et al. 2002). Presumably, these
86 preferences represent selecti on of micro-environments that optimize their ability to capture prey (Whelan 1989, 2001). When vegetation composition or prey availability is spatially variable, differences in foragi ng rates may result (Kilgo 2005, Lyons 2005). However, despite the fact that arthropod bi omass was more abundant in the burned areas, I found no evidence that birds a ttacked prey more frequently in burned areas. Nor was their evidence that attack rates were higher in broadleaf species than in conifer species. My analysis was limited because I lacked suffi cient replication to analyze the foraging behavior of individual species. Because speci es may vary dramatically in their foraging behavior (Holmes and Robinson 1981), lumpi ng all species together may have limited my power to detect differences between burned and unburned habitats. Conclusions Temporal and spatial variation in nest predation and food availability may be important factors explaining post-fire ch anges in bird community composition. I compared predator abundance, arthropod av ailability, and avian foraging behavior between burned and unburned areas in a mixed-c onifer forest in Southern Oregon. This sampling occurred in the third and fourth year after the fire. I found no evidence that the abundance of nest predators differed between burned and unburned areas in either year. Although these data suggest that predators are not an important difference between burned and unburned areas, this hypothesis needs to be tested using more sensitive methods to compare predator communities, pa rticularly mammals, and nest predation rates in burned and unburned areas. There was substantial spatia l and temporal variation in arthropod abundance. The biomass of hemipterans and coleopterans was greater in burned areas in both the third and fourth year after the fire, and overall arthropod biomass wa s greater in the fourth year
87 after the fire. These differences and data from other studies of post-fire arthropod biomass at different times post-fire s uggest that although arthropod abundance may decrease for one to two year s post-fire, it recovers in th e third year, and surpasses unburned sites in the fourth year. This in crease is probably driven in large part by phytophagous insects that feed on hardwoods th at re-sprout after fi re. However, this increase in arthropod biomass wa s not associated with the ra te at which birds attacked prey in burned and unburned areas. I found no evidence of differences in predator activity or the rate of which birds captured prey between burned and unburned area s. I did find that arthropod biomass was greater at burned areas, and that birds in burned areas foraged more frequently in broadleaf vegetation. These post-fire cha nges are likely a result of the increase in broadleaf vegetation that occurred after the fire (see Chapter 3). These results suggest that in this mixed-conifer forest, the eff ects of fire on bird communities are probably largely driven by post-fire changes in vegeta tion structure and composition characteristics of forest vegetation that also influence arthr opod abundance. This is consistent with the results of habitat modeling that showed that post-fire changes in bird abundance are consistent with those predicted by vegetati on characteristics (see Chapter 3). These results would suggest that for these passerine species, managers may be successful at mimicking the effects of fire with mechanical treatments or prescribed fire if these tools can replicate fire effects on vegetation structure and composition. However, such management should also consider other spec ies, such as woodpeckers, amphibians, and mollusks. For these species, mimicking the e ffects of fire may be more complicated than for species that respond primarily to vegeta tion structure and composition. Studies that
88 directly compare these fact ors between burned areas a nd area where vegetation is managed with mechanical treatments will he lp to confirm the results of this study.
89 Table 4-1. The number of foraging observati ons recorded for 16 species of passerine birds in burned and unburned areas of the Little Applegate Valley, Oregon. Species Burned Unburned Black-capped Chickadee, Poecile atricapilla 1 0 Black-headed Grosbeak, Pheucticus melanocephalus 7 7 Brown Creeper, Certhia Americana 0 1 Black-throated Gray Warbler, Dendroica nigrescens 0 1 Bushtit, Psaltriparus minimus 2 1 Chestnut-backed Chickadee, Poecile rufescens 0 1 Dark-eyed Junco, Junco hymealis 2 3 Hermit Warbler, Dendroica occidentalis 1 4 Lazuli Bunting, Passerina amoena 2 1 MacGillivray's Warbler, Oporornis tolmiei 1 0 Mountain Chickadee, Poecile gambeli 0 1 Nashville Warbler, Vermivora ruficapilla 3 3 Red-breasted Nuthatch, Sitta Canadensis 2 1 Warbling Vireo, Vireo gilvus 2 1 Western Tanager, Piranga ludoviciana 1 3 Yellow-rumped Warbler, Dendroica coronata 2 1 Total 26 29
90 Figure 4-1. The proportion of total predator detections (N = 112) composed of eight avian and mammalian taxa in burned and unburned areas of the Little Applegate Valley, Oregon, in 2004 and 2005.
91 Figure 4-2. The proportion of total biomass composed of eight arthropod groups from burned and unburned areas of the Littl e Applegate Valley in 2004 and 2005.
92 Figure 4-3. Spatial (burned and unburned) a nd temporal (2004 and 2005) variation in arthropod biomass samples collected by sw eep netting in the Little Applegate valley of southern Oregon. Bars are me dians, boxes are defined by the upper and lower quartiles, and whiskers are maximum and minimum values. Letters indicate groups that were significantly different (P < 0.05) from each other based on multiple comparisons with a non-parametric Steel-test.
93 Figure 4-4. Foraging rates compared between observation in burned and unburned areas and in conifer and broadleaf vegetation. Bars are medians, boxes are defined by the upper and lower quartiles, and whiskers are maximum and minimum values.
94 CHAPTER 5 CONCLUSION Fire has been, and will remain, one of the major forces that changes forest structure in North America and globally (Pyne 1982, Ag ee 1993). When fires burn naturally, they create a shifting mosaic of burned and unburned areas that maintains a dynamic equilibrium (Thonicke et al. 2001). When fi res are suppressed, thei r absence leads to changes in vegetation characteristics that influence ecological communities and may change patterns of fire severity or frequenc y (Clark et al. 1999). After years of policies that focused simply on fire suppression, land managers in North America are now trying to shift toward management plans that use fire and mechanical treatments to reduce the economic risks of wildfire and restore ecologi cal characteristics that have been changed by fire suppression (Agee 1993, Dombeck et al . 2004, Spies et al. 2006). However, there is concern that these alternatives to natural fire may fail to provide the conditions for wildlife that they are intended to create (Tie demann et al. 2000). In a recent review, Huff et al. (2005) documented that our knowledge of the effects of fire on bird communities in the Pacific Northwest is very limited, and out lined ten questions that , if answered, would help to guide fire management in this region. This dissertation begins to answer these questions about how and why fire changes bi rd communities in mixed-conifer forests of the Pacific Northwest. In Chapter 2, I used simple measures of vegetation volume and the proportion of vegetation comprised of conifer sp ecies to explain spatial patter ns of bird distribution in a watershed in southern Oregon. This analysis demonstrated that the distribution of many
95 birds in this watershed can be predicte d based on vegetation volume, vegetation composition, and, in some cases, their interact ion. These interactions have important applications to the management and conservatio n of forest birds because they suggest that understanding the effects of changing vegeta tion volume requires knowledge of floristic composition. The mechanisms that generate these pattern s require more research. Possibly these interactive effects may repres ent inherent habitat requireme nts of these species, i.e., characteristics of their fundamental niche. Alternatively, if these distributional patterns represent inferior competitive abilities in some environments, then these patterns may represent realized niches (Hutchinson 1965, Holt and Gaines 1992). This difference is important, because fundamental niches s hould apply across a sp eciesâ€™ range, whereas realized niches may be spatially variab le if they are influenced by community composition of a particular location. None -the-less, the predictive power of these vegetation characteristics s uggests that they represent a good framework in which to investigate the effects of fire on bird community composition. In Chapter 3, I used data on bird abunda nce and vegetation that were collected one year before and four years after a wildfire to describe post-fire changes in bird abundance. Previously, the ability to make cau sal inferences about th e effects of wildfire on bird abundance has been limited because study designs have not been randomized and replicated. My study also faced these challenges, but I used pre-fire data and predictions from the habitat models presented in Chapte r 2 to increase my c onfidence that observed changes in bird abundance we re a result of the fire.
96 My results show that in the first year post-fire vegetation volume is dramatically reduced, and that in the following years ther e is a gradual increase in the proportion of the vegetation that is comprise d of broadleaf taxa. During th is 4 year post-fire period, I found evidence that eight out of 29 passerine bird species decreased as a result of the fire, but only one species that increased. The sp ecies that decreased were generally those associated with mature coniferous forest. The strength of these changes varied through time; most species that declined did not do so until the second year of the fire. By the fourth year after the fire, the difference in the proportion of occupied stations in burned and unburned areas was we ll-predicted by the habi tat models presented in Chapter 2. The inability of these models to accurately predict the differences in the first three years after the fire suggests that other factors in addition to the compositional and structural characteristics that birds use as cues for habitat selection may be important immediately after the fire. In Chapter 4, I use data on predator abundance, arthropod abundance, and bird foraging behavior to evaluate mechanisms beyond changes in vegetation structure and composition that may affect bird abundance. I found no evidence that the abundance of nest predators differed between burned and unbur ned areas in either year. There was, however, substantial sp atial and temporal variation in arthropod abundance. Hemipteran and coleopteran biomass was greater in burne d areas in both the third and fourth year after the fire, and overall arthropod biomass was greater in the fourth year after the fire. Thus, although arthropod abundance may decrea se for one to two years post-fire, it appears that it recovers in th e third year, and surp asses unburned sites in the fourth year post-fire. This increase is pr obably driven in large part by phytophagous insects that feed
97 on hardwoods that re-sprout after fire. Foraging observations of passerine birds demonstrated that in burned areas birds forage d more frequently in broadleaf vegetation. Based on the data I have presented in thes e three chapters, I propose that much of the post-fire changes in bird communities can be understood as a result of the increase in broadleaf vegetation that occurred after the fire. These results suggest that for these passerine species, managers may be successful at mimicking the effects of fire with mechanical treatments or prescribed fire if these tools can replicate fire effects on vegetation structure and composition. Often, management has attempted to remove the broadleaf component of vegetation to enhan ce forest productivity (Wagner et al. 2004). I propose that the maintenance of a broadleaf co mponent in post-fire vegetation should be an important consideration of post-fire recove ry plans. Furthermore, such management should also consider other species, such as woodpeckers, amphibians, and mollusks that may be sensitive to habitat features other than vegetation structure and composition.
98 APPENDIX AVIAN FORAGING BEHAVIOR DATA CO LLECTED IN AND ADJACENT TO THE QUARTZ FIRE IN SOUTHERN OREGON In Chapter 4, data on avian foraging beha vior collected in and adjacent to the Quartz Fire in southern Or egon was reported. A summary of these foraging data is presented in Table A-1. This table contains the followi ng fields of information: Species: The species of bird, abbreviated with the following four-let ter codes: BCCH = Black-capped Chickadee, BHGR = Black-h eaded Grosbeak, BTYG = Black-throated Gray Warbler, BUSH = Bushtit, CAVI = Cassinâ€™s Vireo, CBCH = Chestnut-backed Chickadee, DEJU = Dark-eyed Junco, HA WO = Hairy Woodpecker, HEWA = Hermit Warbler, LAZB = Lazuli Bunting, MGWA = MacGillivrayâ€™s Warbler, MOCH = Mountain Chickadee, NAWA = Nashville Warbler, RBNU = Red-breasted Nuthatch, RBSA = Red-breasted Sapsucker, WAVI = Wa rbling Vireo, WEBL = Western Bluebird, WETA = Western Tanager, YRWA = Yellow-ru mped Warbler. Scientific names are given in Table 2-2. Day: Day on which observation was recorded. Month: Month in which observation was recorded. Year: Year in which observation was recorded. Location: The location of the observation reco rded as a RTSTA code (e.g., WRAN02) where the first four letters or numbers id entify the point count ROUTE (e.g., WRAN) and the last two numbers identify the STATION (e.g., 02). A list of UTM coordinates for
99 RTSTA codes is available from the Klam ath Bird Observatory, Ashland, Oregon (www.klamathbird.org ). Burn: A code for whether the location was insi de (B) or outside (U) the Quartz Fire perimeter. Dur: The duration (in seconds) of the observation period. Veg: The vegetation species in which the bi rd was located during the observation, abbreviated with the followi ng four letter codes: PSME = Pseudotsuga menziesii , ARME = Arbutus menziesii , ABSP = Abies spp., CEIN = Ceanothus integerrimus , CADE = Calocedrus decurrens , TODI = Toxicodendron diversiloba , SNAG = dead tree, UNSH = unknown broadleaf shrub, QUKE = Quercus kelloggii , QUGA = Quercus garryana , SALI = Salix spp., PIPO = Pinus ponderosa . Att: The number of prey attacks during the observation. Hop: The number of hops during the observation. Fli: The number of flight s during the observation.
100 Table A-1. Summary of data collected during avian fo raging observations in and adjacent to the Quartz Fire in southern Oregon during 2004 and 2005. Species Day Month Year RTSTA Burn. Dur Veg Att Hop Fli BCCH 17 6 2004 57002 B 35 PSME 1 0 0 BHGB 14 6 2004 KEME08 B 28 ARME 4 6 0 BHGR 5 5 2004 55003 B 106 QUKE 9 20 0 BHGR 17 6 2004 57002 B 52 ARME 0 9 0 BHGR 13 6 2004 KEME05 U 76 PSME 2 7 0 BHGR 14 6 2004 KEME06 U 43 PSME 0 11 0 BHGR 14 6 2004 KEME08 B 65 PSME 3 18 0 BHGR 10 6 2004 KEME15 B 39 CEIN 3 7 0 BHGR 10 6 2004 KEME16 B 16 ARME 3 4 0 BHGR 10 6 2004 KEME16 B 65 ARME 2 6 0 BHGR 28 6 2005 LIKB13 U 85 ARME 2 14 0 BHGR 3 6 2005 LIKB15 U 32 PSME 2 5 0 BHGR 3 6 2005 LIKB15 U 20 PSME 1 2 0 BHGR 9 6 2004 SKAT14 U 21 PSME 2 2 0 BRCR 28 6 2004 7MIL10 U 11 PSME 1 3 0 BTYG 24 5 2004 LIKB03 U 23 ARVI 1 9 0 BUSH 17 5 2005 DUGA19 U 10 PSME 2 1 0 BUSH 10 6 2004 KEME16 B 13 ARME 0 0 0 BUSH 10 6 2004 KEME16 B 59 PSME 0 8 0 CAVI 18 6 2004 55015 B 174 PSME 1 1 0 CAVI 9 5 2005 KEME08 B 72 SNAG 4 14 0 CBCH 28 6 2005 LIKB16 U 25 PIPO 2 1 0 DEJU 28 6 2004 7MLA12 U 17 PSME 3 6 0 DEJU 17 5 2005 DUGA17 B 43 CADE 0 3 0 DEJU 7 6 2004 DUGA18 B 26 PSME 2 6 0 DEJU 17 5 2005 DUGA20 U 39 PSME 3 13 0 DEJU 26 6 2004 WRAN17 U 37 PSME 4 9 0 HAWO 24 5 2004 DUGA12 B 20 SNAG 2 0 0 HAWO 10 6 2004 KEME12 B 11 SNAG 2 0 0 HAWO 10 6 2004 KEME15 B 91 SNAG 2 0 0 HAWO 10 6 2004 KEME16 B 3 PSME 2 0 0 HEWA 17 5 2005 DUGA17 B 27 PSME 0 9 0 HEWA 17 5 2005 DUGA19 U 78 PSME 4 8 0 HEWA 14 6 2004 KEME06 U 12 PIPO 1 7 0 HEWA 25 6 2004 WRAN05 U 10 ABSP 0 5 0 HEWA 26 6 2004 WRAN08 U 21 PSME 2 9 0 LAZB 5 5 2004 55003 B 26 ARME 3 0 0 LAZB 17 6 2004 57003 B 43 SNAG 0 7 0 LAZB 10 6 2004 KEME13 B 18 TODI 2 0 0 LAZB 25 6 2004 WRAN04 U 12 ABSP 0 1 0 MGWA 17 6 2004 57003 B 25 PIPO 1 6 0 MOCH 27 6 2004 60014 U 18 ABSP 0 8 0 NAWA 5 5 2004 55003 B 110 ARME 13 16 0 NAWA 27 6 2004 60014 U 76 SALI 0 17 0 NAWA 27 6 2004 DUGA13 B 27 QUKE 2 8 0 NAWA 17 6 2004 GLAD B 23 PSME 1 3 0
101 Table A-1. Continued. Species Day Month Year RTSTA Burn. Dur Veg Att Hop Fli NAWA 14 6 2004 KEME06 U 23 CEIN 11 7 0 NAWA 9 6 2004 SKAT17 U 18 ARME 1 1 0 RBNU 17 6 2004 57005 B 20 PSME 3 2 0 RBNU 27 6 2004 DUGA12 B 46 ARME 4 4 0 RBNU 14 6 2004 KEME06 U 25 PSME 1 7 0 RBSA 13 6 2004 KEME04 U 94 PIPO 4 0 0 WAVI 9 5 2005 55001 B 31 QUGA 0 9 0 WAVI 17 6 2004 57003 B 29 UNSH 0 1 0 WAVI 27 6 2004 60014 U 205 ARME 2 7 0 WEBL 10 6 2004 KEME15 B 3 SNAG 0 0 0 WETA 16 6 2004 57001 B 68 QUER 1 0 0 WETA 6 5 2004 DUGA18 B 20 CADE 0 5 0 WETA 7 6 2004 DUGA19 U 103 PSME 0 11 0 WETA 17 5 2005 DUGA20 U 30 PSME 2 1 0 WETA 24 5 2004 LIKB03 U 174 PSME 2 0 0 YRWA 17 6 2004 57001 B 21 PSME 1 0 0 YRWA 6 5 2004 DUGA12 B 10 SNAG 1 1 0 YRWA 25 6 2004 WRAN01 U 41 ABSP 2 0 0 YRWA 26 6 2004 WRAN09 B 58 PIPO 3 11 0
102 LIST OF REFERENCES Adams, E. M., and M. L. Morrison. 1993. Effects of forest stand structure and composition on Red-Breasted Nuthatches and Brown Creepers. Journal of Wildlife Management 57 :616-629. Agee, J. K. 1991. Fire history along an eleva tional gradient in th e Siskiyou Mountains, Oregon. Northwest Science 65 :188-199. Agee, J. K. 1993. Fire ecology of Pacific Nort hwest forests. Island Press, Washington, D.C. Alexander, J. D., N. E. Seavy, C. J. Ral ph, and B. Hogoboom. In press. Vegetation and topographical correlates of fire severity from two fires in the Klamath-Siskiyou region of Oregon and California. Inte rnational Journal of Wildland Fire. Altman, B. 2000. Conservation strategy for land birds in lowland and valleys of western Oregon and Washington. Oregon-Washingt on Partners in Flight, Boring, OR. Andersen, A. N., and W. J. Muller. 2000. Arthropod responses to experimental fire regimes in an Australian tropical sava nnah: ordinal-level analysis. Austral Ecology 25 :199-209. Apigian, K. O., D. L. Dahlsten, and S. L. Stephens. 2006. Fire and fire surrogate treatment effects on leaf litter arthropods in a western Sierra Nevada mixedconifer forest. Forest Ecology and Management 221 :110-122. Artman, V. L., and J. F. Downhower. 2003. Wood thrush ( Hylocichla mustelina ) nesting ecology in relation to prescribed burni ng of mixed-oak forest in Ohio. Auk 120 :874-882. Bersier, L. F., and D. R. Meyer. 1994. Bird assemblages in mosaic forests: the relative importance of vegetation structure and floristic composition along the successional gradient. Acta Oecologica 15 :561-576. Bersier, L. F., and D. R. Meyer. 1995. Relationships between bird assemblages, vegetation structure, and floristic com position of mosaic patches in riparian forests. Revue d'Ecologie: la Terre et la Vie 50 :15-33. Bradley, J. E., and J. M. Marzluff. 2003. R odents as nest predators: Influences on predatory behavior and conseque nces to nesting birds. Auk 120 :1180-1187.
103 Brawn, J. D. 2006. Effects of restoring oak savannas on bird communities and populations. Conservation Biology 20 :460-469. Brawn, J. D., S. K. Robinson, and F. R. Thompson. 2001. The role of disturbance in the ecology and conservation of birds. Annua l Review of Ecology and Systematics 32 :251-276. Brotons, L., P. Pons, and S. Herrando. 2005. Colonization of dynamic Mediterranean landscapes: where do birds come from after fire? Journal of Biogeography 32 :789-798. Brown, R. T., J. K. Agee, and J. F. Frankli n. 2004. Forest restoration and fire: Principles in the context of place. Conservation Biology 18 :903-912. Bruno, J. F. 2000. Facilitation of cobble beach plant communities through habitat modification by Spartina alterniflora . Ecology 81 :1179-1192. Cain, J. W., M. L. Morrison, and H. L. Bo mbay. 2003. Predator activity and nest success of willow flycatchers and yellow warblers. Journal of Wildlife Management 67 :600-610. Camprodon, J., and L. Brotons. 2006. Effect s of undergrowth clearing on the bird communities of the Northwestern Med iterranean Coppice Holm oak forests. Forest Ecology and Management 221 :72-82. Cantor, S. B., C. C. Sun, G. Tortolero-L una, R. Richards-Kortum, and M. Follen. 1999. A comparison of C/B ratios from studies us ing receiver operating characteristic curve analysis. Journal of Clinical Epidemiology 52 :885-892. Chappell, C. B., and J. Kagan. 2000. Cea nothus-Manzanita shrublands. Pages 43-44 in D. H. Johnson and T. A. O'neil, editors. Wild life-habitat relation ships in Oregon and Washington. University of Or egon Press, Corvallis, OR. Cheplick, G. P. 2005. Patterns in the di stribution of American beachgrass ( Ammophila breviligulata ) and the density and reproduction of annual plants on a coastal beach. Plant Ecology 180 :57-67. Clark, D. B., M. W. Palmer, and D. A. Clark. 1999. Edaphic factor s and the landscapescale distributions of tropical rain forest trees. Ecology 80 :2662-2675. Collins, S. L. 2000. Disturbance frequency and community stability in native tallgrass prairie. American Naturalist 155 :311-325. Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs. Science 199 :13021310. Crome, F. H. J., M. R. Thomas, and L. A. Moore. 1996. A novel Bayesian approach to assessing impacts of rain forest logging. Ecological Applications 6 :1104-1123.
104 Cushman, S. A., and K. McGari gal. 2004. Hierarchical analysis of forest bird speciesenvironment relationships in the Oregon Coast Range. Ecological Applications 14 :1090-1105. Davis, J. M., and J. A. Stamps. 2004. Th e effect of natal experience on habitat preferences. Trends in Ecology and Evolution 19 :411-416. Dobbs, R. C., P. R. Martin, and T. E. Martin. 1998. Green-tailed Towhee ( Pipilo chlorurus ). in A. Poole and F. Gill, editors. The Birds of North America. The Birds of North America, Inc., Philadelphia, PA. Dombeck, M. P., J. E. Williams, and C. A. Wood. 2004. Wildfire pol icy and public lands: Integrating scientific und erstanding with social c oncerns across landscapes. Conservation Biology 18 :883-889. Easton, W. E., and K. Martin. 1998. The eff ect of vegetation ma nagement on breeding bird communities in British Columbia. Ecological Applications 8 :1092-1103. Fleishman, E., N. McDonal, R. Mac Nally, D. D. Murphy, J. Walters, and T. Floyd. 2003. Effects of floristics, physiognomy and nonnative vegetation on riparian bird communities in a Mojave Desert watershed. Journal of Animal Ecology 72 :484490. Forstmeier, W., and I. Weiss. 2004. Adaptive plas ticity in nest-site selection in response to changing predation risk. Oikos 104 :487-499. Foster, D. R., D. H. Knight, and J. F. Fr anklin. 1998. Landscape pa tterns and legacies resulting from large, infrequent forest disturbances. Ecosystems 1 :497-510. Franklin, J. F., and C. T. Dyrness. 1973. Natural vegetation of Oregon and Washington. Pacific Northwest Forest and Range Expe riment Station, Fore st Service, U.S. Dept. of Agriculture, Portland, Or.,. Gabbe, A. P., S. K. Robinson, and J. D. Brawn. 2002. Tree-species preferences of foraging insectivorous birds: Implicati ons for floodplain forest restoration. Conservation Biology 16 :462-470. Gotway, C. A., and R. D. Wolfinger. 2003. Spatial prediction of counts and rates. Statistics in Medicine 22 :1415-1432. Grinell, J. 1917. The niche-relationshi ps of the California Thrasher. Auk 34 :427-434. Grinnell, J. 1917. The niche-relationshi ps of the California Thrasher. Auk 34 :427-434. Guenette, J. S., and M. A. Villard. 2005. Thre sholds in forest bird response to habitat alteration as quantitative targets fo r conservation. Conservation Biology 19 :11681180.
105 Guisan, A., and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8 :993-1009. Hagar, J., S. Howlin, and L. Ganio. 2004. Short-term response of songbirds to experimental thinning of young Douglas-f ir forests in the Oregon Cascades. Forest Ecology and Management 199 :333-347. Hanowski, J. A. M., and G. J. Niemi. 1995. A comparison of onand off-road bird counts: Do you need to go off road to count birds accurately ? Journal of Field Ornithology 66 :469-483. Hansen, A. J., W. C. McComb, R. Vega, M. G. Raphael, and M. Hunter. 1995. Bird habitat relationships in natural and mana ged forests in the West Cascades of Oregon. Ecological Applications 5 :555-569. Hayes, J. P., J. M. Weikel, and M. M. P. Huso. 2003. Response of birds to thinning young Douglas-fir forests. Ecological Applications 13 :1222-1232. Hill, J. K., C. D. Thomas, and B. Hun tley. 1999. Climate and habitat availability determine 20th century changes in a butterf ly's range margin. Proceedings of the Royal Society of London Series B-Biological Sciences 266 :1197-1206. Holmes, R. T., and S. K. Robinson. 1981. Tree species preferences of foraging insectivorous birds in a northe rn hardwoods forest. Oecologia 48 :31-35. Holt, R. D., and M. S. Gaines. 1992. Analysis of adaptation in heterogeneous landscapes: implications for the evolution of f undamental niches. Evolutionary Ecology 6 :433-447. Hoover, J. P. 2003. Decision rules for site fi delity in a migrator y bird, the Prothonotary Warbler. Ecology 84 :416-430. Huff, M. H., N. E. Seavy, J. D. Alexander, and C. J. Ralph. 2005. Fire and birds in the maritime Pacific Northwest. Studies in Avian Biology 30 . Hutchinson, G. E. 1965. The ecological theat er and the evolutionary play. Yale University Press, New Haven, CT. Hutto, R. L. 1995. Composition of bird communities following stand-replacement fires in Northern Rocky-Mountain (USA) coni fer forests. Conservation Biology 9 :10411058. James, F. C., and N. O. Wamer. 1982. Rela tionships between temperate forest bird communities and vegetation Structure. Ecology 63 :159-171. Janes, S. W. 1994. Variation in the species composition and mean body-size of an avian foliage-gleaning guild along an elevati onal gradient: correlation w ith arthropod body-size. Oecologia 98 :369-378.
106 Johnson, M. D., and T. W. Sherry. 2001. Effects of food availability on the distribution of migratory warblers among habitats in Jamaica. Journal of Animal Ecology 70 :546-560. Jones, D. D., L. M. Conner, T. H. Store y, and R. J. Warren. 2004. Prescribed fire and raccoon use of longleaf pine forests: implications for managing nest predation? Wildlife Society Bulletin 32 :1255-1259. Jones, D. D., L. M. Conner, R. J. Warren, and G. O. Ware. 2002. The effect of supplemental prey and prescribed fire on success of artificial nests. Journal of Wildlife Management 66 :1112-1117. Keller, J. K., M. E. Richmond, and C. R. Smith. 2003. An explanation of patterns of breeding bird species richness and density following clearcutting in northeastern USA forests. Forest Ecology and Management 174 :541-564. Kilgo, J. C. 2005. Harvest-related edge eff ects on prey availability and foraging of hooded warblers in a bottomland hardwood forest. Condor 107 :627-636. Knight, T. M., and R. D. Holt. 2005. Fire gene rates spatial gradient s in herbivory: An example from a Florida sandhill ecosystem. Ecology 86 :587-593. Kotliar, N. B., S. J. Hejl, R. L. Hutto, V. A. Saab, C. P. Melcher, and M. E. McFadzen. 2002. Effects of fire and post-fire sa lvage logging on avian communities in conifer-dominated forests of the wester n United States. Studies in Avian Biology 25 :49-64. Kreisel, K. J., and S. J. Stein. 1999. Bird use of burned and unburned coniferous forests during winter. Wilson Bulletin 111 :243-250. Kunzmann, M. R., K. Ellison, K. L. Purcell, R. R. Johnson, and L. T. Haight. 2002. California Towhee ( Pipilo crissalis ). in A. Poole and F. Gill, editors. The Birds of North America. The Birds of North America, Inc., Philadelphia, PA. Liu, C. R., P. M. Berry, T. P. Dawson, and R. G. Pearson. 2005. Sel ecting thresholds of occurrence in the prediction of species distributions. Ecography 28 :385-393. Lyons, J. E. 2005. Habitat specific foraging of Prothonotary Warble rs: Deducing habitat quality. Condor 107 :41-49. MacArthur, R., and J. W. MacArthur. 1961. On bird species-diversity. Ecology 42 :594598. Mackey, R. L., and D. J. Currie. 2001. Th e diversity-disturbance relationship: Is it generally strong and peaked? Ecology 82 :3479-3492.
107 Martin, T. E. 1988. On the advantage of being different: nest predation and the coexistence of bird species. Proceedings of the National Academy of Sciences of the United States of America 85 :2196-2199. McGarigal, K., and W. C. McComb. 1995. Relationships be tween landscape structure and breeding birds in the Oregon Coast Range. Ecological Monographs 65 :235260. McNeill, L. 1991. Interpolation and smoothing of binomial data for the Southern African Bird Atlas Project. South African Statistical Journal 25 :129-136. Meehan, T. D., and T. L. George. 2003. Short-te rm effects of modera teto high-severity wildfire on a disturbancedependent flycatcher in northwest California. Auk 120 :1102-1113. Metcalfe, N. B. 1984. The effects of habitat on the vigilance of shorebirds: is visibility important? Animal Behaviour 32 :981-985. Mettke-Hofmann, C., C. Ebert, T. Schmidt, S. Steiger, and S. Stieb. 2005. Personality traits in resident and migrat ory warbler species. Behaviour 142 :1357-1375. Mills, G. S., J. B. Dunning, and J. M. Ba tes. 1991. The relations hip between breeding bird density and vegetation volume. Wilson Bulletin 103 :468-479. Moreira, F., A. Delgado, S. Ferreira, R. Borral ho, N. Oliveira, M. Inacio, J. S. Silva, and F. Rego. 2003. Effects of prescribed fire on vegetation structure and breeding birds in young Pinus pinaster stands of northern Port ugal. Forest Ecology and Management 184 :225-237. Moretti, M., M. K. Obrist, and P. Duelli. 2004. Arthropod biodiversity after forest fires: winners and losers in the winter fire regime of th e southern Alps. Ecography 27 :173-186. Morton, M. L., M. E. Pereyra, J. D. Cranda ll, E. A. MacDougall-Shackleton, and T. P. Hahn. 2004. Reproductive effort and return rates in the Mounta in White-crowned Sparrow. Condor 106 :131-138. Munzel, U., and L. A. Hothorn. 2001. A unifi ed approach to simultaneous rank test procedures in the unbalanced one -way layout. Biometrical Journal 43 :553-569. Nagy, L. R., and R. T. Holmes. 2004. Fact ors influencing fecundity in migratory songbirds: is nest predation the most important? Journal of Avian Biology 35 :487-491. National Interagency Fire Center. 20 04. Wildland Fire Statistics. URL http://www.nifc.gov/stats/ wildlandfirestats.html (May 2006).
108 Niwa, C. G., and R. W. Peck. 2002. Influen ce of prescribed fire on carabid beetle (Carabidae) and spider (Araneae) assembla ges in forest litter in southwestern Oregon. Environmental Entomology 31 :785-796. O'Laughlin, J. 2005. Conceptual model for co mparative ecological risk assessment of wildfire effects on fish, with and wit hout hazardous fuel treatment. Forest Ecology and Management 211 :59-72. Odion, D., E. Frost, J. Strittholt, H. Jiang, D. DellaSala, and M. Moritz. 2004. Patterns of fire severity and forest conditions in the western Klamath Mountains, California. Conservation Biology 18 :927-936. Orgeas, J., and A. N. Andersen. 2001. Fire and biodiversity: respons es of grass-layer beetles to experimental fire regimes in an Australian tropical savanna. Journal of Applied Ecology 38 :49-62. Osenberg, C. W., R. J. Schmitt, S. J. Holb rook, K. E. Abusaba, and A. R. Flegal. 1994. Detection of environmental impacts: natu ral variability, effect size, and power analysi. Ecological Applications 4 :16-30. Parker, K. R., and J. A. Wiens. 2005. Assessing recovery following environmental accidents: environmental variation, ec ological assumptions, and strategies. Ecological Applications 15 :2037-2051. Pippin, W. F., and B. Nichols. 1996. Observa tions of arthropod populations following the La Mesa Fire of 1977. Pages 161-165 in C. D. Allen, editor. Fire effects in Southwestern forests: proceedings of th e second La Mesa Fire symposium. USDA Forest Service General T echnical Report RM-GTR-286. Pons, P., and J. M. Bas. 2005. Open-habitat bird s in recently burned ar eas: the role of the fire extent and species' habitat breadth. Ardeola 52 :119-131. Pons, P., P. Y. Henry, G. Gargallo, R. Prodon, and J. D. Lebreton. 2003. Local survival after fire in Mediterranean shrublands : combining capture-recapture data over several bird species. Population Ecology 45 :187-196. Post, D. M. 2002. The long and short of f ood-chain length. Trends in Ecology and Evolution 17 :269-277. Pyne, S. J. 1982. Fire in America: a cultural history of wildland and rural fire. Princeton University Press, Princeton, NJ. R Development Core Team. 2005. R: A langua ge and environment for statistical computing. Version 2.1.1. URL http://www.R-project.org (May 2006). Rabenold, K. N. 1978. Foraging strategies, dive rsity, and seasonality in bird communities of Appalachian spruce-fir fo rests. Ecological Monographs 48 :397-424.
109 Ralph, C. J., G. R. Geupel, P. Pyle, T. E. Martin, and D. F. DeSante. 1993. Handbook of field methods for monitoring landbirds. USDA Forest Service General Technical Report PSW-GTR-144. Robinson, S. K., and R. T. Holmes. 1982. Fo raging behavior of forest birds: the relationships among search tactics, diet, and habitat structure. Ecology 63 :19181931. Rogers, L. E., W. T. Hinds, and R. L. Buschbom. 1976. Genera l weight vs length relationship for insects. Annals of the Entomological Society of America 69 :387389. Root, T. 1988. Environmental-factors associ ated with avian distributional boundaries. Journal of Biogeography 15 :489-505. Rotenberry, J. T. 1985. The role of hab itat in avian community composition: physiognomy or floris tics? Oecologia 67 :213-217. Rushton, S. P., S. J. Ormerod, and G. Ke rby. 2004. New paradigms for modelling species distributions? Journa l of Applied Ecology 41 :193-200. Saab, V. 1999. Importance of spatial scale to habitat use by breeding birds in riparian forests: a hierarchical analys is. Ecological Applications 9 :135-151. Saab, V. A., J. Dudley, and W. L. Thom pson. 2004. Factors influencing occupancy of nest cavities in recently burned forests. Condor 106 :20-36. Sallabanks, R., R. A. Riggs, and L. E. Cobb. 20 02. Bird use of forest structural classes in grand fir forests of the Blue Mountains, Oregon. Forest Science 48 :311-321. Seavy, N. E. In press. Physiological correla tes of habitat associat ion in East African sunbirds. Journal of Zoology. Sedgwick, J. A. 2004. Site fidelity, territory fidelity, and natal philopatry in willow flycatchers ( Empidonax traillii ). Auk 121 :1103-1121. Shochat, E., S. B. Lerman, M. Katti, a nd D. B. Lewis. 2004. Linking optimal foraging behavior to bird community structure in an urban-desert landscape: Field experiments with artificial food patches. American Naturalist 164 :232-243. Siepielski, A. M. 2006. A possible role for re d squirrels in struct uring breeding bird communities in lodgepole pine forests. Condor 108 :232-238. Skinner, C. N. 1995. Change in spatial charact eristics of forest openings in the Klamath Mountains of Northwestern California, USA. Landscape Ecology 10 :219-228.
110 Smucker, K. M., R. L. Hutto, and B. M. St eele. 2005. Changes in bird abundance after wildfire: importance of fire severity a nd time since fire. Ecological Applications 15 :1535-1549. Spies, T. A., M. A. Hemstrom, A. Youngblood, and S. Hummel . 2006. Conserving oldgrowth forest diversity in disturba nce-prone landscapes. Conservation Biology 20 :351-362. Stephens, P. A., S. W. Buskirk, G. D. Hayward, and C. M. Del Rio. 2005. Information theory and hypothesis testi ng: a call for pluralism. Journal of Applied Ecology 42 :4-12. Stephens, S. L. 1998. Evaluation of the effect s of silvicultural and fuels treatments on potential fire behaviour in Sierra Nevada mixed-coni fer forests. Forest Ecology and Management 105 :21-35. Stewart-Oaten, A., and J. R. Bence. 2001. Temporal and spat ial variation in environmental impact assessment. Ecological Monographs 71 :305-339. Stuart-Smith, A. K., and J. P. Hayes. 2003. In fluence of residual tree density on predation of artificial and natural songbird ne sts. Forest Ecology and Management 183 :159176. Swets, J. A. 1988. Measuring the accur acy of diagnostic systems. Science 240 :12851293. Taylor, A. H., and C. N. Skinner. 1998. Fi re history and landscape dynamics in a latesuccessional reserve, Klamath Mountains, California, USA. Forest Ecology and Management 111 :285-301. Thonicke, K., S. Venevsky, S. Sitch, and W. Cramer. 2001. The role of fire disturbance for global vegetation dynamics: coupling fire into a Dynamic Global Vegetation Model. Global Ecology and Biogeography 10 :661-677. Tiedemann, A. R., J. O. Klemmedson, and E. L. Bull. 2000. Solution of forest health problems with prescribed fire: are forest productivity and wildlife at risk? Forest Ecology and Management 127 :1-18. van Mantgem, P., M. Schwartz, and M. B. Keifer. 2001. Monitoring fire effects for managed burns and wildfires: coming to terms with pseudore plication. Natural Areas Journal 21 :266-273. Venables, W. N., and B. D. Ripley. 2002. Mode rn Applied Statistics with S. Springer, New York. Verner, J., and T. A. Larson. 1989. Richness of breeding bird species in mixed-conifer forests of the Sierra-Nevada, California. Auk 106 :447-463.
111 Vigallon, S. M., and J. M. Marzluff. 2005. Is nest predation by Steller's Jays ( Cyanocitta stelleri ) incidental or the result of a specialized search strategy? Auk 122 :36-49. Wagner, R. G., M. Newton, E. C. Cole, J. H. Miller, and B. D. Shiver. 2004. The role of herbicides for enhancing forest produc tivity and conserving land for biodiversity in North America. Wildlife Society Bulletin 32 :1028-1041. Whelan, C. J. 1989. Avian foliage structure pr eferences for foraging and the effect of prey biomass. Animal Behaviour 38 :839-846. Whelan, C. J. 2001. Foliage structure influences foraging of insecti vorous forest birds: An experimental study. Ecology 82 :219-231. Whittaker, R. H. 1960. Vegetation of the Si skiyou Mountains, Oregon and California. Ecological Monographs 30 :280-338. Wiens, J. A., J. T. Rotenberry, and B. Va nhorne. 1986. A lesson in the limitations of field experiments: shrubsteppe birds and habitat alteration. Ecology 67 :365-376. Wootton, J. T. 1998. Effects of disturbance on species diversity: A multitrophic perspective. American Naturalist 152 :803-825.
112 BIOGRAPHICAL SKETCH Nathaniel (Nat) E. Seavy grew up on Puget Sound in Washington State and attended The Evergreen State College, in Olympia. At Evergreen, Nat developed an interest in ornithology that was fostered by Dr. Steven Herman. Since 1993, Nat has worked on ornithological research project s in Washington, Oregon, Guatemala, and Uganda. These projects have included resear ch on the breeding biology of raptors and owls, foraging ecology and physiology of tropica l birds, and habita t associations of passerine birds. Nat received his masterâ€™s degree from the Department of Zoology at the University of Florida in 2001. After receivi ng his Ph.D., Nat plans to remain involved with research on mechanisms that generate pa tterns of animal distribution and predicting how environmental change will alter these patterns.