IMPACTS OF STRUCTURE RETENTION ON AVIAN ECOLOGY IN MANAGED FORESTS OF THE PACIFIC NORTHWEST By Daniel W. Linden A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Fisheries and Wildlife Ecology, Evolutionary Biology, and Behavior 2011     ABSTRACT IMPACTS OF STRUCTURE RETENTION ON AVIAN ECOLOGY IN MANAGED FORESTS OF THE PACIFIC NORTHWEST By Daniel W. Linden Structure retention is a practice used in managed forests to assist the conservation of biological diversity, whereby green trees, dead trees i.e., snags , and downed wood are retained during timber harvest. This activity is recognized as beneficial; however, there is little scientific support to guide the management prescriptions e.g., density, patch sizes, distribution pattern . I quantified wildlife responses to structure retention attributes in harvest units across 4 regions in the Pacific Northwest, USA, during the summers of 2008– 2010. These 4 regions encapsulated ecological and management variability of intensively managed forests ranging from Washington to northern California. Bird species were observed at retention sites across the 4 regions and white‐headed woodpeckers Picoides albolarvatus were studied more intensively in California. I used a hierarchical modeling framework to separately model the processes related to data collection e.g., detection probability from those related to the state variables of interest e.g., species occupancy . This framework allowed for the examination of factors which influenced each process. The results provide empirical support for management strategies that can improve the effectiveness of structure retention for addressing biodiversity objectives. In the first chapter, I report on a multi‐species occurrence model which estimated occupancy and detection probabilities for all bird species observed at retention sites. Retained tree count was associated with an increased occupancy probability for 70% of the observed species. The community response to tree count was consistent across all study     areas and years – species richness estimates increased with tree count and approximated a species‐area curve. Distance to nearest mature forest did not significantly affect occupancy probability for any observed species, and therefore, had no significant relationship with species richness. These results suggest that the diversity of birds using structure retention in harvest units can be maximized at patches of 10–15 rotation age trees. In the second chapter, I used a multistate site occupancy model to estimate the probabilities of occupancy and nesting for white‐headed woodpeckers in harvest units where structure retention was present. Snag density had a significant positive association with nesting probability. Mature forest proportion was negatively associated with nesting probability, though there was considerable uncertainty. High occupancy 0.98 and nesting 0.89 probabilities suggest that current structure retention policies have provided the necessary habitat conditions for white‐headed woodpecker nesting in harvest units of northern California. Forest managers can maximize nesting probability by retaining 2 snags/ha during harvest. In the third chapter, I examined reproductive success for white‐headed woodpeckers using models of nest survival and number of young fledged. We documented a high nest success rate 0.85 and found that successful nests were most likely to produce 3 young. None of the habitat variables that were examined at multiple scales were able to significantly describe variation in either metric of reproductive success. Current forest management practices in northern California appear to be providing habitat conditions that are conducive to the species’ persistence in this region.     ACKNOWLEDGEMENTS There are many people to whom I am grateful for being a part of my experience as a doctoral student. First, I would like to thank my advisor, Dr. Gary Roloff, for convincing me to return to Michigan State for another degree and for supporting me in every way possible. Gary always treated me as an intellectual peer and for that I am appreciative. I thank my graduate committee members, Dr. A.J. Kroll, Dr. Catherine Lindell, and Dr. Brian Maurer, for providing support and improving my research. Dr. Andrew McAdam also provided a valuable perspective during my program. Special thanks to A.J. for acting as a second mentor and for always delivering an insightfully direct opinion. The National Council for Air and Stream Improvement, Inc., NCASI provided the majority of funding for my study, as well as the member companies of the Washington Forest Protection Association, Oregon Forest Industries Council, and American Forest Resource Council. Dr. Larry Irwin of NCASI was particularly instrumental in securing support from the many project partners. I thank the project partners from Hancock Timber Resource Group, Longview Fibre Paper and Packaging, Inc., Plum Creek Timber Company, Inc., Rayonier, Inc., Sierra Pacific Industries, Timber Products, Inc., and Weyerhaeuser Company for providing property access and logistical support. Special thanks to Tim McBride of Hancock and Stu Farber of Timber Products for being actively engaged in the project. Dr. Dan Varland afforded me the opportunity to worth with some local teachers in Washington on outreach projects and was also a good friend. A number of people were critical to the success of the data collection. Laurie Clark, Dennis Rock and Suzanne Rock, all of NCASI, provided field assistance and helped iv    tremendously with field logistics. Both Laurie and the Rocks always made me feel welcome in their homes and often provided some much needed sustenance – I thank them for their friendships. A number of quality technicians participated in the data collection, including Lauren Bailey, Pete Datema, Adam Dillon, Krissy Frey, Becky Fuda, Jamie Garren, James Harrington, Blake Hatling, Stacey Hollis, Vanessa Hudson, Kori Hutchison, Darren Kite, John Longcore, Eva Masin, Elaina Snyder, and Nick VanDyken. John and Kori were especially good friends and I am honored that my project was responsible for the union of such a great couple. We will always have Mt. Shasta. Scholarships and fellowships from the Rocky Mountain Goat Foundation, the Department of Fisheries and Wildlife, the FW Graduate Student Organization, the Ecology, Evolutionary Biology, and Behavior Program, and the Graduate School made my life easier. The support staff in the Department of Fisheries and Wildlife was always helpful in addressing my needs with great expertise and organization. Finally, I would like to thank my friends and family. Many friends have come and gone during my time at MSU. I have been fortunate to share such good times with so many great people over the years. Playing countless Sundays of football thanks to Ace Sarnelle , attending Spartan basketball and football games, and participating with the Art’s Bar soccer team were among the many highlights. My parents, Laurie and Fred, were especially supportive during my time at MSU and I can never thank them enough for all that they have provided me. My sisters and extended family also put up with my absence and never made me feel guilty for being away. And to Leslie – words cannot express the gratitude I feel for having introduced myself to you so bluntly and for later suggesting we watch the Red Sox together. Best decision ever. v    TABLE OF CONTENTS LIST OF TABLES .................................................................................................................................................... viii  LIST OF FIGURES ..................................................................................................................................................... ix  INTRODUCTION ........................................................................................................................................................ 1  CHAPTER 1 CONSERVING AVIAN RICHNESS THROUGH STRUCTURE RETENTION IN MANAGED FORESTS OF THE PACIFIC NORTHWEST ...................................................................................................... 6  Abstract .................................................................................................................................................................... 6  1. Introduction...................................................................................................................................................... 8  2. Methods ........................................................................................................................................................... 11  2.1. Study areas ............................................................................................................................................ 11  2.2. Retention site selection and measurement ............................................................................. 12  2.3. Bird surveys .......................................................................................................................................... 13  2.4. Modeling framework ......................................................................................................................... 14  3. Results .............................................................................................................................................................. 20  4. Discussion ...................................................................................................................................................... 23  . 5. Management implications ....................................................................................................................... 28  CHAPTER 2 MULTISTATE SITE OCCUPANCY ESTIMATION OF WHITE‐HEADED WOODPECKERS PICOIDES ALBOLARVATUS IN MANAGED FORESTS ......................................................................... 45  Abstract ................................................................................................................................................................. 45  1. Introduction................................................................................................................................................... 47  2. Methods ........................................................................................................................................................... 50  2.1. Study area .............................................................................................................................................. 50  . 2.2. Woodpecker surveys ......................................................................................................................... 51  2.2. Habitat covariates ............................................................................................................................... 52  2.4. Modeling framework ......................................................................................................................... 54  3. Results .............................................................................................................................................................. 58  4. Discussion ...................................................................................................................................................... 60  . 5. Management implications ........................................................................................................................ 64  CHAPTER 3 FACTORS AFFECTING WHITE‐HEADED WOODPECKER PICOIDES ALBOLARVATUS REPRODUCTIVE SUCCESS IN MANAGED FORESTS WITH STRUCTURE RETENTION ............ 71  vi    Abstract ................................................................................................................................................................. 71  1. Introduction................................................................................................................................................... 72  2. Methods ........................................................................................................................................................... 74  2.1. Study area .............................................................................................................................................. 74  . 2.2. Nest searching and monitoring .................................................................................................... 76  . 2.2. Model covariates ................................................................................................................................. 77  2.4. Statistical analysis .............................................................................................................................. 79  2.4.1. Nest survival ...................................................................................................................................... 79  2.4.2. Number of young fledged ............................................................................................................ 80  2.4.3. Model parameter estimation ...................................................................................................... 82  3. Results ............................................................................................................................................................... 83  4. Discussion ....................................................................................................................................................... 84  . 5. Management implications ........................................................................................................................ 87  APPENDIX A ............................................................................................................................................................. 97  Appendix A.1. WinBUGS code for community occupancy model chapter 1 ....................... 97  Appendix A.2. WinBUGS code for multistate occupancy model chapter 2 ....................... 100  Appendix A.3. WinBUGS code for nest survival model chapter 3 ......................................... 102  Appendix A.4. WinBUGS code for offspring model chapter 3 ................................................. 103  LITERATURE CITED ........................................................................................................................................... 106  vii    LIST OF TABLES CHAPTER 1 Table 1.1. Attributes of harvest units selected for observing bird use at structure retention sites in managed forests of the Pacific Northwest. ................................................................................. 30  Table 1.2. List of species detected using retention sites from 2008–2009 in managed forests of the Pacific Northwest, including habitat group based on nesting and/or foraging requirements , common name, scientific name, American Ornithologist’s Union alpha code, and total number of detections. ...................................................................................................................... 31  CHAPTER 2 Table 2.1. Posterior summaries for model parameters means, standard deviations SD , and 95% credible intervals lower 0.025, upper 0.975 from the multistate model estimating white‐headed woodpecker occupancy and nesting probabilities during 2010 among harvest units in California, USA. Parameters include intercepts and regression coefficients of standardized variables from the logit‐linear models for probabilities of occupancy 1 , nesting given occupancy 2 , detection of occupancy given the site is occupied without nesting 1 , detection of occupancy given the site is occupied with nesting 2 , and detection of nesting given the site is occupied with nesting . .............. 65  CHAPTER 3 Table 3.1. Summary of potential model covariates describing habitat collected at multiple scales for white‐headed woodpecker nests from 2009–2010 in northern California, USA. Habitat variables were measured at the level of the nest tree, nest site, stand, and landscape for all monitored nests n 89 . Variables in bold were selected for use in the models of reproductive metrics. .......................................................................................................................................... 89  Table 3.2. Count of white‐headed woodpecker nests that successfully fledged young during 2009 and 2010 on private timberlands of northern California, USA. Mean and standard error SE for number of young fledged are also presented for each year. ................................. 90  Table 3.3. Posterior summaries for model parameters means, standard deviations SD , and 95% credible intervals 0.025 and 0.975 from the logit linear models for nest survival and number of young fledged. Temporal covariates were year‐specific linear and quadratic 2 and nest completion date for effects of ordinal date for daily nest survival , 2 . Habitat covariates were nest height number of young . , . , nest tree diameter , snag density in the harvest unit and mature forest proportion in a 1‐km radius surrounding the nest viii    , . .................... 91  LIST OF FIGURES CHAPTER 1 Figure 1.1. Map of study area locations within ecoregions and hydrologic units in the Pacific Northwest, USA. Region codes represent the following areas: CWA central Washington; SWA southwest Washington; SOR southwest Oregon; NCA northern California. .................................................................................................................................................................. 33  Figure 1.2. Mean probabilities of occupancy and detection for all species detected within the 4 regions during 2008 filled squares and 2009 open squares . Each species‐specific probability is defined by the mean of the posterior distribution for the intercept in the logit‐linear models. Region codes represent the following areas in the Pacific Northwest: CWA central Washington; SWA southwest Washington; SOR southwest Oregon; NCA northern California. ......................................................................................................................................... 35  Figure 1.3. Posterior estimates mean with 95% credible interval of the regression coefficients from logit‐linear models of detection for each species. The coefficients correspond to the following covariates: a ordinal date, b squared ordinal date, and c previous detection. Species are ordered by increasing mean estimate of the coefficient and distinguished by habitat group into open/generalist species open squares and forest species filled squares . ...................................................................................................................................... 37  Figure 1.4. Posterior estimates mean with 95% credible interval of the regression coefficients from logit‐linear models of occupancy for each species. The coefficients correspond to the following covariates: a tree count, b squared tree count, c edge distance, and d squared edge distance. Species are ordered by increasing mean estimate of the coefficient and distinguished by habitat group into open/generalist species open squares and forest species filled squares . ............................................................................................ 39  Figure 1.5. Mean estimated site richness for all species a, b , forest species c, d , and open/generalist species e, f at each retention site compared to tree count in 2008 left and 2009 right . Study areas are colored with central Washington CWA blue, southwest Washington SWA green, southwest Oregon SOR yellow, and northern California NCA red. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. ............................. 41  Figure 1.6. Mean estimated site richness for all species a, b , forest species c, d , and open/generalist species e, f at each retention site compared to edge distance m in 2008 left and 2009 right . Study areas are colored with central Washington CWA blue, southwest Washington SWA green, southwest Oregon SOR yellow, and northern California NCA red. ....................................................................................................................................... 43  ix    CHAPTER 2 Figure 2.1. Harvest units selected to survey white‐headed woodpecker occurrence and reproduction during 2010 near Mt. Shasta, California, USA. .............................................................. 67  Figure 2.2. Model predicted relationship between conditional nesting probability, 2 , of white‐headed woodpeckers and harvest unit snag density snags 25.4cm diameter/ha on industrial timberlands in northern California. Mean probability with 95% credible intervals are presented. ...................................................................................................................................... 68  Figure 2.3. Model predicted relationship between conditional nesting probability, 2 , of white‐headed woodpeckers and proportion of mature forest quadratic mean diameter 60cm within a 1‐km radius on industrial timberlands in northern California. Mean probability with 95% credible intervals are presented. ...................................................................... 69  Figure 2.4. Model predicted mean probabilities of detection with 95% credible intervals over the 4 survey intervals. The probabilities are detection of occupancy given the site is occupied without nesting 1 , open triangles , detection of occupancy given the site is occupied with nesting 2 , filled squares , and detection of nesting given the site is occupied with nesting , open squares . ................................................................................................. 70  CHAPTER 3 Figure 3.1. Harvest units selected to study white‐headed woodpecker reproductive success from 2009–2010 near Mt. Shasta, California, USA. ................................................................................. 93  Figure 3.2. Predicted relationship between daily nest survival and ordinal date for white‐ headed woodpeckers during 2009 and 2010 in northern California, USA. Lines represent mean probabilities solid and 90% credible intervals dashed . .................................................. 94  . Figure 3.3. Predicted relationship between expected number of young fledged and ordinal date of nest completion for white‐headed woodpeckers during 2009 and 2010 in northern California, USA. Lines represent mean probabilities solid and 90% credible intervals dashed . ................................................................................................................................................................... 95  x    INTRODUCTION Loss of biodiversity caused by human activities is recognized as a major threat to ecological communities across the planet Wilson, 1999 . This threat is related not only to ethical and aesthetic values, but also goods and services derived from diverse ecosystem properties Hooper et al., 2005 . Biodiversity is hypothesized to serve as insurance that counteracts environmental fluctuations which could otherwise severely alter the properties of ecosystems Folke et al., 1996; Naeem and Li, 1997 . The underlying premise is that communities with greater biodiversity are more resilient to environmental disturbances. As such, the development of policies that incorporate the maintenance of biodiversity into resource management has become more widespread Redford and Richter, 1999 . Biodiversity is often used synonymously with species richness, though the former incorporates more than simply taxonomic diversity. The concept of biodiversity refers to the diversity of living organisms and their interactions with the environment that exist within and across all scales—spatial, temporal, and biotic—of ecological organization Noss, 1990; Angermeier and Karr, 1994; Hooper et al., 2005 . The complexity of this definition is difficult for humans to comprehend thereby exacerbating problems already associated with effective resource management when sustaining biodiversity becomes the conservation goal Bunnell and Huggard, 1999 . The species pool within a community is often used as a surrogate for biodiversity because it represents an entity that is easy to measure and understand Bunnell and Huggard, 1999 . Using taxonomic diversity alone may not be effective for quantifying biodiversity, as species typically do not have equal 1    impacts on the processes within an ecosystem due to differences in their functional traits Chapin et al., 2000; Hooper et al., 2005 . Functional diversity attempts to address both species and ecological processes Petchey and Gaston, 2002 . A reduction in ecosystem functions is hypothesized to parallel a reduction in biodiversity for a given community, though this relationship is complex and at times ambiguous Loreau et al., 2001 . There are still many unanswered questions regarding the effects of biodiversity on ecosystem productivity and stability, and the efficiency of resource use by organisms; the context dependency of these relationships makes the formulation of consistent ecological models difficult Naeem, 2002 . It can be postulated that since species must be occupying fundamentally different niche spaces to coexist within a given community, any loss of species will result in a loss of ecosystem function; the magnitude and significance of this loss are proportional to the difference in functional traits between the lost species and its neighbors Lawton et al., 1998 . Functional traits that are unique and possessed by only a few species present the greatest threat of lost ecosystem function when biodiversity is reduced Marcot and Vander Hayden, 2001 . Communities with a high level of functional redundancy among species, which is usually correlated with high biodiversity, are more resilient to changes in environmental conditions Naeem, 1998; Marcot and Vander Hayden, 2001 . Keystone species typically occupy a unique niche space with little redundancy, resulting in their relatively disproportionate influence on the ecosystem Power et al., 1996 . Especially important are keystone species that act as ecosystem engineers, which according to Lawton and Jones 1995:142 , “directly or indirectly modulate the availability of resources other than themselves to other species, by causing physical state changes in 2    biotic or abiotic materials.” While single species management is generally criticized for being one‐dimensional, the importance of keystone species is explicit and represents a unification of ecosystem management practices for the conservation of biodiversity Simberloff, 1998; Callicott et al., 1999 . Despite the uncertainty in its utility and meaning, the maintenance of biodiversity has become a goal for forest management Bunnell and Johnson, 1998 . Forest management evolved during the last century from a focus on the extraction of valuable commodities to one that recognized multiple interests and sought to maintain the essential elements of forest ecosystems. Aubry 2007 summarized the history of this evolution in the Pacific Northwest region of the USA, where an increased understanding of forest ecology and relationships between vertebrate species and unique structural elements of vegetation communities e.g., riparian and old‐growth forests led to the enactment of multiple environmental laws and major policy changes for public lands during the 1980s. The concept of a “new forestry” was proposed as a means to balance the conflicting goals of timber extraction and biodiversity conservation Franklin, 1989 . Among the components of new forestry was legacy retention, a stand‐scale silvicultural practice whereby legacies of the original stand e.g., large live and dead trees, and other structural habitat elements are retained during harvest Franklin, 1989 . Cavity‐nesting species require dead trees i.e., snags for nest sites, and some primary cavity excavators act as keystone species and ecosystem engineers by creating cavities for other species Aubry and Raley, 2002; Bednarz et al., 2004; Martin et al., 2004 . Structure retention would, therefore, provide a structural diversity more similar to that of naturally disturbed forests than would traditional clearcutting. This practice could potentially facilitate a wider range of ecological 3    functions and species and, thus, promote biodiversity Hansen et al., 1995; Franklin et al., 2002 . Implementation of structure retention practices are mandated for both public and private nonfederal forests in the Pacific Northwest by state forest practices rules Washington Forest Practices Board, 2002; Oregon Department of Forestry, 2005; California Department of Forestry and Fire Protection, 2007 . These rules attempt to reduce the impacts of timber harvesting on ecosystem properties both biotic and abiotic through operational guidelines related to harvesting and replanting activities e.g., road construction on steep slopes and proximity to wetlands . Retention of structural features is incorporated into forest practices rules by restricting harvest within riparian zones and requiring minimum densities of standing live and dead trees and down logs each with minimum size dimensions within harvest units. Some private landowners also participate in forest stewardship programs that may require additional structural retention measures to meet a third‐party certification. For example, as of 2011 the Sustainable Forestry Initiative enrolled 74 million ha of forestlands in North America Sustainable Forestry Initiative, 2011 . The certification standard includes an indicator specifically pertaining to retained structures. While these management efforts are likely beneficial, it is unclear how specific properties of retained structures e.g., spatial distribution, density, size actually influence biodiversity Franklin et al., 2002 . Our goal was to quantify the contribution of retained structures for providing wildlife habitat and influencing biodiversity within managed forests. The National Council for Air and Stream Improvement NCASI recently identified the need for research to evaluate the effectiveness of structure retention policies, including their contribution to 4    wildlife diversity and their role in providing habitat for cavity‐nesting species NCASI, 2008 . We quantified the physical attributes of retained structures and examined relationships between structure attributes and wildlife use, with special emphasis on cavity‐nesting birds. We focused on retained structures that were specifically preserved during harvesting to meet or exceed regulations set by state forest practices rules, including large green trees, dead trees or snags , and patches of these elements. Results of the study are presented in 3 separate chapters. The first chapter examines use of retained structures by the entire bird community to understand how structure attributes influence species richness across 4 regions of the Pacific Northwest ranging from Washington to northern California. The second and third chapters focus on white‐headed woodpecker Picoides albolarvatus ecology in harvest units of northern California. White‐headed woodpeckers may serve as a keystone species in dry forests of western North America where they are abundant, given their role as primary excavators. The analyses examine how habitat attributes at multiple scales influence stand‐level nesting probability chapter 2 and reproductive success chapter 3 for white‐headed woodpeckers. Each chapter concludes with a discussion of management implications. 5    CHAPTER 1 CONSERVING AVIAN RICHNESS THROUGH STRUCTURE RETENTION IN MANAGED FORESTS OF THE PACIFIC NORTHWEST Abstract Structure retention is a practice used in managed forests to assist the conservation of biological diversity, whereby green trees, dead trees i.e., snags , and downed wood are retained during timber harvest. This activity is recognized as beneficial; however, there is little scientific support to guide the management prescriptions e.g., patch sizes, distribution pattern . We quantified the response of birds to structure retention in timber harvest areas located in the Pacific Northwest. We used a hierarchical community model to examine how size tree count and location distance to forest edge of retention sites influenced the species richness of birds using sites. The modeling framework integrated multiple species‐specific occupancy models that accounted for imperfect detection to produce estimates of species richness. We sampled a biogeoclimatic gradient by selecting harvest units within four separate regions two in Washington, one each in Oregon and California that support different forest types. Observations were conducted at a random selection of retention sites e.g., patches, individual trees within harvest units to record bird use in the breeding seasons of 2008 and 2009. Estimated occupancy and detection probabilities differed by species and region. Retained tree count was associated with an increased occupancy probability for 70% of the observed species, including both forest‐ dependent and open/generalist birds. The community response to tree count was consistent across all study areas and years – species richness estimates increased with tree 6    count and approximated a species‐area curve. Edge distance did not significantly affect occupancy probability for any observed species, and therefore, had no significant relationship with species richness. These results suggest that the diversity of birds using structure retention in harvest units can be maximized at patches of 10–15 rotation age trees. Forest managers are encouraged to group green‐trees around high‐quality snags and other unique wildlife trees where possible, and to vary prescriptions across stands to provide habitat heterogeneity at the landscape scale. Keywords: avian richness, forest practices rules, hierarchical community modeling, managed forests, occupancy, Pacific Northwest, structure retention 7    1. Introduction Conservation of biological diversity is a consideration during forest management Lindenmayer and Franklin, 2002 . The effectiveness of structure retention as a means to conserve biological diversity within intensively managed forests is poorly understood NCASI, 2008 . Structure retention is a management practice whereby green trees, dead trees i.e., snags , and downed wood are retained during timber harvest. Structure retention provides residual habitat elements of mature forests that might otherwise be lost during timber harvesting and forest regeneration activities; these habitat elements are critical resources for numerous forest‐dwelling wildlife Thomas et al., 1979; Swanson and Franklin, 1992; Bull et al., 1997; Hunter and Bond, 2001; Rosenvald and Lõhmus, 2008 . For example, Rose et al. 2001 suggested that 93 wildlife species have habitat requirements associated with snags in Washington and Oregon, including 63 species of birds that use snags for nesting i.e., in cavities or foraging. Whether structure retention ultimately impacts biological diversity depends on how resultant patterns of habitat complexity and the amounts of critical resources like snags meet the requirements of individual wildlife species at relevant scales. Thus, the difficulty of managing for biodiversity i.e., accommodating the requirements of many species becomes apparent Bunnell and Johnson, 1998 . Structure retention is currently regulated at the stand‐level on state and private lands in the Pacific Northwest Washington, Oregon, and California by state forest practices rules, which include criteria for retaining minimum densities and diameters of green trees, snags and downed logs during forest management activities Washington Forest Practices Board, 2002; Oregon Department of Forestry, 2005; California Department of Forestry and Fire Protection, 8    2007 . Additionally, forest practices rules regulate activities in and around riparian management zones and other sensitive areas e.g., unstable slopes . These rules were developed to balance timber management objectives with the broader conservation needs of the forested landscape. The forest practices rules pertaining to structure retention were developed to help alleviate the potential negative impacts of even‐aged forest management on snags and snag‐dependent wildlife species NCASI, 2008 . Diameter retention rules e.g., minimum tree diameter 25cm are based on the requirement by cavity‐nesting species for decaying trees and snags that are sufficiently large Thomas et al., 1979; Zarnowitz and Manuwal, 1985; Bull et al., 1997 . Density retention rules e.g., 10 trees/ha are intended to ensure that current and future snag resources are not limiting to wildlife species Neitro et al., 1985; Bunnell et al., 2002 . Green‐tree retention is intended to serve as the source of future snags, but it may also influence stand‐level occurrence of some wildlife species depending on the density and pattern of green trees in the harvest unit Chambers et al., 1999; Schieck and Hobson, 2000; Bunnell et al., 2002; Walter and Maguire, 2005; Preston and Harestad, 2007 . In some states, forest practices rules require a specific distribution of structure retention e.g., Washington mandates a maximum distance of ~240m between any given point and a retained tree in the harvest unit , though little empirical evidence exists to support such rules. Other states allow forest managers more flexibility in deciding on the pattern i.e., clumped or dispersed and location of structures. Forest managers can maximize the utility of retained structures as a habitat resource by implementing retention patterns with the greatest positive impact on the wildlife community. 9    Evaluation of management actions requires that the biologically relevant quantity of interest i.e., state variable be properly defined for monitoring Yoccoz et al., 2001 , and species richness is a common choice for state variable in the context of biodiversity conservation. Species richness can be problematic as a management target because heterogeneity in species detectability can invalidate naïve estimates of species richness Boulinier et al., 1998 . Additionally, species typically do not equitably impact processes within an ecosystem due to differences in their functional traits Chapin et al., 2000; Hooper et al., 2005 . Hence the interpretation of species richness without some consideration of species detectability and identity can be misleading. A hierarchical modeling framework that incorporates imperfect species detection during sampling and generates species‐specific detection and occupancy proportion of sites occupied probabilities can be used to produce valid estimates of species richness Dorazio et al., 2006; Kery and Royle, 2008; Royle and Dorazio, 2008 . This type of multi‐species occupancy model has been used to evaluate the effects of management actions and landscape changes on species richness for entire avian communities and subsets of functionally similar species that may have different responses to specific habitat alterations e.g., Ruiz‐Gutierrez et al., 2010; Zipkin et al., 2010 . Our study used a hierarchical multispecies occupancy model to understand how attributes of structure retention sites in timber harvest units affect avian species richness within private industrial forests of the Pacific Northwest. Retention sites, hereafter, refer to single trees or patches of multiple trees that are left standing during harvesting operations to meet or exceed the requirements mandated by state forest practices rules. We focused on two attributes of retention sites that can be easily modified during 10    harvesting operations and likely impact occupancy by individual bird species: 1 number of trees retained i.e., tree count ; and 2 location within the harvest unit i.e., distance to forest edge . In addition to total species richness, we estimated the richness of a functional group that forages or nests or both in the forest canopy and would presumably be absent from recently harvested forests that lack structure retention. Given that structure retention is intended to help conserve biological diversity, the response by species that are typically negatively affected by timber harvest should be a priority for evaluating the effectiveness of structure retention practices. 2. Methods 2.1. Study areas Our study was conducted in 4 separate areas in the Pacific Northwest Washington, Oregon, and California that spanned multiple biogeoclimatic regions i.e., sections, sensu Bailey, 1988 , including the Northern Cascades, Oregon and Washington Coast Ranges, and the Southern Cascades Figure 1.1 . We refer to each study area by the general location in the state within which it occurred: central Washington CWA , southwest Washington SWA , southwest Oregon SOR , and northern California NCA . All study areas were located within forested watersheds on private industrial ownership with a long history of management second and third rotation stands . Watersheds contained a heterogeneous matrix of stand age classes and some had mixed ownership e.g., the NCA study area was juxtaposed with the Shasta‐Trinity National Forest . We identified all harvest units that had been logged 3–12 years prior using a geographic information system GIS . For each study area, we calculated summary statistics on stand area and elevation and randomly 11    selected ~20 harvest units that represented average conditions available on the landscape i.e., area and elevation values were 1 standard deviation from the study area mean . In total, we selected 84 harvest units that fell into 2 general area classes, small 4–12ha and large 22–61ha , and ranged from 20–1830m in elevation, and from 41.1–47.7° latitude Table 1.1 . 2.2. Retention site selection and measurement We examined the harvest units in each study area to identify retention sites that could be selected for sampling using 1‐m resolution aerial photography collected from the National Agriculture Imagery Program NAIP; http://www.apfo.usda.gov/ during 2005– 2006. We field‐verified all photo‐identified retention sites during reconnaissance surveys in 2006. Reconnaissance surveys included a thorough census of each harvest unit to identify additional retention sites not visible from the imagery e.g., individual snags . We delineated retention sites in the field based on the felling distance of trees at and around the site. For example, an individual tree was considered a distinct retention site only if the distance to the nearest adjacent tree was greater than the height of either tree. GPS coordinates were recorded at the center of each retention site and edge distance was calculated in a GIS to the nearest mature forest 40 years old as determined by interpretation of the NAIP imagery. We used a stratified random selection to identify retention sites for further sampling in each study area. Stratification was necessary to ensure that retention sites with a range of tree counts were selected given the prevalence of retention sites with tree count 1. The strata were defined by the tree count at each retention site and consisted of 4 groups: 1 tree, 2–5 trees, 6–15 trees, and 16 trees. Selected retention sites located within the same harvest unit were required to be 150m 12    apart to decrease spatial dependence, resulting in 1–9 sites per harvest unit depending on the area and shape of the unit. For each selected retention site, we used variable radius plots to tally and measure trees 12.7cm diameter at breast height dbh with a basal area factor 10 prism. 2.3. Bird surveys We observed birds using retention sites during the breeding season May–July at 241 sites in 2008 and 203 sites in 2009; 95 sites were surveyed in both years, resulting in 349 unique retention sites over the duration of the study. Each survey consisted of 2 observers recording all birds that interacted with a retention site e.g., perched on a branch, foraged on a trunk for a duration of 30min starting at a randomly selected time between sunrise and 5 hours later. Observers were located 25m to the east of the retention site to avoid disturbance and obtain the best lighting for visual detection; surveys began after a 2‐min settling period upon arrival. Observers recorded the species, location e.g., trunk, branch , and behavior e.g., perching, foraging of all individuals detected at a retention site, in addition to the times recorded to the nearest second at which the individual arrived and departed. Both observers made an effort to detect birds visually and aurally, though one observer was primarily responsible for recording while the other focused solely on viewing the retention; these duties were alternated. Surveys were conducted during adequate wind 12km/h and weather no rain or fog conditions Ralph et al., 1995 . We trained observers for 2 weeks prior to sampling with portable media players to improve their abilities in identifying bird species by sight and sound. 13    2.4. Modeling framework We used the hierarchical modeling framework described in Royle and Dorazio 2008 that has been developed and applied in other studies e.g., Dorazio et al., 2006; Kery and Royle, 2008; Dorazio et al., 2010; Ruiz‐Gutierrez et al., 2010; Zipkin et al., 2010 . The framework uses a state‐space formulation that separates the ecological process true occurrence for a given species, which is unobserved from the observation process detection of species presence ; observations or detections are modeled conditional on the latent process of occurrence Royle and Kery, 2007 . This separation allows for proper partitioning of uncertainty and examination of covariates that are distinctive to each process. The multispecies model estimates species‐specific occupancy and detection probabilities which can be combined to calculate community‐level metrics such as the species richness of a site while accounting for species that were never observed. A major benefit of the multispecies approach is that the species‐specific model parameters are all treated as random effects governed by a common community‐level distribution, resulting in a more parsimonious model with greater precision for parameter estimates, especially for rare species Royle and Dorazio, 2008 . The precision of parameter estimates is improved for any species with sparse detection data because the estimates for each individual species are informed by the data across all species Dorazio et al., 2010 . The approach is more efficient than estimating fixed parameters separately for each species in a large multispecies dataset, which would be limited to species having an adequate number of detections Royle and Dorazio, 2008 . To create the repeated measures design necessary for estimating occupancy and detection probabilities MacKenzie et al., 2002 , we divided each 30 minute bird survey 14    into three 10‐min intervals. Following MacKenzie and Royle 2005 , we interpreted occupancy as “use” given that the closure assumption was not valid – we assumed most bird species had territories that were larger than retention sites and, therefore, would be periodically unavailable for detection during surveys. We also assumed that detection was primarily a function of that availability i.e., the probability that a species is available for detection, given that it is present at a site and not a function of imperfect detection by the observers. We were confident that 2 observers could adequately detect all species that were present at a retention site during a survey, given the clear path of sight and small viewing window necessary to observe sites. To account for the potential lack of independence among observation intervals, we estimated detection probabilities for intervals with previous detections separately from intervals with no previous detections Mordecai et al., 2011 . We constructed a detection history for each species at retention site which indicated for each survey interval in year whether a species was detected 1 or undetected 0 . Our model notation is consistent with the state‐space formulation presented by Royle and Kery 2007 for modeling occupancy and detection probabilities. For a species to be detected during a survey interval, the species had to be periodically present at the retention site during a given year; therefore, the detections conditional on the unobserved occupancy state where such that were ~ Bernoulli is the probability that species is detected at retention site during interval in year , given that species is actually present at retention site in year . If species is not present at retention site in year then 0 and 15    0 with a probability of 1. Note that it is possible to observe a detection history where 0, 0, 0 even when 1 in situations where the species is never detected at the retention site during the survey that year even though it is actually present at some point. We modeled the unobserved occupancy state such that ~ Bernoulli where represents the probability that species is present at retention site in year t. We built logit‐linear models for the probabilities of detection and occupancy to incorporate the effects of covariates for each process. The logit transformation is necessary to allow predicted values in the linear models to range across all real numbers while constraining the back‐transformed probabilities to be between 0 and 1. We assumed that species‐specific detection and occupancy probabilities were also specific to an individual study area , and likely varied across years. In the model for detection, we included survey date as a potential covariate with both linear and quadratic terms to accommodate peaks of availability that may occur at any point during the season Royle and Dorazio, 2008 . We also included a binary variable to indicate whether a species had been detected in a previous survey interval Mordecai et al., 2011 . We defined the model of detection as: logit where 1 2 2 3 is the mean logit‐scale detection probability for species in study area during year ; 1 and 2 are the regression parameters for the linear and quadratic effects of survey date; and 3 is the regression parameter estimating the effect of a previous detection, . 16    We estimated occupancy separately for each year with a multi‐season model which had an initial occupancy probability in year 1 probability in year 2 1 followed by a subsequent occupancy 2 . Occupancy in year 2 was dependent on the unobserved occupancy state in year 1 1 and a function of either survival presence at a retention site in year 2 given presence in year 1 or colonization presence at a retention site in year 2 given absence in year 1 . The models of occupancy for each year incorporated the two site covariates of management interest, number of trees and edge distance, with both linear and quadratic terms. The number of trees was restricted to stems 25cm diameter given that this was the minimum diameter resulting in compliance with the state forest practices rules NCASI, 2008 . We were not interested in modeling occupancy dynamics per se i.e., effects of covariates on colonization and survival , thus, covariates were assumed to have the same additive effect on occupancy probability for both years. Similar to our detection model, we specified separate mean probabilities for each study area. We defined the occupancy model in year 1 as: logit where 1 1 2 2 2 4 3 is the mean logit‐scale occupancy probability for species in study area during year 1 at mean values for the covariates, and the terms are the regression parameters for the linear and quadratic effects of tree count 1 , 2 and edge distance 3 , 4 at retention site . Accordingly, the occupancy model for year 2 was: logit 2 1 1 1 1 17    2 2 where and are the mean logit‐scale probabilities of survival continued presence at a site and colonization new presence at a site , respectively, for species in study area at mean values for the covariates. Occupancy probability in year 2 is defined by survival when 1 1 and by colonization when 1 0. The regression parameters for tree count 1 , 2 and edge distance 3 , 4 are specified the same as in year 1. As part of the multi‐species hierarchical modeling framework, we treated the parameters for occupancy and detection models as random effects with each species‐ specific parameter being drawn from a common distribution specified by a mean and variance. The mean and variance quantified the community‐level responses across species, assuming that the species under consideration were ecologically similar see below and the heterogeneity across species could be described by a normal distribution Kery and Royle, 2008 . We specified 2 habitat groups of bird species, forest and open, with separate parameter distributions to account for potential differences in ecological responses to forest management. The forest group represented species that require mature trees to fulfill some life requisites e.g., cavity/canopy nesters, bark/canopy foragers; Martin, 1995 and could be considered “forest dependent”; the open group represented all other species which were either habitat generalists or were known to prefer relatively open vegetation types e.g., shrubs, young forest . As an example, we defined the parameter representing the linear effect of tree count on occupancy probability as: 1 where 1 ~ 1 , 1 is the mean response across forest species and 1 is the standard deviation. Each parameter in the models for occupancy and detection was 18    assigned its own hyperparameters i.e., and , and the species‐specific parameter estimate was drawn from the distribution specified by the hyperparameters. We calculated site‐level richness in each year, occupancy states across species such that ∑ , by summing the unobserved 1 for the 1, … , species that were observed. We also calculated the site‐level richness for each subset of species in the forest and open groups. We did not use data augmentation to incorporate unobserved species into the estimates of richness, as outlined in Dorazio et al. 2006 and illustrated in other applications of hierarchical multispecies models e.g., Dorazio et al., 2010; Zipkin et al., 2010 . Russell et al. 2009a argued that using “unseen, hypothetical species” as a basis for management recommendations to regulatory agencies would be problematic. Our estimates of species richness are still improved over the observed counts for individual sites by accounting for those species that were observed in the study area but never detected at a given sampling site. We examined relationships between site‐level richness and the covariates associated with occupancy edge distance and tree count while acknowledging that the model does not formally quantify these relationships see Zipkin et al., 2010 . Site‐level richness is a derived quantity determined by the collection of species‐ specific occupancy probabilities for a given retention site, with the unobserved occupancy state being contingent on covariate values. We present the relationships between richness and retention site covariates graphically. We used a Bayesian analysis to estimate the parameters and calculate community‐ level summaries with Markov chain Monte Carlo MCMC methods in WinBUGS Spiegelhalter et al., 2003 . WinBUGS uses Gibbs sampling Geman and Geman, 1984 to draw samples of all unknown quantities from the joint probability distributions specified 19    by the models. In this way, a posterior probability distribution for each parameter is estimated, from which summary statistics such as means and credible intervals the Bayesian analog to a confidence interval can be calculated. We operated WinBUGS with program R R Development Core Team 2011 through the R2WinBUGS package Sturtz et al., 2005 . We chose non‐informative prior distributions for all hyperparameters. The prior distribution for each of the mean hyperparameters was specified as a uniform distribution between 0 and 1 on the probability scale for model intercepts and a diffuse normal distribution with mean 0 and variance 1000 for regression coefficients. The prior distribution for each of the standard deviation hyperparameters was specified in terms of the precision i.e., inverse variance using a gamma distribution with shape 0.1 and scale 10. All covariate values were standardized to have a mean of 0 and a unit variance of 1. We examined model results based on 3 chains of 75,000 iterations after discarding the first 25,000 iterations and thinning by 50; this process resulted in 3,000 values forming the posterior distribution for each parameter. We determined model convergence by examining trace plots of the posterior distributions for each chain and by assessing the R‐hat statistic, or scale reduction factor, which should be 1.1 for all parameters Gelman et al., 2003 . WinBUGS code for model specification is presented in the Appendix A.1 . 3. Results We observed 47 and 45 bird species using retention sites in 2008 and 2009, respectively, with a total of 54 species across both years. Total detections were relatively low for every species, indicating that birds were rarely observed using retention sites 20    during the surveys Table 1.2 . Average species‐specific detection probabilities reflected the rarity of observations and, along with occupancy probabilities, were variable within and across study areas and years Figure 1.1 . Average species‐specific detection probabilities ranged from 0.01–0.07 in CWA, 0.01–0.07 in SWA, 0.04–0.57 in SOR, and 0.01–0.15 in NCA ; average species‐specific occupancy probabilities ranged from 0.05–0.63 in CWA, 0.02–0.98 in SWA, 0.02–0.33 in SOR, and 0.14–0.93 in NCA. There were no significant correlations between estimated occupancy and detection probabilities for any of the study areas Figure 1.2 , indicating that the probability of a bird being detected at a retention site i.e., being present in the part of its territory that overlaps a site was not significantly related to the probability of the bird ever using the retention site i.e., having a territory that overlaps a site . Our assessments of model fit indicated that convergence was achieved: R‐hat was 1.1 for all model parameters mean 1.01 and trace plots suggested adequate mixing. Covariates affecting occupancy and detection probabilities in the log‐linear models differed by species and habitat group in the significance of their regression coefficients Figures 1.3, 1.4 ; significance is indicated when the 95% credible interval CI does not overlap zero. Detection probabilities were shown to increase with survey date for 4 species including American Robin, Band‐tailed Pigeon, Cedar Waxwing, and Dark‐eyed Junco Figure 1.3a . A single species, Hairy Woodpecker, exhibited a significantly negative quadratic relationship between detection probability and survey date Figure 1.3b , suggesting that detection probability was higher during the middle of the survey season than at either start or end. Detection probability was found to increase significantly in a 21    given time interval after a detection had occurred in a previous time interval for 8 species Figure 1.3c . Occupancy probability for most species 38 of 54; 70% exhibited a significant positive linear response to tree count; habitat group had a small influence on the distribution or precision of parameters Figure 1.4a . The 16 species not exhibiting a significant response to tree count included 8 forest species Black‐capped and Chestnut‐ backed Chickadee, Downy and Hairy Woodpecker, Mountain and Western Bluebirds, Pileated Woodpeckers, Northern Flicker and Tree Swallow and 8 open/generalist species American Crow, Bewick’s Wren, Lazuli Bunting, Mourning Dove, Rufous Hummingbird, Spotted Towhee, and Townsend’s Solitaire . All of the forest species not showing a significant relationship with tree count were cavity‐nesters that likely used individual snags more often than green patches. Seven species exhibited a significantly positive quadratic response to tree count including 1 forest species Western Tanager and 5 open species American Robin, Chipping Sparrow, Stellar’s Jay, Swainson’s Thrush, and White‐ crowned Sparrow . Edge distance was not a significant predictor of occupancy probability for any species Figure 1.4c, d . Mean site‐level avian richness was estimated to range from ~1–32 for all species, ~1–12 for forest species, and ~1–22 for open species in both years Figures 1.5, 1.6 . Considering that occupancy probabilities for most species exhibited positive linear responses to tree count Figure 1.4a , as expected the estimates of site‐level richness also increased with tree count in both years for the entire bird community and each habitat group Figure 1.5 . The forms of the relationships between tree count and site‐level richness suggested power functions with a maximum richness at around 10–15 trees for all 22    species and open species, with some differences between study areas Figure 1.5 . Forest species appeared to increase with tree count but the trend was mostly linear without a clear asymptote. The absence of significant responses by species to edge distance Figure 1.4c resulted in no apparent relationship between mean site‐level richness and edge distance, aside from generally lower ranges of estimates at distances 300m Figure 1.6 . 4. Discussion The effectiveness of structure retention at meeting wildlife conservation objectives may be enhanced by using specific strategies that influence species use of the resources provided by the management actions. Our model results suggested that tree count had a positive effect on the predicted species richness of birds using retention sites in managed forests of the Pacific Northwest. The relationship between species richness and tree count was similar to a power function, which has often been used to describe species‐area relationships Connor and McCoy 1979 . There was little difference between the habitat groups Figures 1.4, 1.5 , indicating that both forest and open/generalist species responded similarly to attributes of structure retention, though forest species richness did not appear to reach a maximum. This may have resulted from many forest species, particularly cavity‐nesters, being more influenced by snags than live trees in the harvest units; retention sites with more green trees also tended to have more snags. One interpretation of the open species response is that larger patches of retained trees offer a greater amount and diversity of resources e.g., food, cover that are beneficial to numerous different species, regardless of specific habitat preferences. Our habitat grouping distinguished species based on certain life history characteristics Martin, 1995 23    but the simple dichotomy would not preclude some similarities in habitat use, especially for generalist species. Ruiz‐Gutierrez et al. 2010 cautioned on the use of specialist/generalist classifications using previous literature given that species responses do not always follow preconceived expectations. Our classification of forest and open/generalist species was used to allow the parameter distributions to differ for the species in each group; in the absence of a difference as dictated by the data , classification ends up having no impact on the model results. As expected, many species in the open group were ubiquitous and regularly observed throughout the harvest units, while forest species were almost exclusively observed at retention sites D.W. Linden, personal observation . Retention site use for open species may have required a short movement from the early successional habitat within the harvest unit, while for forest species it may have required movement from the adjacent mature forest habitat. In either case, the probability of movement to a retention site was likely influenced by the resources provided by the site, resulting in a similar response to tree count by both habitat groups. Our conclusions are generally consistent with previous research in western North America. Most studies of green‐tree retention effects on the avian community in the Pacific Northwest have focused on stand‐level responses of species richness and abundance to varying intensities of harvest, with some combination of unharvested control forests, total‐ harvest clearcuts, and partial‐harvest or variable‐retention stands serving as treatment units Beese and Bryant, 1999; Chambers et al., 1999; Preston and Harestad, 2007 . These studies have shown that moderate‐intensity harvests typically result in stands with a species composition more similar to unharvested forests than to clearcuts and can actually increase species richness as a response to greater habitat complexity, though forest‐ 24    dependent birds often exhibit a decreased abundance. Preston and Harestad 2007 suggested that forest‐dependent birds increased their use of retention with increasing patch size ranging from 0.25–2ha, though their design did not allow for formal inferences regarding patch size. In boreal forests of Alberta, large residual patches 100 trees supported bird communities that were more similar to old contiguous forests than small patches 10 trees immediately following stand‐replacing disturbances i.e., fire or harvest and those differences gradually decreased with time 15, 30 and 60 years post‐ disturbance as regenerating trees formed a closed canopy Schieck and Hobson, 2000 . These studies highlight the importance of green‐tree retention for providing habitat to members of the forest bird community. Edge distance was not a significant predictor of use for any observed species and, thus, appeared to have no influence on species richness at retention sites Figure 1.6 . Richness estimates were generally low at distances 300m, but the low sample size at those distances n 7 precludes robust conclusions. While it is possible that bird species responding negatively to edge distance were never observed using retention sites, the lack of an effect may also be attributed to the limited range of distances observed: distance to adjacent forest was 50m for 50% and 200m for 92% of observed retention sites. The distances we observed may have been too small to limit movement by most bird species, especially given that the edge contrast induced by age and structural differences i.e., regenerating vs. mature forest could be relatively benign Ries et al., 2004 . Additionally, the calculation of edge distance did not take into account the presence of additional nearby retention sites which may have diluted real effects. There are few examples in the literature to assist our understanding of territory‐scale movements by birds in search of 25    resources outside the preferred habitat type. Desrochers and Fortin 2000 observed Black‐capped Chickadee flocks regularly foraging along forest boundaries at distances up to 100m from the edge in a matrix of forest and agriculture. Fraser and Stutchbury 2004 found that an area‐sensitive forest bird was willing to move 1km across extensive areas of non‐forest in search of a mate, with no indication that movement was restricted to forest corridors. Given the high dispersal capability of birds, the distances to adjacent forest for retention sites observed in our study were likely too small to affect site use. The absence of large distances to adjacent forest are likely an artifact of existing forest practices rules which contain restrictions on the size and juxtaposition of harvest units. Our dynamic multi‐species hierarchical model enabled an estimation of species‐ specific habitat use across multiple years while accounting for imperfect detection. Our approach followed a modeling framework Dorazio and Royle, 2005 that has proven effective for estimating multi‐species responses to environmental changes resulting from management practices Russell et al., 2009a; Zipkin et al., 2010 and landscape‐scale forest fragmentation Zipkin et al., 2009; Ruiz‐Gutierrez et al., 2010 . The major benefit to this approach stems from the specification of species as a random effect governed by a community‐level distribution which allows rarely detected species to be incorporated into estimates of species richness and represents a parsimonious approach to modeling multi‐ species datasets Royle and Dorazio, 2008 . Our sampling methodology had several limitations in the context of understanding the benefits of structure retention to avifauna. We interpreted occupancy as use because retention sites were typically smaller than the expected territory sizes of the bird species observed in the harvest units and, thus, availability of birds with territories that overlap 26    retention sites determined our observed differences in detection. Use was defined as the detection of a species at a retention site, but the ecological values of different detections were not always equal: an individual perching for a short time on the branch of a retained green tree is qualitatively different than an individual foraging on a trunk or using a retained snag for a cavity nest. While we observed the full spectrum of use during the study, the low sample size of detections prevented these distinctions from being made for this analysis. The covariates included in our analysis, tree count for stems 25cm and edge distance, represented two attributes that were hypothesized to affect probability of use and were attributes directly linked to operational management decisions and the forest practices rules. However, tree count and edge distance may not have captured the range of biologically relevant characteristics that influence retention site use. We chose tree count over patch size because we felt that birds would perceive retention sites based on the number of trees serving as substrate for resources as opposed to “habitat islands” with different areas; the two measures tree count and area were also highly correlated. Tree count also represented a simplified description of attributes at a retention site. Snags provide a different set of resources than live trees and these differences can positively or negatively impact habitat use by a species depending on life history requirements Thomas et al., 1979; Bull et al., 1997 ; the same is true for snags with differences in decay state, tree species, or size Bull, 2002 . Unfortunately, the opportunistic nature of snag retention in harvest units made it difficult to sample a wide enough range of snag attributes to examine their effects on retention site use. Snag retention is one of the most difficult management 27    objectives for industrial forests given that snags have limited lifespans i.e., amount of time standing and that snag creation is naturally a highly stochastic process NCASI, 2008 . 5. Management implications Most bird species using retention sites in harvest units of the Pacific Northwest appear to respond positively to the number of trees retained and, therefore, forest managers can increase the utility of structure retention to avifauna by strategically grouping trees to maximize species richness. We contend a minimum group size for green tree retention of 10–15 trees with diameters 25cm is a valid management target for managed forests of the Pacific Northwest that will influence avian species diversity in harvest units. Retention sites with 10–15 large trees will typically range in area from approximately 0.01–0.02ha depending on the target planting density e.g., 900 trees/ha in Snoqualmie, WA; T. McBride, Hancock Forest Management, personal communication . Our observations suggest that single green trees are rarely used compared to larger patches and given the risk of blowdown, are not useful for meeting tree and snag retention objectives. Given that snag retention is often opportunistic, centering green‐tree retention groups around high‐quality snags Bull, 2002 could serve to guide the placement of retention sites. Regardless of the plans for green trees, retention should not be abandoned for single snags or unique wildlife trees that cannot be grouped for operational reasons; these resources should always be left in the harvest unit where possible. While the current study was not able to assess stand‐level differences in structure retention, it is understood that forest management prescriptions which are applied similarly across local scales i.e., stands can reduce habitat heterogeneity at the larger scale i.e., landscapes Lindenmayer 28    and Franklin, 2002 . Therefore, our management target should not be viewed as an unvarying prescription; rather, varying the group sizes of green‐tree retention across harvest units will help create habitat diversity at the landscape scale.   29    Table 1.1. Attributes of harvest units selected for observing bird use at structure retention sites in managed forests of the Pacific Northwest. Regiona CWA SWA SOR NCA n 22 20 20 23 Area ha Mean Min Max 8 37 40 9 4 24 22 4 12 57 61 11 Elevation m Mean Min Max 350 170 610 1500 200 20 300 1250 470 330 800 1830 Dominant tree speciesb WH, DF, RC, Ald WH, DF, RC, Ald, SS DF, WH WF, PP, DF, RF, IC a Regions are ordered by latitude, from north to south. Region codes represent the following areas in the Pacific Northwest: CWA central Washington; SWA southwest Washington; SOR southwest Oregon; NCA northern California. b Dominant tree species codes represent the following: Ald Red alder Alnus rubra ; DF Douglas‐fir Pseudotsuga menziesii ; IC California incense‐cedar Calocedrus decurrens ; PP ponderosa pine Pinus ponderosa ; RC western redcedar Thuja plicata ; RF red fir Abies magnifica ; SS Sitka spruce Picea sitchensis ; WF white fir Abies concolor ; WH western hemlock Tsuga heterophylla . 30    Table 1.2. List of species detected using retention sites from 2008–2009 in managed forests of the Pacific Northwest, including habitat group based on nesting and/or foraging requirements , common name, scientific name, American Ornithologist’s Union alpha code, and total number of detections. Group Common name scientific name FOREST Western Tanager Piranga ludoviciana Mountain Chickadee Poecile gambeli Northern Flicker Colaptes auratus Yellow‐rumped Warbler Dendroica coronata Red‐breasted Nuthatch Sitta canadensis White‐headed Woodpecker Picoides albolarvatus Black‐headed Grosbeak Pheucticus melanocephalus Hairy Woodpecker Picoides villosus Red‐breasted Sapsucker Sphyrapicus ruber Warbling Vireo Vireo gilvus Chestnut‐backed Chickadee Poecile rufescens Tree Swallow Tachycineta bicolor Mountain Bluebird Sialia currucoides Pileated Woodpecker Dryocopus pileatus Black‐throated Gray Warbler Dendroica nigrescens Hutton's Vireo Vireo huttoni Violet‐green Swallow Tachycineta thalassina Western Bluebird Sialia mexicana Downy Woodpecker Picoides pubescens Townsend's Warbler Dendroica townsendi Black‐capped Chickadee Poecile atricapillus OPEN Dark‐eyed Junco Junco hyemalis American Robin Turdus migratorius Chipping Sparrow Spizella passerina Pacific‐slope Flycatcher Empidonax difficilis Steller's Jay Cyanocitta stelleri Willow Flycatcher Empidonax traillii Band‐tailed Pigeon Columba fasciata Swainson's Thrush Catharus ustulatus Cedar Waxwing Bombycilla cedrorum Winter Wren Troglodytes troglodytes Dusky Flycatcher Empidonax oberholseri Olive‐sided Flycatcher Contopus cooperi 31    AOU code WETA MOCH NOFL YRWA RBNU WHWO BHGR HAWO RBSA WAVI CBCH TRES MOBL PIWO BTYW HUVI VGSW WEBL DOWO TOWA BCCH DEJU AMRO CHSP PSFL STJA WIFL BTPI SWTH CEDW WIWR DUFL OSFL detections 41 40 35 30 19 19 17 17 12 12 10 9 6 6 4 4 3 3 2 2 1 62 30 27 26 20 20 19 18 16 13 12 12 Table 1.2 cont’d Wilson's Warbler Wilsonia pusilla Song Sparrow Melospiza melodia White‐crowned Sparrow Zonotrichia leucophrys Rufous Hummingbird Selasphorus rufus Cassin's Finch Carpodacus cassinii MacGillivray's Warbler Oporornis tolmiei Spotted Towhee Pipilo maculatus American Goldfinch Carduelis tristis Evening Grosbeak Coccothraustes vespertinus Green‐tailed Towhee Pipilo chlorurus Orange‐crowned Warbler Vermivora celata Purple Finch Carpodacus purpureus Western Wood‐Pewee Contopus sordidulus Bewick's Wren Thryomanes bewickii Townsend's Solitaire Myadestes townsendi Common Raven Corvus corax Golden‐crowned Kinglet Regulus satrapa Lazuli Bunting Passerina amoena American Crow Corvus brachyrhynchos Mourning Dove Zenaida macroura Red Crossbill Loxia curvirostra 32    WIWA SOSP WCSP RUHU CAFI MGWA SPTO AMGO EVGR GTTO OCWA PUFI WEWP BEWR TOSO CORA GCKI LAZB AMCR MODO RECR 12 9 9 7 6 6 6 4 4 4 4 4 4 3 3 2 2 2 1 1 1 Figure 1.1. Map of study area locations within ecoregions and hydrologic units in the Pacific Northwest, USA. Region codes represent the following areas: CWA central Washington; SWA southwest Washington; SOR southwest Oregon; NCA northern California. 33    Figure 1.1 cont’d Ü SWA CWA Washington Oregon SOR NCA Ecoregions Hydrologic Units 0 75 150 300 km California 34    Figure 1.2. Mean probabilities of occupancy and detection for all species detected within the 4 regions during 2008 filled squares and 2009 open squares . Each species‐specific probability is defined by the mean of the posterior distribution for the intercept in the logit‐linear models. Region codes represent the following areas in the Pacific Northwest: CWA central Washington; SWA southwest Washington; SOR southwest Oregon; NCA northern California. 35    Figure 1.2 cont’d SOR 0.00 Detection probability 0.05 0.10 Detection probability 0.00 0.10 0.20 0.30 0.40 0.50 0.15 NCA 0.0 0.2 0.4 0.6 0.8 Occupancy probability 1.0 0.0 1.0 CWA 0.00 Detection probability 0.01 0.03 0.05 Detection probability 0.02 0.04 0.06 0.07 SWA 0.2 0.4 0.6 0.8 Occupancy probability 0.0 0.2 0.4 0.6 0.8 Occupancy probability 1.0 0.0 36    0.2 0.4 0.6 0.8 Occupancy probability 1.0 Figure 1.3. Posterior estimates mean with 95% credible interval of the regression coefficients from logit‐linear models of detection for each species. The coefficients correspond to the following covariates: a ordinal date, b squared ordinal date, and c previous detection. Species are ordered by increasing mean estimate of the coefficient and distinguished by habitat group into open/generalist species open squares and forest species filled squares . 37    1.0 Figure 1.3 cont’d b) α2 −1.0 Regression coefficient −0.5 0.0 0.5 Regression coefficient −1.0 −0.5 0.0 0.5 1.0 a) α1 0 10 20 30 40 Species index 50 20 30 40 Species index 0 50 −1 Regression coefficient 0 1 2 3 c) α3 0 10 38    10 20 30 40 Species index 50 Figure 1.4. Posterior estimates mean with 95% credible interval of the regression coefficients from logit‐linear models of occupancy for each species. The coefficients correspond to the following covariates: a tree count, b squared tree count, c edge distance, and d squared edge distance. Species are ordered by increasing mean estimate of the coefficient and distinguished by habitat group into open/generalist species open squares and forest species filled squares . 39      −1.5 −1.0 Regression coefficient −1.0 −0.5 0.0 0.5 Regression coefficient −0.5 0.0 0.5 1.0 0 0 10 c) β3 10 1.0 −3 Regression coefficient 1 2 3 Regression coefficient −2 −1 0 1 2 0 4 a) β1 3 Figure 1.4 cont’d 20 30 40 Species index 20 30 40 Species index 50 50 40  b) β2 0 0 10 d) β4 10 20 30 40 Species index 50 20 30 40 Species index 50 Figure 1.5. Mean estimated site richness for all species a, b , forest species c, d , and open/generalist species e, f at each retention site compared to tree count in 2008 left and 2009 right . Study areas are colored with central Washington CWA blue, southwest Washington SWA green, southwest Oregon SOR yellow, and northern California NCA red. For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of this dissertation. 41    0:10 Figure 1.5 cont’d 0:10 0:10 q qq qq qq qq qq 2008 q q q q q q q 0:10 q q q q q 2009 q q q q q q q q q q q q 5 10 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 0 q 0:10 30 q q q q q a 0 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 5 10 20 q q q 5 10 15 20 Tree count q 25 0 Richness (ALL) q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 20 30 q 30 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 0 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q b 5 10 15 20 Tree count 25 30 4 q 2 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 0:10 0 q 5 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q c 25 q q q 8 q q q 10 15 20 Tree count q 10 12 q q q 0 q q q q q q q q q q q q q q q q q q q 6 6 q q q q 4 8 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 2 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 30 0 5 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q d 0 q 10 12 Richness (FOREST) q 10 15 20 Tree count 25 30 q 10 5 q 0 q q 0:10 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 0 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 20 q q q q q q q q q q q q q q q q q q q q 15 q q q q q q 10 q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 15 q q q 5 q q q q q q q q q q e q 5 10 15 20 Tree count 25 0 q 20 Richness (OPEN) q 30 0 42    q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q 5 10 15 20 Tree count q q 25 q f 30 Figure 1.6. Mean estimated site richness for all species a, b , forest species c, d , and open/generalist species e, f at each retention site compared to edge distance m in 2008 left and 2009 right . Study areas are colored with central Washington CWA blue, southwest Washington SWA green, southwest Oregon SOR yellow, and northern California NCA red. 43    0:10 Figure 1.6 cont’d 0:10 0:10 q qq qq qq qq qq 2008 q q q q q q 0:10 q q q q q q q q q q 2009 q q q q q q q q 5 10 q q q q q q q q q q q qq qq q q q q q q qq q q q q q q q q qqqq q q q q q q q q q q qq qq q qq q q qqq q qq q q q qq q q q q q qq qq q q qq q q q q qq q q qqq q q q q qq q q qq q qq qq q qq q q qq qqq q q q qq q q qq q q qq q q q q q q qq q qq qq q q q q q qq q q q q q q q qq q qq q q q qq q q q q qq q qq q q q q q qq q q qq q q q qq q q q q qq qq qq q q q qqqq q q q qq qq q q qq qq q q qq q q q q q q q q q q q q qq q qq q q qq q qqq q q q q qq 30 q q q q q qq q q q q q q q q q q qq q q q q q q q q q q q q q q q q q 0 q q q a q 20 20 q q q q 5 10 Richness (ALL) q q q q q qq q q qq q qqq q q q qq q qq qq q q qqq q qq q q q q qq q q q qq q q q 0 30 q 0 q 0:10 q 100 200 300 Edge distance (m) 400 q q q q q q q qq qq q q q q q q qq q qq q q q q qq qq q q q q qq q q q qq q q qq q q q q q q q q q q q q q q qq q q q qq q q q q q qq q q q q q q q qqq q q q q q q q qq q q q q qq q qq q qq q q q q q q qq q q q qq q q qq q q q q qq q qq q qq q q q q q q qq qq q q q q q q qq q q q q qq q q q q qq q q qq q q q qqqq qq q q q q q qq q q q q qq q q qqqq q qq qq q qq q qq qq q q q q q q qq q q qq q q qq q q q q q q qq q qq q q q q qq q qqq q q q q q q qq q q qq q q qq q q q qq qq q qqq qq qq q q q q q q q q q q q q q q q q q q qq q q q q qq q qq q q qq q q q qqq q qqq q qq q q q q qq q q qq q q q q 0 b q q q 100 200 300 Edge distance (m) 400 4 q 2 q 10 12 8 q q q 0 q 6 6 q 4 8 q c 2 q q q q q q q q qq q q q q q q q q q q q q q q qq q q q q q qq q q q q qq q q qq q q qq q q q q q qq q q q qq qq q qqq qq q q qq q q q qq q q q q q q q q q q q q q qq q q q qq q q q q q q q q q q q q q q q q q q q q q q q q q q q q q q qqq q q qq qq q q q qq qq q q q q q q q q q qqq q qq q q qq q q q qq q q q q q qq qq q q q q qq q q q q qqqqqqqq qq q q qq q q q qqqq qq qqqqq q q qq q q q q q qq q q qqq qqq qq q q qq q qq q qqqq q qq q q qq q q q q q q q q q qq qq q q q q qq q q q q q q q q q q q q q q q q qq q q q qq qqq q q qq q q q q q q qq q q qq qq qq qq q qq q q q qq qq qq q q q 0:10 0 q q q q q q q q q q q qq q q q qq qq q q q q q q q q q q q qq q q qq q q q q q q q q qq q q q q q q q q q q q qq q q q q q qq q q q qq q q q q q q q qq q q q q q qq q q q q q q q q q q qq q q q q q q qq q q qq q q qq q q q q q q q q q q q qq q q q q q q q q q q q q q q q q q qq q q qq q q q q q q q qqqq q q qqq q q q q qq q q q q q qq q q q q qq q q qq qq q qq q qq q q q q q qq q q q q q q q qq qq q q qq q q q q q qq q qq q qq qq q q q qq q q q q qq q q q q q q q qq qq qq q q q q q q q q q q q q q q q q q q q q q q q q q q qq qq q q q q q qq q q q qq qqq q q qq q q q qq q q q qq q q q q q q qq q q q q q q qq q d q q q 0 q 10 12 Richness (FOREST) q q 100 200 300 Edge distance (m) 400 0 100 200 300 Edge distance (m) 400 q q 5 q 0 q q 0:10 q q 0 100 200 300 Edge distance (m) 20 15 10 10 q e 5 15 q qq q q q qq q q qq q qq q q qq q qq q q q qq qq q q q q q q q qq q q q qq q qq qq q qq q q q qq q q qq q q q qq q q qq q q q q q q q q qq q qq q q qqq q q q q q q q q qq q q q q q q q q q q q q q q q qq q q qq q qqqqq qqq q q qq q q q q qqq q q q q qq q qq qq q q q q q q q q q qq q q q q qq q q qq q q q q qq q q qq q q q qq q q q q q q q q q qq q q q q qq q qq q q qq qqqqq q qq q qq q q q q q q q q q qq q qq qqqq q qq qq q qq q qq q q q q q q q q q qq qq q q q q qq q q qq qqq qq q q q q q q q q q q q qq q q q q q q q qq q q q q q qq q q q qq q q q q q q qqq q qqq q qqq q q q qq q qq q q q q q q 0 q 20 Richness (OPEN) q 400 0 44    q q q q qq q q q q q qqq q q q q q qq q q qqq q q q q q qq qq q qq q q q q q q qq q q q q q q q q q q q q q q q q q q q q qq q q q q q q qq qq q q q q q q q q q qq q q q q q q q q q q q qq qq q q q q q q q q q qq q q qq q qq q q q q q q q q q q qq q q q qqq q qqqqq q qqq q q q q q q q qq qq q qq qq q q q q q q q q q qq q qq q q q q q q q q q q q q q qq q q q q q q qq qq q q qqq q q q q q q q q q qq q q qq q q q q q q q q q qq q q q q qq q q qq qq qq q q qq q qq q qq q q q qqq qq q qq q q q q q qq q q qq q q q q q q q q q q q qq q q q q q q qq q q q q qq qq q qqq q qqq q q q q qq q q q q q qq q q q qqq q q 100 200 300 Edge distance (m) f q q q 400 CHAPTER 2 MULTISTATE SITE OCCUPANCY ESTIMATION OF WHITE‐HEADED WOODPECKERS PICOIDES ALBOLARVATUS IN MANAGED FORESTS Abstract Structure retention is used to address biodiversity conservation objectives in managed forests by providing current and future snag resources for wildlife. The practice is considered critical to wildlife in intensively managed forests where timber production is the primary goal, but few studies have demonstrated a clear link between stand‐level attributes of structure retention and metrics of species habitat use, such as occupancy and reproduction. In this study, we used a Bayesian multistate site occupancy model to understand how structure retention influences the probability of occupancy and nesting by white‐headed woodpeckers Picoides albolarvatus on private industrial forests in California, USA during the 2010 breeding season. Our objective was to estimate state probabilities i.e., occupancy and nesting for white‐headed woodpeckers at the harvest‐ unit scale and identify relationships between each state probability and habitat covariates that can be directly impacted by forest management decisions, including snag density in the harvest unit and proportion of mature forest on the landscape. Mean occupancy probability was estimated to be 0.98 95% credible interval CRI : 0.89–1.00 and the probability of nesting given occupancy was 0.93 95% CRI: 0.75–0.99 , resulting in an unconditional nesting probability of 0.89 95% CRI: 0.72–0.97 . Snag density was positively associated with nesting probability, while neither covariate was good at predicting occupancy. The model indicated a negative effect of mature forest proportion on 45    nesting probability, though there was considerable uncertainty. Both detection of occupancy given nesting and detection of nesting increased with survey period, while detection of occupancy at sites without nesting decreased over time. The high occupancy and nesting probabilities in 2010 suggest that current structure retention policies have provided the necessary habitat conditions for white‐headed woodpecker nesting in harvest units of northern California. Harvest units with 2 snags/ha stem diameters 25.4cm had a probability of nesting that approached 1, suggesting a minimum threshold of retention that could serve as the biological basis for management targets. Keywords: Bayesian framework, California, managed forests, multistate occupancy, nesting probability, structure retention, white‐headed woodpecker 46    1. Introduction Structure retention is a management practice that is recognized as critical to biodiversity conservation in managed forests Bunnell et al., 1999 given that legacy trees can create landscape‐level structural diversity that might otherwise be lost Franklin, 1989 and provide current and future snag resources for the numerous wildlife species that depend on them Thomas et al., 1979; Neitro et al., 1985 . Wildlife species that require dead and dying trees to meet certain life history requirements are negatively impacted by forest management actions that remove these resources, such as traditional even‐aged timber harvesting Bunnell et al., 1999; Franklin et al., 2002 and salvage logging after natural disturbances Lindenmayer et al., 2004; Lindenmayer and Noss, 2006 . Cavity‐ nesting birds are especially vulnerable to silvicultural systems that do not provide the density and quality of snags necessary to maintain viable populations through time Newton, 1994 . Primary cavity‐nesters often have a disproportionately large influence on the diversity and abundance of bird species by acting as ecosystem engineers Jones et al., 1994 , and some woodpeckers have been hypothesized to play the role of keystone species in forest ecosystems Bednarz et al., 2004; Martin et al., 2004 . Forest management practices that address the habitat needs of woodpeckers have the potential to affect entire species assemblages Daily et al., 1993; Drever and Martin, 2010 . Examining the response by primary cavity‐nesters to differences in stand‐level snag resources may assist our understanding of the ecological benefit provided by structure retention in managed forests. White‐headed woodpeckers Picoides albolarvatus are a species that will likely benefit from forest practices that incorporate structure retention. This woodpecker is a 47    relatively weak primary excavator and an omnivore, requiring decayed snags for nesting and large, cone‐producing pines Pinus spp. for foraging Milne and Hejl, 1989; Dixon, 1995 . The species is historically associated with old pine forests containing large diameter trees and an open canopy Garrett et al., 1996 , though they have been found using managed forests with smaller diameter trees Kozma, 2011 . Pine seeds are considered a vital resource for overwinter survival of white‐headed woodpeckers Raphael and White, 1984 , but they may also serve as the primary food item at other times of year, including the breeding season Ligon, 1973; Dixon, 1995 ; forest types that support multiple species of cone‐producing pines contain the highest abundance of white‐headed woodpeckers Garrett et al., 1996 . While the species is considered to be relatively common in California, it is “poorly studied” compared to other woodpeckers in North America Garrett et al., 1996 and considered a sensitive and/or threatened species by three states Idaho, Oregon, Washington, USA and the U.S. Forest Service. Several recent studies have increased our knowledge of white‐headed woodpecker ecology Wightman et al., 2010; Hollenbeck et al., 2011; Kozma, 2011 and suggest that forest disturbance events e.g., wildfire which increase snag density and decrease live tree cover can improve habitat quality for the species. In a previous study examining tree retention effects on bird communities, we observed frequent nesting by white‐headed woodpeckers in clearcut harvest units with tree retention in northern California D.W. Linden, personal observation . Identifying the factors that contribute to probability of nesting by white‐ headed woodpeckers in managed forests will help to guide retention policies for areas where the species occurs. 48    Occupancy modeling provides a method for estimating species occurrence probabilities and habitat relationships while accounting for imperfect detection, which has been recognized as an important and pervasive consideration for ecological studies attempting to understand species distributions MacKenzie et al., 2002; MacKenzie et al., 2006 . Failure to account for imperfect detection can bias both the estimates of species occurrence probabilities and the parameters of model covariates, such as habitat attributes Gu and Swihart, 2004 . While estimates of occupancy probability can improve understanding of species presence, presence alone may not be a particularly useful metric for evaluating conservation and management strategies that are aimed at impacting species persistence. A multistate occupancy model extends the basic scenario of presence/absence to situations where there is more than one category of species occupancy, such as the probability of reproduction given presence, and there is uncertainty about the true state of the species at a given site Nichols et al., 2007; MacKenzie et al., 2009; Martin et al., 2009 . The model allows detection probabilities to vary according to the estimated true state, thus, accounting for differences in detection that might occur when certain states result in a species being more or less detectable e.g., adult birds actively feeding young . More importantly, environmental covariates that influence the state probabilities can be evaluated in the model. We used a multistate occupancy model to estimate probabilities of occurrence and reproduction of white‐headed woodpeckers within an intensively managed forest landscape in northern California, USA. The multistate model allowed us to examine how forest management practices that impact stand‐ and landscape‐level habitat attributes might influence state probabilities, while accounting for differences in detection 49    probability that might result from changes in nesting behavior during the breeding season Russell et al., 2009b . Our objective was to estimate occupancy and nesting probabilities of white‐headed woodpeckers at the harvest‐unit scale and identify relationships between each state probability and habitat covariates that can be directly impacted by forest management decisions, including snag density in the harvest unit and proportion of mature forest on the landscape. We hypothesized that white‐headed woodpeckers would have higher occupancy and nesting probabilities in harvest units that retained greater densities of snags having a diameter at breast height dbh 25.4cm due to increased availability of nest sites Milne and Hejl, 1989; Kozma, 2009; Wightman et al., 2010 . In addition, we hypothesized that both state probabilities would be positively associated with the proportion of mature forest with large diameter trees e.g., 50cm surrounding the harvest units 1‐km radius that could be used as foraging habitat for white‐headed woodpeckers during fall and winter, but may also serve as an important resource during the breeding season Dixon, 1995 . 2. Methods 2.1. Study area We conducted the study in northern California, east of Mt. Shasta, on private industrial timberlands surrounded by the Shasta‐Trinity National Forest centered around 41°21ˊN, 121°50ˊW Figure 2.1 . The area was covered by second‐ and third‐growth, mixed‐conifer forest that was historically dominated by ponderosa pine Pinus ponderosa and is now mostly white fir Abies concolor . Selective logging and fire suppression in the last century have altered the overstory composition similar to other dry forests of western 50    North America Hessburg et al., 2005 . Additional overstory species in decreasing order of prevalence included Douglas‐fir Pseudotsuga menziesii , red fir Abies magnifica , California incense‐cedar Calocedrus decurrens and sugar pine Pinus lambertiana , and non‐forested areas were typically chaparral vegetation composed primarily of manzanita Arctostaphylos spp. . True fir Abies spp. dominance generally increased with elevation, and elevations ranged from 1250–1950m. The study area was characterized by a mosaic of stand size classes: ~41% was comprised of mature stands with a quadratic mean diameter QMD; diameter of tree with average basal area 60cm, ~36% contained younger stands with a QMD ranging 25–60cm, and the remaining ~20% of stands were evenly split among past harvests QMD 12–25cm , relatively recent harvests QMD 3–12cm and new harvests planted with seedlings U.S. Forest Service, 2007 . We selected all harvest units that were clearcut with retention between 3–6 yrs prior to 2010 in the study area n 65 . Recent harvests 1 yr old were not selected to avoid potential disturbance effects from harvesting and replanting activities. Harvest units had a mean area 9ha range 4–15ha and typically contained 1–6 patches ~0.1ha of green‐tree retention that were randomly dispersed throughout the unit, in addition to varying densities of snags that were retained during harvest including stumps 1m tall . 2.2. Woodpecker surveys We used repeated, independent surveys across multiple visits to each harvest unit from June–July 2010 to estimate state and detection probabilities for white‐headed woodpeckers. During each survey, a single observer walked to the center of the harvest unit and after a 2‐min settling period, recorded all detections of woodpeckers in the harvest unit over a 10‐min observation period. Sex and behavior of detected individuals 51    were recorded when possible and detections that occurred before or after the 10‐min period typically while entering/exiting the harvest unit were recorded separately. We used passive as opposed to broadcast surveys Dudley and Saab, 2003 to prevent individuals in nearby harvest units from being enticed into entering the focal harvest unit. At the conclusion of the 10‐min observation, areas of the harvest unit where woodpecker behavior suggested the presence of a nest e.g., individuals viewed carrying food or entering a snag cavity were searched and carefully observed for up to 30min or until a nest was located. All surveys were conducted between sunrise and 5 hours post under optimal weather conditions e.g., no rain or heavy wind . Intervals between surveys were approximately 7–10 days with a maximum of 4 visits to each harvest unit. We used a removal design Farnsworth et al., 2002 to construct a detection history see Modeling framework for each harvest unit, where visits to a given site were terminated once evidence of reproduction was found. This design recognizes that subsequent detection of a state may lack independence from previous detections under certain circumstances, as in this case, due to observer experience of finding a nest. This heterogeneity in detection probability could be modeled explicitly but would not increase the precision of estimates for the state and detection probabilities, nor provide any additional useful information MacKenzie and Royle, 2005; Saracco et al., 2011 . 2.2. Habitat covariates We used a combination of field‐collected and remotely‐sensed data to describe the habitat covariates we hypothesized would influence the state probabilities of occupancy and nesting. Snag density was measured at each harvest unit by a complete tally of all dead wood 12.7cm dbh and 1m tall. Previous experience with white‐headed woodpeckers at 52    these sites suggested tall stumps were a frequently used substrate for cavity nests and, thus, were tallied with standing snags. Snags 5m outside the harvest unit boundary were also included. We used a Geographic Information System GIS to determine boundaries of harvest units by examining shapefiles outlining Timber Harvest Plans THPs as provided by the California Department of Forestry ftp://ftp.fire.ca.gov/forest/ and 2009 aerial photography from the National Agricultural Imagery Program NAIP; http://datagateway.nrcs.usda.gov/ . Snag tallies for stems 25.4cm diameter were divided by the estimated area of the harvest unit to derive the snag density; the diameter cutoff represented the lower limit for white‐headed woodpecker cavity nests observed in the field and in previous studies Milne and Hejl, 1989; Kozma, 2009 . We calculated the proportion of mature forest within a 1‐km radius from the center of each harvest unit within a GIS to estimate the availability of large trees for foraging. Mature forest was designated by areas containing trees with a mean dbh 60cm, as determined by the California Wildlife Habitat Relationships CWHR classification provided in the CALVEG geodatabase U.S. Forest Service, 2007 . The 1‐km radius represented the estimated home range of a white‐headed woodpecker Garrett et al., 1996; Hollenbeck et al., 2011 . We considered using forest patches that were pine dominant according to the CALVEG geodatabase, but these forest types were relatively rare at the higher elevations typical of our study area where true firs were the most numerous tree species. We acknowledge that our classification of mature forest did not guarantee the presence of large diameter pines, though they were frequently observed in the forests surrounding our harvest units including a 250cm dbh sugar pine; D. W. Linden, personal observation . 53    2.4. Modeling framework We developed a multistate model using the hierarchical modeling framework presented by Royle and Dorazio 2008 and following the notation of Nichols et al. 2007 for estimating occupancy with multiple states. Here we considered each harvest unit to represent one of 1, … , 65 sites each with 1, … , 4 surveys. Each of the 4 survey intervals spanned 10 days, during which a site could be visited no more than once depending on whether a previous visit had detected a nest. The observed state at site could be categorized into 1 of 3 states 1 ; and occupied with nesting : unoccupied 0 ; occupied without nesting 2 . We modeled detections conditional on the latent true state such that | ~ , where latent true state at site , and is the observed state at site during survey , is the is the probability of detecting state during survey . We used the notation of Nichols et al. 2007 which reparameterizes the detection probabilities such that: 1 the probability of detecting occupancy given that the site is occupied without nesting Pr 1| 1 ; 2 the probability of detecting occupancy given that the site is occupied with nesting Pr 1| 2 ; and the probability of detecting a nest given that the site is occupied with nesting Pr 2| 2 . Detection histories for each site, , consisted of 4 surveys regardless of whether the site was actually visited 4 times; if a site was not visited during survey interval then NA. We modeled the latent true occupancy state such that ~ , where the probability of occupancy state at site . Once again, we followed the Nichols et al. 54    is 2007 notation that reparameterizes a more general multinomial occupancy model MacKenzie et al., 2009 where the state probabilities are defined as follows: 1 1 or 2 Pr probability that site is occupied probability that site has a nest Pr 2 / Pr as 1 ∗ 2 1 1 Pr 1 Pr 2 ; and 2 2 , given that the site is occupied the the 1 or 2 2 . The unconditional nesting probability is specified 2 . This parameterization allows for the examination of covariates affecting each probability separately Nichols et al., 2007 , which is more biologically interesting than examining relationships with the unconditional probability of occupancy 1 MacKenzie et al., 2009 . without nesting, Pr We examined the effects of covariates on state and detection probabilities using logit‐linear models. Detection probabilities were modeled as follows: logit 1 logit 2 logit 10 11 20 21 30 31 where 10 , 20 , and 30 are the logit‐transformed detection probabilities; and 11 , 21 , and 31 are the regression coefficients for the variable survey number which simply indexes the 1, 2, 3, or 4 and allows for the detection probabilities to increase, decrease, or remain constant over the course of the entire study. We hypothesized that the detection probabilities 2 and would increase over the survey intervals as woodpeckers became more conspicuous during nesting activities that required frequent movement e.g., returning to cavity nest to feed young . 55    Our logit‐linear models for the occupancy state probabilities were as follows: logit 1 10 11 12 logit 2 20 21 22 where 10 and 20 are the logit‐transformed probabilities of occupancy and nesting given occupancy, respectively; 11 and 21 are regression coefficients for the variable , representing the snag density at site ; and 21 and 22 are regression coefficients for the variable , which describes the proportion of mature forest in a 1‐km radius around site . We considered quadratic terms for but this relationship was not supported by the data for either state probability and thus, we removed it for simplicity. All covariates were standardized to have a mean of 0 and unit variance of 1 to facilitate comparison of effect sizes between the selected covariates. We incorporated spatial random effects into our logit‐linear models for the occupancy state probabilities, given that our distribution of sites and average distance between sites was small relative to the dispersal capability and home‐range size of white‐ headed woodpeckers, such that we might expect some spatial autocorrelation in the state probabilities. The spatial random effects could potentially account for unmeasured environmental covariates or spatial population processes e.g., source‐sink dynamics that affected the species distribution in our study area Chelgren et al., 2011 . Therefore, the logit‐linear models for 1 and 2 included additional terms, 13 and 23 , representing zero‐mean random variables that were spatial correlated with distance according to an exponential function Diggle et al., 1998, as referenced in Chelgren et al., 2011 . These spatial random effects were each assigned multivariate normal distributions, such that 56    2 13 ~ MVN 0, 1 Σ and 23 ~ MVN 0, 2 2 2 2 Σ ; the variance terms 1 and 2 serve as the variances of the random effects, while Σ is the between‐site correlation matrix. The elements of the exp correlation matrix were calculated by an exponential function, 1, 2 where controls the rate of decay in correlation between any 2 sites separated by distance 1, 2 ; larger values of result in a more rapid decay. It should be noted that we chose the same correlation matrix for each of the spatial random effects, but allowed the variances to differ. We implemented the multistate model in WinBUGS Spiegelhalter et al., 2003 , using the R2WinBUGS package Sturtz et al., 2005 from R R Development Core Team 2011 . WinBUGS is a software program designed for Bayesian model inference using Markov chain Monte Carlo MCMC methods with Gibbs sampling Geman and Geman, 1984 to estimate model parameters by sampling from the joint probability distribution. Posterior distributions for all model parameters are estimated, from which means and credible intervals CRIs – the Bayesian equivalent to a confidence interval can be calculated. The model specification easily facilitates the incorporation of spatial random effects. We used non‐informative priors for most model parameters in the logit‐linear models, including a uniform distribution between 0 and 1 on the probability scale for intercepts 10 , 20 , 30 , 10 , 20 and a normal distribution with mean 0 and variance 100 for covariates of detection 11 , 21 , 31 and occupancy state 11 , 21 , 12 , 22 . Similar to Chelgren et al. 2011 , we found the terms for the spatial random effects to be particularly sensitive and were unable to use typical uninformative 2 2 priors on those parameters. The priors we used for 1 and 2 were hierarchical half‐ 57    normal mean 0, variance 2 , df 1 with a uniform U 0, 1 prior on ; this specification recognizes that the spatial random effects are likely small with low variance Gelman, 2006 . Our prior for was uniform U 1,30 which dictated a reasonable range in magnitude of correlation with increasing distance between sites; for example, the range in potential correlation for sites separated by 0.5km was 0.28–0.96, while for sites separated by 10km it was 0–0.42 for context, see Figure 2.1 . We examined posterior distributions for model parameters by running 3 chains of 100,000 iterations, discarding the first 50,000 iterations, and thinning by 25; this resulted in 6,000 values forming the posterior distribution for each model parameter. We assessed model convergence by examining trace plots and ensured that the scale reduction factor, or R‐hat statistic, was 1.1 and the effective sample size was 100 for all model parameters Gelman and Hill, 2007 . We assessed the significance of regression coefficients based on whether the estimated credible intervals overlapped zero; strong support for a covariate effect was indicated when the 95% CRI did not overlap zero, while weak support was indicated when the 90% CRI did not overlap zero. WinBUGS code for model specification is presented in the Appendix A.2 . 3. Results We detected white‐headed woodpeckers at 58 of the 65 harvest units and found cavity nests at 42 harvest units, resulting in a naïve estimate of 0.89 95% CRI: 0.79–0.95 for occupancy probability and 0.72 95% CRI: 0.60–0.82 for nesting probability, given occupancy. We detected nests across all 4 survey intervals, but only 5 harvest units were discovered to have nesting in the last survey. Snag densities for trees 25.4cm dbh 58    averaged 3.35 snags/ha standard deviation SD : 2.55, range: 0.24–12.53 within harvest units while the proportion of mature forest surrounding harvest units 1km‐radius averaged 0.45 SD: 0.19, range: 0–0.87 . A simple logistic regression comparing harvest units with detected nests to those without revealed no significant effects of snag density or proportion of mature forest on naïve nesting probability. The hierarchical multistate model which accounted for imperfect detection suggested higher estimates for both occupancy and nesting probabilities, with 1 95% CRI: 0.89–1.00 and 2 0.98 0.93 95% CRI: 0.75–0.99 for sites with average values of the covariates. For the unconditional probability of nesting in the study area, 1 ∗ 2 0.89 95% CRI: 0.72–0.97 . Model results suggested only one covariate with strong support 95% CRI not overlapping zero for a relationship with state probability – snag density, which appeared to increase the conditional nesting probability Table 2.1 . Examination of the change in predicted nesting probability with increasing snag density indicates that the probability of a white‐headed woodpecker nesting in a harvest unit approaches 1 at densities 2 snags/ha Figure 2.2 . The model also suggested weak support 90% CRI not overlapping zero for a relationship between conditional nesting probability and proportion of mature forest; conditional probability of nesting appeared to decrease with an increasing proportion of mature forest, though uncertainty in the predicted probability is evident at high proportions Figure 2.3 . Detection probabilities were variable depending on the estimated true state and the survey interval Table 2.1; Figure 2.4 . While none of the linear trends with survey interval had strong support, all had weak support from the data. The probability of detecting 59    occupancy by white‐headed woodpeckers was considerably lower for sites that did not contain a nest than for sites that did Figure 2.4 . In addition, the detection probability decreased over time for sites without nesting and increased for sites with nesting; 1 ranged from 0.46 95% CRI: 0.23–0.71 in survey 1 to 0.17 95% CRI: 0.01–0.50 in survey 4, while 2 ranged from 0.79 95% CRI: 0.68–0.88 in survey 1 to 0.95 95% CRI: 0.82– 0.99 in survey 4 Figure 2.4 . The probability of detecting a nest also increased with time and ranged from 0.42 95% CRI: 0.30–0.55 in survey 1 to 0.66 95% CRI: 0.48–0.85 in survey 4. 4. Discussion Structure retention in harvest units is considered to be a critical component of biodiversity conservation in managed forests Bunnell and Huggard, 1999 , yet little information exists regarding the response by wildlife species to specific practices within intensively managed landscapes where timber production is the primary goal. Our results suggest that white‐headed woodpeckers had a high probability of occupancy across young 3–6 yr old harvest units that employed structure retention and that the density of retained snags was a significant predictor of woodpecker nesting probability in industrial forests of northern California. Our study illustrates the tangible benefits of structure retention to the population persistence of a primary cavity‐nester and suggests biologically‐based targets for retention policies that can impact avian communities. We found evidence for our hypothesis that nest‐site availability might be limiting in harvest units and those units with higher densities of retained snags would be more likely to contain a white‐headed woodpecker nest. Other studies of white‐headed woodpeckers 60    have suggested that decay condition may be more important than density of snags Raphael and White, 1984; Wightman et al., 2010 , given that not all snags are considered available for nest sites when the condition of the wood is unsuitable for cavity excavation Bagne et al., 2008 . Snag decay condition increases with the time‐since‐death of the tree but also interacts with cause of death e.g., fire vs. insect and tree species Morrison and Raphael, 1993 . We did not include snag decay condition or species in our model due to the lack of variation across harvest units in our study: for snags 25.4cm dbh, 97% were in moderate to late stages of decay and 86% were true fir Abies spp. . Thus, we felt snag density was an appropriate proxy for nest‐site availability and chose the 25.4cm dbh threshold to reflect white‐headed woodpecker nest‐site preferences demonstrated by previous studies Milne and Hejl, 1989 and our own observations 95% of nest trees 25.4cm dbh n 89 , D. W. Linden, unpublished data . The model indicated weak support for a negative relationship between mature forest proportion within a 1‐km radius and white‐headed woodpecker nesting probability. This result was counter to our hypothesis that mature forest, defined by stands with an average diameter 60cm, may provide foraging habitat for white‐headed woodpeckers and increase the probability that a harvest unit would be occupied and contain a nest. Previous studies have cited the importance of large‐diameter pine forests as foraging habitat for white‐headed woodpeckers due to their reliance on pine seeds during the winter and throughout the year Dixon, 1995; Garrett et al., 1996; Kozma, 2011 . The study area supported multiple pine species e.g., ponderosa and sugar , yet the forest composition was dominated by firs and the remotely‐sensed data could not predict the location of individual mature pine trees; thus, our definition of mature forest may have been over‐ 61    simplified. Hollenbeck et al. 2011 found that QMD was a predictor of white‐headed woodpecker habitat in Oregon but acknowledged both the limitation of remotely‐sensed data for estimating QMD and the possibility that QMD did not properly quantify the presence of large‐diameter pines. At the stand level, canopy cover within mature forests in our study area was consistently high 60% and would not qualify as the “open canopy” that white‐headed woodpeckers appear to prefer Garrett et al., 1996 . At the landscape scale, a uniform pattern of harvest units resulting from state forest practices rules, which restrict the adjacency of timber harvests California Department of Forestry and Fire Protection, 2007 , may have created a suitable juxtaposition of open and closed forests Hollenbeck et al., 2011 and provided the necessary conditions for white‐headed woodpecker nesting. This would explain why the proportion of mature forest decreased nesting probability, as higher proportions effectively eliminate that juxtaposition. We used a hierarchical multistate occupancy model Nichols et al., 2007; MacKenzie et al., 2009 to estimate the probability of multiple species states, specifically occupancy and nesting, while accounting for imperfect detection and state uncertainty. The multistate model estimated detection probability both for the presence of individuals and the presence of cavity nests, allowing us to distinguish differences between detection processes that were dependent on the true state of the species at the site. The detection probabilities of occupancy and nesting for sites with a nest suggested that woodpeckers were more likely to be detected as the breeding season progressed, which may have been a function of changes in behavior due to the activities associated with rearing young. Given that detection probabilities were 1, a failure to incorporate detection would have lead to biased estimates of the state probabilities. A simple logistic regression suggested that the 62    naïve estimate of overall nesting probability of white‐headed woodpeckers in young harvest units was 0.65, much lower than the probability estimate from the multistate model 1∗2 0.89 . The differences in occupancy probability between the simple model 0.89 and multistate model 1 0.98 were smaller, suggesting that state uncertainty was responsible for more bias than imperfect detection. The simple model also found that snag density was not a good predictor of nesting which opposed the significant relationship estimated by the multistate model and further illustrates the bias that can occur with regression coefficients when imperfect detection is ignored in presence/absence models Gu and Swihart, 2004 . Our estimates of white‐headed woodpecker occupancy and nesting probability are restricted to young harvest units 3–6 yrs old during the breeding season in 2010, which limits some of our inference. We did not survey recent harvest units 1 yr old due to concerns over the impact of the recent disturbance on nest site selection, though we acknowledge that woodpeckers may have been using the new cuts. Anecdotally, we observed a successful white‐headed woodpecker nest in one unit that had been harvested between the 2009 and 2010 breeding seasons which indicates non‐zero occupancy and nesting probabilities in new harvest units. We also did not survey older harvest units 10– 15 yrs old which prevented us from examining how occupancy changes over time, though we expect detection probability would have been severely decreased in older cuts given that regeneration in these units was often 2–3m in height and our survey methods relied on a clear view of the entire harvest unit. Based on previous observations in the study area, we suspect that decreased snag densities, due to longer decay periods, and increased 63    cover for nest predators likely decreased the probability of white‐headed woodpecker nesting in older harvest units. 5. Management implications Our study supports the ecological benefit of structure retention in managed forests, as illustrated by white‐headed woodpecker nesting in harvest units of northern California. Results from the multistate occupancy model suggest that nesting probability was significantly increased by the retained snag density in harvest units, with probabilities approaching 1 at 2 snags/ha stems 25.4cm dbh . State forest practices rules in California require the retention of all non‐merchantable safe snags during harvest California Department of Forestry and Fire Protection, 2007 . This policy appears to be providing the habitat conditions necessary for supporting white‐headed woodpeckers in dry conifer forests of northern California, given the high estimates of occupancy 0.98 and overall nesting probability 0.89 . Where safety concerns or site conditions prevent the retention of natural snags, high‐cut stumps 2m in height may provide an easy and effective alternative given the propensity by white‐headed woodpeckers to create cavity nests relatively close to the ground 3m; Milne and Hejl, 1989 , though these created snags may take 3–5 years or longer before they are suitable for cavity excavation Arnett et al., 2010 . Additional research should identify links between white‐headed woodpeckers and secondary cavity‐nesters to understand whether management practices that support these woodpeckers can also support other members of the avian community. 64    Table 2.1. Posterior summaries for model parameters means, standard deviations SD , and 95% credible intervals lower 0.025, upper 0.975 from the multistate model estimating white‐headed woodpecker occupancy and nesting probabilities during 2010 among harvest units in California, USA. Parameters include intercepts and regression coefficients of standardized variables from the logit‐linear models for probabilities of occupancy 1 , nesting given occupancy 2 , detection of occupancy given the site is occupied without nesting 1 , detection of occupancy given the site is occupied with nesting 2 , and detection of nesting given the site is occupied with nesting 65    . Table 2.1 cont’d Parameter Occupancy 10 11 12 Mean SD 0.025 0.975 3.942 0.678 1.003 1.261 1.065 0.904 2.110 –1.680 –0.890 6.944 2.605 2.719 2.641 3.088 –1.366 0.940 1.151 0.797 1.116 1.116 –3.063 4.790 5.714 0.099 0.394 0.751 0.904 0.653 –1.306 –0.581 2.295 2.024 –0.558 0.584 0.543 0.403 –1.964 –0.103 0.179 1.474 –0.675 0.354 0.518 0.255 –1.741 –0.072 0.304 0.885 1 effect effect Nesting | occupancy 2 20 21 effect 22 effect Detection of occupancy | occupancy without nesting 1 10 11 effect Detection of occupancy | occupancy with nesting 2 20 21 effect Detection of nesting | occupancy with nesting 30 31 effect 66    Figure 2.1. Harvest units selected to survey white‐headed woodpecker occurrence and reproduction during 2010 near Mt. Shasta, California, USA. Harvest units Shasta-Trinity NF Private land Roads Oregon California 0 2 4 67    8 km Ü Figure 2.2. Model predicted relationship between conditional nesting probability, 2 , of white‐headed woodpeckers and harvest unit snag density snags 25.4cm diameter/ha on industrial timberlands in northern California. Mean probability with 95% credible 0.8 0.6 0.4 0.2 0.0 Probability of Nesting | Occupancy 1.0 intervals are presented. 0 1 2 3 Snag density (snags/ha) 68    4 Figure 2.3. Model predicted relationship between conditional nesting probability, 2 , of white‐headed woodpeckers and proportion of mature forest quadratic mean diameter 60cm within a 1‐km radius on industrial timberlands in northern California. Mean 0.8 0.6 0.4 0.2 0.0 Probability of Nesting | Occupancy 1.0 probability with 95% credible intervals are presented. 0.0 0.2 0.4 0.6 Mature forest proportion 69    0.8 1.0 Figure 2.4. Model predicted mean probabilities of detection with 95% credible intervals over the 4 survey intervals. The probabilities are detection of occupancy given the site is occupied without nesting 1 , open triangles , detection of occupancy given the site is occupied with nesting 2 , filled squares , and detection of nesting given the site is , open squares . 0.6 0.4 0.0 0.2 Detection probability 0.8 1.0 occupied with nesting 1 2 3 p j1 4 1 2 p j2 70    3 4 1 2 3 δj 4 CHAPTER 3 FACTORS AFFECTING WHITE‐HEADED WOODPECKER PICOIDES ALBOLARVATUS REPRODUCTIVE SUCCESS IN MANAGED FORESTS WITH STRUCTURE RETENTION Abstract The importance of woodpeckers to forest bird communities and their sensitivity to forest management practices make them a suitable target for addressing conservation objectives in managed forests. Understanding the impact of management activities on woodpecker populations can be improved by examining how metrics of reproductive success relate to altered habitat conditions. We estimated nest survival and number of young fledged for white‐headed woodpeckers Picoides albolarvatus located on private timberlands in northern California, USA from 2009–2010. We documented a high nest success rate 0.85, 95% credible interval: 0.59–0.96 and found that successful nests were most likely to produce 3 young. None of the habitat variables that were examined at multiple scales were able to describe variation in either metric of reproductive success. There was weak support for a date effect on number of young fledged, with more young fledged from nests that finished during the middle of the season. The high fecundity exhibited by woodpeckers in our study area suggests that the population may serve as a source. Current forest management practices appear to be providing habitat conditions that are conducive to the species’ persistence in this region. Keywords: Bayesian framework, California, fecundity, managed forests, nest survival, reproductive success, structure retention, white‐headed woodpecker 71    1. Introduction Woodpeckers are often regarded as indicators of forest ecosystem health and integrity due to their reliance on habitat features e.g., large dead trees, or snags most characteristic of old, unmanaged forests Angelstam and Mikusinski, 1994; Drever et al., 2008 . In addition, the role of woodpeckers as ecosystem engineers and keystone species can influence the biodiversity of forest wildlife communities Daily et al., 1993; Jones et al., 1994; Mikusinski et al., 2001; Martin et al., 2004; Virkkala, 2006 . Woodpeckers require snags as substrate for foraging and nesting. They excavate cavities for rearing young and rarely reuse previous cavities, creating a valuable resource for secondary cavity‐nesters Martin and Eadie, 1999 . In North America, secondary cavity‐nesters rely almost exclusively on woodpeckers as their source for tree cavities Cockle et al., 2011 . Forest management activities that remove large trees and snags are detrimental to many forest vertebrates, and cavity‐nesting species are especially vulnerable due to nest‐site limitation Newton, 1994 . Setting habitat targets that fulfill the resource requirements for woodpeckers has the potential to be a useful forest management strategy to address biodiversity conservation objectives Drever and Martin, 2010 . The white‐headed woodpecker Picoides albolarvatus is an omnivorous primary excavator associated with pine forests characterized by large trees and open canopies in western North America Garrett et al., 1996 . This woodpecker requires moderately decayed snags for nesting substrate and large mature pine trees for foraging; it relies on pine seeds as a primary food source during the winter and possibly throughout the year Ligon, 1973; Dixon, 1995 . Recent studies of the white‐headed woodpecker have emphasized habitat attributes affecting the nesting ecology of the species which may be 72    modified by forest management practices that include salvage logging in burned forests Wightman et al., 2010 , fuel reduction in dry forests Hollenbeck et al., 2011 , and methods to improve timber production Kozma, 2009, 2011 . These studies estimated nest survival as a function of biotic and abiotic factors, including habitat variables such as density of large snags and live trees. Additional information included habitat suitability models to predict nest locations Wightman et al., 2010; Hollenbeck et al., 2011 and detailed nest demographic data e.g., average clutch size, number of young fledged Kozma, 2009 . This research has contributed much to our knowledge of white‐ headed woodpecker ecology, particularly for the regions i.e., Oregon and Washington, USA where the species is considered threatened by state and federal natural resource agencies due to population declines that have likely resulted from reductions in the amount of old‐growth pine forests Kozma, 2009 . Unfortunately, only one field study exists on white‐headed woodpeckers in the core of the species range in California Raphael and White, 1984 , where habitat selection and population ecology may differ compared to areas on the range periphery Hardie and Hutchings, 2010 . Our objective was to examine factors influencing white‐headed woodpecker reproductive success in managed forests of northern California. Previous research investigating the impact of structure retention practices on forest bird communities in the region discovered a high prevalence of white‐headed woodpeckers nesting in recent clearcuts. We found the probability of nesting was significantly increased in harvest units with greater densities of retained snags, and estimated high probabilities of occupancy 0.98 and nesting 0.89 across harvest units in the study area Chapter 2 . Here we use nest monitoring data to estimate two components of reproductive success, the 73    probabilities of nest survival and number of young fledged. Nest survival alone does not properly quantify reproductive success, given that it represents one of several events necessary for an individual to contribute to recruitment in a population Etterson et al., 2011 . For avian species that lay a single clutch during the breeding season, like most woodpeckers Martin, 1993 , the sequence of events defining fecundity involve successful pairing and nest construction, egg laying and incubation, rearing of young to independence, and post‐fledging survival Etterson et al., 2011 . Nests that survive to produce 1 young provide an opportunity for recruitment, and this opportunity increases with the number of young fledged. Therefore, quantifying the impacts of forest management on woodpecker populations can be improved through examination of multiple processes that affect recruitment. Our specific objectives were to 1 estimate reproductive parameters of white‐ headed woodpeckers in forests managed for timber production and 2 examine factors that influence these parameters, particularly habitat variables that are altered by forest management practices. We hypothesized that resource availability might influence nest survival but would likely explain more variation in the number of young fledged. 2. Methods 2.1. Study area Our study was conducted on private industrial timberlands surrounded by the Shasta‐Trinity National Forest east of Mt. Shasta in Siskiyou county, northern California centered around 41°21ˊN, 121°50ˊW Figure 3.1 . The area was covered by second‐ and third‐growth, mixed‐conifer forest that was historically dominated by ponderosa pine Pinus ponderosa and is now mostly white fir Abies concolor . Selective logging and fire 74    suppression in the last century have altered the overstory composition similar to other dry forests of western North America Hessburg et al., 2005 . Additional overstory species in decreasing order of prevalence included Douglas‐fir Pseudotsuga menziesii , red fir Abies magnifica , California incense‐cedar Calocedrus decurrens and sugar pine Pinus lambertiana . Non‐forested areas were typically chaparral vegetation composed primarily of manzanita Arctostaphylos spp. . Elevations ranged from 1250–1950m, with true fir Abies spp. dominance generally increasing with elevation. Size classes of forest stands were variable: ~41% was comprised of mature stands with a quadratic mean diameter QMD; diameter of tree representing average basal area 60cm, ~36% contained younger stands with a QMD ranging 25–60cm, and the remaining ~20% of stands were evenly split among past harvests QMD 12–25cm , relatively recent harvests QMD 3–12cm and new harvests planted with seedlings U.S. Forest Service, 2007 . We selected all harvest units that were clearcut with retention between 3–6 yrs prior to 2009 in the study area n 65 . Recent harvests 1 yr old were not selected to avoid potential disturbance effects from harvesting and replanting activities. Harvest units had a mean area 9ha range 4–15ha and typically contained 1–6 patches ~0.1ha of green‐tree retention that were randomly dispersed throughout the unit, in addition to varying densities of snags that were retained during harvest including stumps 1m tall . Slopes ranged from 0–100% mean 28% and were variable across harvest units but mostly homogenous within harvest units, as part of forest planning. Mean monthly temperatures at lower elevations ranged from 2°C in January to 20°C in July; annual precipitation was 124cm with 75% falling as snow from November‐March National 75    Climatic Data Center, National Oceanic and Atmospheric Administration; http://www.ncdc.noaa.gov/ . 2.2. Nest searching and monitoring We searched each harvest unit for white‐headed woodpeckers with a passive survey that involved recording detections of individuals and using behavioral cues e.g., individuals viewed carrying food or entering a snag cavity to determine the presence of a cavity nest. Harvest units were searched systematically from June–July in 2009 and 2010 on days with good weather conditions no rain or heavy wind . Harvest units that were not found to contain a nest were revisited every ~10 days with priority given to those sites where 1 adult white‐headed woodpecker was detected. Nests were monitored with a custom‐built cavity‐viewing camera system using methods similar to Huebner and Hurteau 2007 . During each monitoring visit we recorded date, time, number of eggs or young and estimated nest age based on the size and feather development of nestlings. We attempted to visit each active nest every 3–4 days to accurately determine nest fate. Nests were considered successful if an empty nest was preceded by the observation of large nestlings that appeared close to fledging based on behavior Martin and Geupel, 1993 . Failure was assumed for all nests that were found empty at a date considered too early for fledging according to estimated nest age or when evidence indicated predation. Fledging date for successful nests was estimated as the midpoint between the dates of the last nest visit when the cavity was found to be empty and the visit prior. Initiation date for successful nests was estimated by assuming a total nesting period egg‐laying through fledging of 40 days Wightman et al., 2010 and subtracting the total nesting period from the estimated fledging date. The number of 76    fledged young was estimated as the highest count of nestlings recorded after 80% completion of the nesting period for a given nest Nappi and Drapeau, 2009 ; we chose this criterion to account for nests with a fledging period that spanned multiple days. 2.2. Model covariates We included temporal and biotic factors as model covariates that might influence reproductive success. The temporal factors included ordinal date and year. We included a year effect due to higher winter snowfall and cooler spring temperatures in 2010, which resulted in a delayed onset of nesting. We did not include abiotic factors, such as maximum temperature and precipitation, which are sometimes hypothesized to influence nest survival Wightman et al., 2010; Saab et al., 2011 because we predicted these variables would have little impact. Despite the cooler spring temperatures in 2010, maximum daily temperatures converged quickly between the years to become similar during our nesting observation period June–July and precipitation during that time was minimal and mostly invariant. Therefore, while weather conditions differed between the years prior to the nesting period, daily differences in weather during the nesting period were small and unlikely to explain much variation. Maximum daily temperatures were also highly correlated with date in both years r 0.85 , and we were more interested in comparing the effects of covariates that could be incorporated into both metrics of reproductive success i.e., daily survival rate as a function of date, and number of young fledged as a function of nest completion date . The biotic factors included in the models were described by habitat variables measured at the levels of the nest site, harvest unit, and surrounding landscape. Field measurements were conducted at the completion of nest monitoring to avoid disturbing 77    active nests. We recorded species, height, dbh and decay class Bull et al., 1997 of nest snags, as well as the cavity entrance height and directional bearing. We estimated the live and dead tree basal area surrounding each nest site using variable‐radius plots with a basal area factor 10 prism; this variable‐radius method emphasizes larger trees. We measured snag density at each harvest unit by tallying all standing dead wood 12.7cm dbh and 1m in height, including snags 5m from the harvest unit boundary. Tall stumps were known to be used for cavity nests by white‐headed woodpeckers Milne and Hejl, 1989 and, thus, were included in the snag tallies. Harvest unit boundaries were determined within a geographic information system GIS by examining shapefiles of Timber Harvest Plans THP; California Department of Forestry; ftp://ftp.fire.ca.gov/forest/ and 2 sets of 1‐m resolution aerial photography from the National Agricultural Imagery Program NAIP collected in 2005 and 2009 http://datagateway.nrcs.usda.gov/ . We used the calculated harvest unit area to determine snag density for stems 25.4cm diameter, which represented the lower limit for white‐headed woodpecker cavity nests observed in the field and in previous studies Milne and Hejl, 1989; Kozma, 2009 . We used a GIS to calculate the proportion of mature forest within a 1‐km radius of all nest trees to estimate the availability of large trees for foraging; the 1‐km radius represented the estimated home range of a white‐headed woodpecker Garrett et al., 1996; Hollenbeck et al., 2011 . Mature forest was designated by areas containing trees with a QMD 60cm, as determined by the California Wildlife Habitat Relationships CWHR classification provided in the CALVEG geodatabase U.S. Forest Service, 2007 . We calculated mature forest proportion for both years separately due to timber harvesting that took place between the 2009 and 2010 field seasons. The locations of these new harvest units were mapped with a combination of the 78    THP shapefiles indicating the areas with planned harvests and 2010 Landsat Thematic Mapper imagery http://landsat.gsfc.nasa.gov/ given that the most current NAIP photography was collected prior to the recent harvesting. The CALVEG geodatabase was then updated accordingly. Our selection of model covariates describing habitat was reduced upon examination of the data to include only those factors that exhibited variation and had a biologically meaningful interpretation Table 3.1 . Basal areas of live trees, snags, and all stems 25.4cm dbh were low across nest sites, with 80%, 82%, and 64% of monitored nests having a basal area of 0 for each category, respectively. Snag tree species and decay class also exhibited little variation 84% of nests in true fir and 89% in moderately decayed snags ; therefore, these attributes were not included in the modeling. 2.4. Statistical analysis 2.4.1. Nest survival We modeled nest survival using a generalized linear modeling approach that estimates the probability of a nest surviving a specified time interval as a function of site‐ and time‐specific covariates Rotella et al., 2004 . The approach accounts for data structures that are typical of avian nest survival studies where entry of nests is staggered, nest visit intervals are unequal, and the exact date of nest success or failure is unknown Dinsmore et al., 2002 . In the model specification, nest survives from time to a probability of 1 with , known as the daily survival rate DSR when the time interval is specified in days. The required information for each nest includes the first day the nest is discovered, the last day the nest is checked, the day prior to the last check, and the nest fate success 1, failure 0 . This information is used to construct an encounter history for 79    each nest and the estimation of is obtained by maximizing the likelihood of the product of the probabilities for the observed encounter histories Dinsmore et al., 2002 . Relating individual covariates to the survival probability can be achieved with a logit link function. Following the notation of Royle and Dorazio 2008 , we constructed a model of nest survival such that, logit 0 4 where 1 2 2 6 5 3 is the daily survival rate for nest ; 0 is the logit‐transformed survival probability at average covariate values; 1 and 2 are year‐specific coefficients for the linear and quadratic effects of date, respectively; and 3 , 4 , 5 and 6 are coefficients for the linear effects of nest height, nest tree dbh, snag density in the harvest unit, and proportion of mature forest surrounding the nest, respectively. The average probability of a nest surviving the entire nesting period is calculated by ∏ 1 , where 40 for white‐headed woodpeckers Wightman et al., 2010 . 2.4.2. Number of young fledged We modeled the number of young fledged from successful nests using an ordinal logistic regression which can handle count data that are underdispersed i.e., variance mean Agresti, 2002 . While underdispersion of reproductive data should be a common occurrence for organisms with small clutch/brood sizes 5 having a mean near the midpoint of the range, and has been reported for numerous avian species Cohen, 1988 , this issue appears to be rarely addressed in the literature where simple average sizes are reported and compared but see Paradis et al., 2000 . Ordinal logistic regression estimates 80    the probabilities of an ordinal response that can take Pr Pr 1 Pr 1, 2, … , categories where ; cumulative probabilities are related to covariates with a logit link function Gelman and Hill, 2007 . The logit‐linear models have different intercepts but the same beta coefficients for covariates that are hypothesized to affect the cumulative probabilities, Pr . Thus, the model makes the proportional odds assumption – the effects of covariates are the same across different cumulative logits e.g., the logistic regression comparing 1 to 1 is the same as that for comparing 2 to 2 Agresti, 2002 . Our model for number of young fledged was specified similarly to the nest survival model such that, logit 0 1 4 where . 2 . 6 5 2 3 is the cumulative probability that number of young is at nest , with 1 i.e., the probability that is the maximum observed number of young is 1 ; is an ordered cutoff parameter for the ordinal responses satisfying 1 2 1 ; and represents the summation of covariate effects on the categorical ⋯ probability distribution, with an intercept 0 , year‐specific parameters, 1 and 2 , for the linear and quadratic effects of nest completion date, and additional parameters 3 , 4 , 5 , and 6 for the effects of site‐specific habitat covariates, as explained in the nest survival model. With this model specification, a positive value for a coefficient increases the expected number of young for a given nest. This allows the interpretation of covariate 81    effects to remain consistent between the models for daily survival rate and number of young fledged. 2.4.3. Model parameter estimation We used a Bayesian approach with Markov chain Monte Carlo MCMC to estimate model parameters for both survival and number of young fledged in WinBUGS Spiegelhalter et al., 2003 . WinBUGS uses Gibbs sampling Geman and Geman, 1984 to draw samples from the joint probability distribution of each specified model and estimate posterior distributions for each model parameter. The posterior distributions are summarized by means and credible intervals, which are the Bayesian analogs to confidence intervals. We specified the models and formatted the data inputs using the R2WinBUGS package Sturtz et al., 2005 in R R Development Core Team 2011 . All covariates were standardized to have a mean of 0 and unit variance of 1 to allow comparison of effect sizes. Non‐informative priors were specified for all parameters in the logit‐linear models, with a uniform distribution from 0 to 1 for the intercepts and diffuse normal distributions for parameters describing covariate effects mean 0, variance 100 . Diffuse normal distributions were also specified for the cutoff parameters, with values restricted to satisfy the ordering. We examined the posterior distributions for each set of model parameters after running 3 chains of 100,000 iterations and discarding the first 50,000 iterations. Posterior chains were thinned by 25 to reduce autocorrelation, resulting in 6,000 values forming the posterior distribution for each parameter. Model convergence was assessed by examination of trace plots and the scale reduction factor, or R‐hat statistic. Convergence was assumed when the R‐hat statistic was 1.1 for all model parameters Gelman and Hill, 2007 . We assessed the significance of regression coefficients based on 82    whether the estimated credible intervals overlapped zero; strong support for a covariate effect was indicated when the 95% CRI did not overlap zero, while weak support was indicated when the 90% CRI did not overlap zero. WinBUGS code for model specification is presented in the Appendix A.3, A.4 . 3. Results We monitored 89 nesting attempts by white‐headed woodpeckers from 2009–2010, nearly evenly split between the years with 45 in 2009 and 44 in 2010. Nests were well distributed across the study area: of the 65 harvest units searched each year, 42 were found to have 1 nest in 2009 and 41 in 2010. Nest occupancy shifted among harvest units between the years, as 53 harvest units had 1 nest during the course of the study, and 30 harvest units had 1 nest in both years. A total of 3 harvest units in each year were found to have 2 nests, with 2 harvest units having multiple nests each year. We observed low nest failure in both years, with only 6 of the 89 nesting attempts failing to produce 1 young. All 6 nest failures were assumed to be the result of depredation given lack of evidence otherwise Wightman et al., 2010 , and 3 nest failures were attributed to black bear Ursus americanus based on observations of fallen snags with cavity nests torn open we frequently observed bears both in and around harvest units during the study . The nesting stage upon first discovery was incubation for 22% of nests and nestling for the remaining 78% across the 2 years. The average completion dates for successful nests were estimated at 5 July in 2009 range: 24 June–24 July and 13 July in 2010 range: 28 June–2 August . The number of young fledged from successful nests was similar across years Table 3.2 and ranged from 1–5, with an overall average count of 83    2.94 fledglings SE 0.10 . As expected, this distribution could not be properly described by a Poisson distribution due to severe underdispersion, given that 94% of nests fledged between 2–4 young and the variance 0.84 was much lower than the mean Table 3.2 . The models for daily survival and number of young fledged indicated that none of the temporal or biotic habitat variables had strong support for an effect on either metric of reproductive success Table 3.3 . The daily survival rate estimated at average covariate values was 0.996 95% CRI: 0.987–0.999 , resulting in a period survival rate assuming a 40‐day nesting period of 0.85 95% CRI: 0.59–0.96 . We found weak support for a quadratic date effect on daily survival rate with survival decreasing during the middle portion of the nesting period Figure 3.2 , but the estimate was imprecise. Likewise, we found weak support for a quadratic effect of nest completion date on the number of young fledged: nests that fledged in the middle of the season had a marginally higher expected number of young Figure 3.3 . Support for the quadratic date effect in each model was higher for data from 2010 than that for 2009 Table 3.3; Figures 3.2–3.3 , though the uncertainty makes it difficult to determine whether the annual variation was actually significant. At average values for covariates, the probability that number of young fledged would be 3 was 0.79 95% CRI: 0.62–0.90 . 4. Discussion Reproductive success is a fundamental element of wildlife population dynamics, and management strategies which aim to assist species persistence should address the environmental factors that drive reproductive success. Reproductive success was high for white‐headed woodpeckers nesting on industrial timberlands in northern California during 84    2009 and 2010. We estimated high daily survival rates and subsequently high nest success 0.85 , with the majority of successful nests likely to fledge 3 young. Our nest success estimate is similar to those for white‐headed woodpeckers in unburned managed forests of central Washington 0.84; Kozma, 2009 and higher than those in burned 0.76; Wightman et al., 2010 and unburned 0.39; Hollenbeck et al., 2011 forests of central Oregon. Kozma 2009 described white‐headed woodpecker nesting demographics and found successful nests averaged 2.54 young/nest, which was lower than sympatric hairy woodpeckers Picoides villosus ; this finding was used to hypothesize that white‐headed woodpeckers might be nesting in suboptimal habitat on managed forests in Washington. Nappi and Drapeau 2009 calculated source‐sink status for black‐backed woodpeckers Picoides arcticus using average adult and juvenile survival estimates from the literature for Picoides woodpeckers and suggested that nest productivity nest success young/nest needed to be 2 for the population to serve as a source, under intermediate to high adult and juvenile survival. Assuming intermediate survival rates for adults 0.65 and juveniles 0.33 and using the logic from Nappi and Drapeau 2009 , nest productivity in our study area was ~2.5, indicating that this population of white‐headed woodpeckers may serve as a source. Our study suggests that current forest management practices in northern California are providing habitat conditions which facilitate white‐headed woodpecker population persistence in this region. The few nest failures in our study made it difficult to assess factors influencing nest survival, especially when considering that 3 of the 6 failures were attributed to a large and mobile generalist predator i.e., black bear . The susceptibility of cavity nest depredation to black bears is going to be more a function of snag condition than habitat, given that 85    heavily decayed snags can be easily pushed over and ripped apart. White‐headed woodpeckers are a relatively weak primary excavator Garrett et al., 1996 , so the majority of snags that serve as cavity nests are likely to be at least moderately decayed. Wightman et al. 2010 found similarly low nest depredation rates for white‐headed woodpeckers which they attributed to the lack of live tree density and low cover for small mammalian predators e.g., squirrels that are often responsible for cavity nest depredation Martin, 1995 . The nest sites we observed were also characterized by a lack of live tree cover due to their location within clearcut harvest units. In this context, the removal of tree cover resulting from timber harvest likely has a similar effect to that from wildfire, though other differences between these two disturbance events have important implications for forest ecosystems Lindenmayer and Noss, 2006 . Kozma and Kroll in press found that shrub cover decreased nest survival of white‐headed woodpeckers and hypothesized that woodpeckers had a reduced ability to defend nests against small mammalian predators in areas with high shrub cover. The nests observed in our study had a low average cavity height 3.1m , which is typical of the species Milne and Hejl, 1989 ; thus, the lack of shrub cover in harvest units intensively managed for timber production may provide optimal conditions by increasing nest defense capabilities. Our hypothesis that the variation in number of young fledged could be explained by resource availability as quantified by habitat attributes in and around nest sites was not supported by the data. White‐headed woodpeckers rely on various insects for foraging during the breeding season that likely become more available as summer temperatures increase Elchuk and Wiebe, 2003 which may explain the later timing in their nest initiation compared to sympatric species like hairy woodpeckers Kozma, 2009 . Relative 86    temperatures are typically higher in clearcut harvest units which can increase insect abundance and has been shown to increase woodpecker abundance in stands with tree retention Edworthy et al., 2011 . Thus, food resources across the harvest units in our study area may not have been limiting during the breeding season. Woodpeckers have generally low clutch sizes because of stable food resources that do not confer advantages to fluctuations in reproductive investment from year to year, which is more typical of migratory species that exploit stochastic food resources Wiebe et al., 2006 . While strong excavating species e.g., hairy woodpeckers often have access to a stable food resource by drilling for bark burrowing beetles throughout the year, the use of pine seeds by white‐ headed woodpeckers may result in the same advantages in regions where pine species are a common/dominant overstory tree. Landscapes that contain a mosaic of open and closed canopy forest with presence of mature pine may therefore provide both foraging resources i.e., insects and pine seeds required by white‐headed woodpeckers Wightman et al., 2010 . The high, stable fecundity exhibited in our study area suggests that white‐headed woodpeckers were occupying high quality habitat. 5. Management implications Forest management on private industrial timberlands in the Pacific Northwest is characterized by structure retention policies which mandate that minimum densities of snags and live trees are left behind during harvesting operations. In northern California, these retention policies, in addition to restrictions on harvest unit size, appear to provide nest site conditions for white‐headed woodpeckers that support relatively high nest success and fecundity. We recommend that forest managers retain every available snag 87    that can be safely left standing during harvest operations to maintain the availability of nest sites for woodpeckers. Stands with low snag densities should be supplemented with high‐cut stumps, which have been shown to provide a suitable nesting substrate for white‐ headed woodpeckers. We also recommend that live tree retention should include mature pines e.g., ponderosa and sugar where forest composition permits, given the importance of this food resource for winter survival of white‐headed woodpeckers. Uncertainty exists regarding the temporal dynamics of white‐headed woodpecker occupancy in recent harvest units, specifically the duration over which stands with snag retention can adequately provide nesting opportunities. Harvest units may provide a relatively small window of opportunity given the change in habitat conditions due to snag decay and falling, which reduces nest site availability, and growth of regenerating trees, which increases cover for nest predators. Future research should examine how current landscape‐level forest planning influences the temporal and spatial distribution of nesting habitat for white‐headed woodpeckers. 88    Table 3.1. Summary of potential model covariates describing habitat collected at multiple scales for white‐headed woodpecker nests from 2009–2010 in northern California, USA. Habitat variables were measured at the level of the nest tree, nest site, stand, and landscape for all monitored nests n 89 . Variables in bold were selected for use in the models of reproductive metrics. Variable Mean SD Min Nest tree level Tree diameter cm Nest height m 53.0 24.2 3.1 1.9 17.5 0.9 2.3 4.8 1.7 4.7 0.6 1.4 0 0 0 2.5 1.4 0.3 41 14 9 Nest site level Basal area m2/ha a All stems Live trees Snags Stand level Snag density #/ha Landscape level Mature forest % b a For trees 25.4cm dbh. b Within 1‐km radius of nest tree. 89    152.4 11.0 25.3 25.3 6.9 6.1 71 Max Table 3.2. Count of white‐headed woodpecker nests that successfully fledged young during 2009 and 2010 on private timberlands of northern California, USA. Mean and standard error SE for number of young fledged are also presented for each year. Year 2009 2010 1 2 1 Number of young fledged 2 3 4 15 15 10 11 14 13 90    5 1 1 Mean SE 2.84 0.14 3.05 0.14 Table 3.3. Posterior summaries for model parameters means, standard deviations SD , and 95% credible intervals 0.025 and 0.975 from the logit linear models for nest survival and number of young fledged. Temporal covariates were year‐specific linear and quadratic effects of ordinal date for daily nest survival number of young . , , nest tree diameter . , 2 and nest completion date for 2 . Habitat covariates were nest height , snag density in the harvest unit and mature forest proportion in a 1‐km radius surrounding the nest 91    , . Table 3.3 cont’d Parameter Mean 0.975 0.713 1.729 1.419 4.340 –3.657 –5.737 7.109 3.320 –0.217 1.981 1.715 –0.377 6.137 1.396 0.986 –0.060 3.657 –0.438 –0.291 0.715 –0.377 0.441 0.396 0.480 0.498 –1.366 –1.061 –0.236 –1.439 0.392 0.479 1.653 0.548 0.073 0.622 –0.053 –0.321 1.711 0.560 0.311 0.406 –3.355 –0.432 –0.662 –1.127 3.772 1.777 0.559 0.473 –0.357 0.217 –0.797 0.062 –0.048 0.041 –0.055 –0.019 5      6 Number of young fledged 0 1,1 . 2009 1,2 . 2010 2 2009 2,1 . 2 2010 2,2 . 3 4 5      6 0.025 5.543 –0.221 –2.621 Nest survival 0 1,1 2009 1,2 2010 2 2009 2,1 2 2010 2,2 3 4 SD 0.226 0.221 0.292 0.218 –0.496 –0.394 –0.623 –0.435 0.400 0.473 0.518 0.408 92    Figure 3.1. Harvest units selected to study white‐headed woodpecker reproductive success from 2009–2010 near Mt. Shasta, California, USA. Harvest units Shasta-Trinity NF Private land Roads Oregon California 0 2 4 93    8 km Ü Figure 3.2. Predicted relationship between daily nest survival and ordinal date for white‐ headed woodpeckers during 2009 and 2010 in northern California, USA. Lines represent 0.99 0.98 0.97 2009 0.96 Daily survival rate (DSR) 1.00 mean probabilities solid and 90% credible intervals dashed . 175 180 185 190 195 200 180 185 190 195 200 0.99 0.98 0.97 2010 0.96 Daily survival rate (DSR) 1.00 170 170 175 Ordinal date 94    Figure 3.3. Predicted relationship between expected number of young fledged and ordinal date of nest completion for white‐headed woodpeckers during 2009 and 2010 in northern California, USA. Lines represent mean probabilities solid and 90% credible intervals dashed . 3 2 2009 0 1 # Fledglings/nest 4 5 190 200 210 190 200 210 3 2 2010 0 1 # Fledglings/nest 4 5 180 180 Ordinal date 95    APPENDIX 96    APPENDIX A Appendix A contains the WinBUGS code for all models in chapters 1–3. The code is annotated to provide general descriptions of model components. The model specifications are consistent with the formatting in WinBUGS. WinBUGS requires that normal distributions be specified in terms of precision inverse variance instead of standard deviation; precision is indicated by . Greek symbols from model specifications in the text are typically spelled out in the model code e.g., psi, mu . Index letters in the code are consistent with those in the text, though ordering is sometimes changed for looping purposes. Note that data are required for these models to run in WinBUGS and are not provided here. The names of all data types vectors, matrices, lists required by each model are indicated at the end of the code. Appendix A.1. WinBUGS code for community occupancy model chapter 1 model{ # Define prior distributions for community-level model parameters # Priors defined separately, where appropriate, for the following: # 1) FOREST and OPEN species (f) # 2) study area (h) # 3) year (t) # Occurrence (u) probabilities in year 1 for (f in 1:2){ for (h in 1:4){ u.prob[f,h] ~ dunif(0,1) mu.u[f,h] <- logit(u.prob[f,h]) tau.u[f,h] ~ dgamma(0.1,0.1) }} # Detection (v) probabilities for (f in 1:2){ for (h in 1:4){ 97    for (t in 1:2){ v.prob[f,h,t] ~ dunif(0,1) mu.v[f,h,t] <- logit(v.prob[f,h,t]) tau.v[f,h,t] ~ dgamma(0.1,0.1) }}} # Survival (phi) probabilities for (f in 1:2){ for (h in 1:4){ phi.prob[f,h] ~ dunif(0,1) mu.phi[f,h] <- logit(phi.prob[f,h]) tau.phi[f,h] ~ dgamma(0.1,0.1) }} # Colonization (gamma) probabilities for (f in 1:2){ for (h in 1:4){ gamma.prob[f,h] ~ dunif(0,1) mu.gamma[f,h] <- logit(gamma.prob[f,h]) tau.gamma[f,h] ~ dgamma(0.1,0.1) }} # Parameters for detection covariates (no separate distribution for FOREST vs. OPEN) mu.a1 ~ dnorm(0, 0.001) mu.a2 ~ dnorm(0, 0.001) mu.a3 ~ dnorm(0, 0.001) tau.a1 ~ dgamma(0.1,0.1) tau.a2 ~ dgamma(0.1,0.1) tau.a3 ~ dgamma(0.1,0.1) # Parameters for the occupancy covariates for (f in 1:2){ mu.b1[f] ~ dnorm(0, 0.001) mu.b2[f] ~ dnorm(0, 0.001) mu.b3[f] ~ dnorm(0, 0.001) mu.b4[f] ~ dnorm(0, 0.001) tau.b1[f] ~ dgamma(0.1,0.1) tau.b2[f] ~ dgamma(0.1,0.1) tau.b3[f] ~ dgamma(0.1,0.1) tau.b4[f] ~ dgamma(0.1,0.1) } #Create priors for species i from the community level prior distributions. for (i in 1:n) { for (h in 1:4){ u[i,h] ~ dnorm(mu.u[FD[i],h], tau.u[FD[i],h]) v[i,h,1] ~ dnorm(mu.v[FD[i],h,t], tau.v[FD[i],h,t]) v[i,h,2] ~ dnorm(mu.v[FD[i],h,t], tau.v[FD[i],h,t]) phi[i,h] ~ dnorm(mu.phi[FD[i],h], tau.phi[FD[i],h]) gamma[i,h] ~ dnorm(mu.gamma[FD[i],h], tau.gamma[FD[i],h]) 98    } a1[i] ~ dnorm(mu.a1, tau.a1) a2[i] ~ dnorm(mu.a2, tau.a2) a3[i] ~ dnorm(mu.a3, tau.a3) b1[i] b2[i] b3[i] b4[i] ~ ~ ~ ~ dnorm(mu.b1[FD[i]], dnorm(mu.b2[FD[i]], dnorm(mu.b3[FD[i]], dnorm(mu.b4[FD[i]], tau.b1[FD[i]]) tau.b2[FD[i]]) tau.b3[FD[i]]) tau.b4[FD[i]]) #Estimate the Z matrix (true occurrence for species i at site j) for (j in 1:J) { # occupancy in year 1 logit(psi[j,i,1]) <- u[i,studyarea[j]] + b1[i]*trees[j] + b2[i]*trees[j]*trees[j] + b3[i]*dist[j] + b4[i]*dist[j]*dist[j] Z[j,i,1] ~ dbern(psi[j,i,1]) # occupancy in year 2 logit(psi[j,i,2]) <- phi[i,studyarea[j]]*Z[j,i,1] + gamma[i,studyarea[j]]*(1-Z[j,i,1]) + b1[i]*trees[j] + b2[i]*trees[j]*trees[j] + b3[i]*dist[j] + b4[i]*dist[j]*dist[j] Z[j,i,2] ~ dbern(mu.pi[j,i,1]) #Model detections at site j and survey period k in year t. for (t in 1:2) { for (k in 1:K) { logit(p[j,k,i,t]) <- v[i,h,t] + a1[i]*date[j,t] + a2[i]*date[j,t]*date[j,t] + a3[i]*pdet[j,k,i,t] mu.p[j,k,i,t] <- p[j,k,i,t]*Z[j,i,t] y[j,k,i,t] ~ dbern(mu.p[j,k,i,t]) }}}} #Calculate point level richness estimates for the whole community, FOREST, and OPEN species. for(t in 1:2) { for(j in 1:J) { N[j,t]<- sum(Z[j,1:n,t]) N.forest[j,t]<- inprod(Z[j,1:n,t],forest[1:n]) N.open[j,t]<- inprod(Z[j,1:n,t],open[1:n]) }} } #end model #Data: # y = detection array for site j, survey k, species i, year t # n = number of species # K = number of surveys # studyarea = vector of study area codes (1-4) for each site j # FD = indicator of habitat group (1 or 2) 99    # date = matrix of ordinal dates for site j in year t # pdet = array indicating previous detection # trees = vector of tree counts for each site j # dist = vector of edge distances for each site j # forest = vector of dummy variable for FOREST species # open = vector of dummy variable for OPEN species Appendix A.2. WinBUGS code for multistate occupancy model chapter 2 model{ # state 1 = unoccupied, state 2 = occupied, state 3 = occupied w/ nest for (i in 1:N){ state[i,1] <- 1-psi1[i] #state 1 state[i,2] <- psi1[i]*(1-psi2[i]) #state 2 state[i,3] <- psi1[i]*psi2[i] #state 3 States[i] ~ dcat(state[i,]) #logit occupancy logit(psi1[i]) #logit occupancy logit(psi2[i]) #Mean of spatial Zero[i] <- 0 model for occupied (state = 2) <- b10 + b11*snags[i] + b12*forest[i] + b13[i] model for occupied w/ nest (state = 3) <- b20 + b21*snags[i] + b22*forest[i] + b23[i] random effects for (j in 1:J) { p[i,1,j,1] <- 1 p[i,1,j,2] <- 0 p[i,1,j,3] <- 0 # Pr(detect state1 | state1) # Pr(detect state2 | state1) # Pr(detect state3 | state1) p[i,2,j,1] <- 1-p1[i,j] p[i,2,j,2] <- p1[i,j] p[i,2,j,3] <- 0 # Pr(detect state1 | state2) # Pr(detect state2 | state2) # Pr(detect state3 | state2) p[i,3,j,1] <- 1-p2[i,j] p[i,3,j,2] <- p2[i,j]*(1-d[i,j]) p[i,3,j,3] <- p2[i,j]*(d[i,j]) logit(p1[i,j]) <- a10 + a11*j logit(p2[i,j]) <- a20 + a21*j logit(d[i,j]) <- a30 + a31*j y[i,j] ~ dcat(p[i,States[i],j,]) }} 100    # Pr(detect state1 | state3) # Pr(detect state2 | state3) # Pr(detect state3 | state3) #Priors for spatial random effects b13[1:N] ~ spatial.exp(Zero[1:N],standX[],standY[], tau.sp[1], alpha[1], 1) b23[1:N] ~ spatial.exp(Zero[1:N],standX[],standY[], tau.sp[2], alpha[2], 1) for (w in 1:2){ alpha[w] ~ dunif(1,30) upsilon[w] ~ dunif(0.1,3) tau.ups[w] <- pow(upsilon[w],-2) sigma[w] ~ djl.dnorm.trunc(0,tau.ups[w],0.5,2) tau.sp[w] <- pow(sigma[w],-2) } #Priors for a10.prob ~ a20.prob ~ a30.prob ~ b10.prob ~ b20.prob ~ a10 a20 a30 b10 b20 a11 a21 a31 b11 b12 b21 b22 } <<<<<~ ~ ~ ~ ~ ~ ~ other model parameters dunif(0,1) dunif(0,1) dunif(0,1) dunif(0,1) dunif(0,1) logit(b10.prob) logit(b20.prob) logit(b30.prob) logit(a10.prob) logit(a20.prob) dnorm(0, dnorm(0, dnorm(0, dnorm(0, dnorm(0, dnorm(0, dnorm(0, 0.001) 0.001) 0.001) 0.001) 0.001) 0.001) 0.001) #end model #Data: # y = observed state at site i during survey j # N = number of sites # J = number of surveys # snags = snag density at site j # forest = proportion mature forest within 1-km radius of site j # standX = x coordinate for site j # standY = y coordinate for site j 101    Appendix A.3. WinBUGS code for nest survival model chapter 3 model { for(i in 1:N){ for(t in (first[i]+1):last[i]){ logit(phi[i,t])<- b0 + b1[year[i]]*date[t] + b2[year[i]]*date[t]*date[t] + b3*nestht[i] + b4*nestdbh[i] + b5*snagdensity[stand.id[i]] + b6*forest[i] mu[i,t]<-phi[i,t]*y[i,t-1] y[i,t]~dbern(mu[i,t]) }} #Priors for logit models b0.prob ~ dunif(0,1) b0 <- logit(b0.prob) #Different prior for each year for (yr in 1:2){ b1[yr]~dnorm(0,0.001) b2[yr]~dnorm(0,0.001) } b3 ~ dnorm(0,0.001) b4 ~ dnorm(0,0.001) b5 ~ dnorm(0,0.001) b6 ~ dnorm(0,0.001) } #end model #Data: # y = status (0=failed,1=active,NA=unknown) of nest i during time t # N = number of nests # first = first time period t that nest i was discovered # last = last time period t nest i was observed to be active (y=1) # year = year that nest i was observed (1 or 2) # date = ordinal date of time period t # nestht = cavity height of nest i # nestdbh = diameter of snag in which nest i occurred # stand.id = harvest unit within which nest i occurred # snagdensity = density of snags in harvest unit for nest i # forest = proportion of mature forest in 1-km radius around nest i 102    Appendix A.4. WinBUGS code for offspring model chapter 3 model { for (i in 1:N){ y[i] ~ dcat(P[i,]) #Probability for each observed number of offspring (1-5) P[i,1] <- Q[i,1] P[i,2] <- Q[i,2] - Q[i,1] P[i,3] <- Q[i,3] - Q[i,2] P[i,4] <- Q[i,4] - Q[i,3] P[i,5] <- 1 - Q[i,4] for (j in 1:4){ logit(Q[i,j]) <- lambda[j] - mu[i] } mu[i] <- b0 + b1[year[i]]*date[t] + b2[year[i]]*date[t]*date[t] + b3*nestht[i] + b4*nestdbh[i] + b5*snagdensity[stand.id[i]] + b6*forest[i] } # Ordered cut points for underlying continuous latent variable # Restrict the priors so that lamba[k] < lamba[k+1] lambda[1] ~ dnorm(0, 0.001)I(-20,lambda[2]) lambda[2] ~ dnorm(0, 0.001)I(lambda[1],lambda[3]) lambda[3] ~ dnorm(0, 0.001)I(lambda[2],lambda[4]) lambda[4] ~ dnorm(0, 0.001)I(lambda[3],20) #Priors for logit models b0.prob ~ dunif(0,1) b0 <- logit(b0.prob) #Different prior for each year for (yr in 1:2){ b1[yr]~dnorm(0,0.001) b2[yr]~dnorm(0,0.001) } b3 ~ dnorm(0,0.001) b4 ~ dnorm(0,0.001) b5 ~ dnorm(0,0.001) b6 ~ dnorm(0,0.001) } #end model #Data: # y = number of young fledged by nest i # N = number of nests # year = year that nest i was observed (1 or 2) # date = estimated ordinal date that nest i fledged # nestht = cavity height of nest i 103    # nestdbh = diameter of snag in which nest i occurred # stand.id = harvest unit in which nest i occurred # snagdensity = density of snags in harvest unit for nest i # forest = proportion of mature forest in 1-km radius around nest i 104    LITERATURE CITED 105    LITERATURE CITED Agresti, A., 2002. Categorical data analysis. 2nd edition. Wiley‐Interscience, New York, USA. Angelstam, P., Mikusinski, G., 1994. Woodpecker assemblages in natural and managed boreal and hemiboreal forest ‐ a review. Annales Zoologici Fennici 31, 157‐172. Angermeier, P.L., Karr, J.R., 1994. Biological integrity versus biological diversity as policy directives ‐ protecting biotic resources. Bioscience 44, 690‐697. Arnett, E.B., Kroll, A.J., Duke, S.D., 2010. Avian foraging and nesting use of created snags in intensively‐managed forests of western Oregon, USA. Forest Ecology and Management 260, 1773‐1779. Aubry, K.B., 2007. Changing perceptions of the role of managed forests as wildlife habitat in the Pacific Northwest. In: Harrington, T.B., Nicholas, G.E. Eds. , Managing for wildlife habitat in Westside production forests. General Technical Report PNW‐GTR‐ 695. U.S. Department of Agriculture, Forest Service, pp. 3‐18. Aubry, K.B., Raley, C.M., 2002. The pileated woodpecker as a keystone habitat modifier in the Pacific Northwest. In: Laudenslayer, W.F., Jr., Shea, P.J., Valentine, B.E., Weatherspoon, C.P., Lisle, T.E. Eds. , Proceedings of the symposium on the ecology and management of dead wood in western forests. General Technical Report PSW‐ 181. U.S. Department of Agriculture, Forest Service, Pacific Southwest Research Station, pp. 257‐274. Bagne, K.E., Purcell, K.L., Rotenberry, J.T., 2008. Prescribed fire, snag population dynamics, and avian nest site selection. Forest Ecology and Management 255, 99‐105. Bailey, R.G. 1988. Ecogeographic analysis: a guide to the ecological division of land for resource management. Miscellaneous Publication. U.S. Department of Agriculture, Forest Service, Washington, DC, USA. Bednarz, J.C., Ripper, D., Radley, P.M., 2004. Emerging concepts and research directions in the study of cavity‐nesting birds: Keystone ecological processes. Condor 106, 1‐4. Beese, W.J., Bryant, A.A., 1999. Effect of alternative silvicultural systems on vegetation and bird communities in coastal montane forests of British Columbia, Canada. Forest Ecology and Management 115, 231‐242. Boulinier, T., Nichols, J.D., Sauer, J.R., Hines, J.E., Pollock, K.H., 1998. Estimating species richness: The importance of heterogeneity in species detectability. Ecology 79, 1018‐1028. 106    Bull, E.L., 2002. The value of coarse woody debris to vertebrates in the Pacific Northwest. In: Laudenslayer, W.F., Jr., Shea, P.J., Valentine, B.E., Weatherspoon, C.P., Lisle, T.E. Eds. , Proceedings of the symposium on the ecology and management of dead wood in western forests. General Technical Report PSW‐181. U.S. Department of Agriculture, Forest Service, Pacific Southwest Research Station, pp. 171‐178. Bull, E.L., Parks, C.G., Torgersen, T.R. 1997. Trees and logs important to wildlife in the interior Columbia River basin. General Technical Report PNW‐391. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, p.55. Bunnell, F.L., Boyland, M., Wind, E., 2002. How should we spatially distribute dying and dead wood? In: Laudenslayer, W.F., Jr., Shea, P.J., Valentine, B.E., Weatherspoon, C.P., Lisle, T.E. Eds. , Proceedings of the symposium on the ecology and management of dead wood in western forests. General Technical Report PSW‐181. U.S. Department of Agriculture, Forest Service, Pacific Southwest Research Station, pp. 739‐752. Bunnell, F.L., Huggard, D.J., 1999. Biodiversity across spatial and temporal scales: problems and opportunities. Forest Ecology and Management 115, 113‐126. Bunnell, F.L., Johnson, J.F., 1998. Policy and practices for biodiversity in managed forests: The living dance. UBC Press, Vancouver, BC, Canada. Bunnell, F.L., Kremsater, L.L., Wind, E., 1999. Managing to sustain vertebrate richness in forests of the Pacific Northwest: relationships within stands. Environmental Reviews 7, 97‐146. California Department of Forestry and Fire Protection. 2007. California Forest Practice Rules. California Department of Forestry and Fire Protection, Resource Management, Forest Practice Program, Sacramento, USA. Callicott, J.B., Crowder, L.B., Mumford, K., 1999. Current normative concepts in conservation. Conservation Biology 13, 22‐35. Chambers, C.L., McComb, W.C., Tappeiner, J.C., 1999. Breeding bird responses to three silvicultural treatments in the Oregon Coast Range. Ecological Applications 9, 171‐ 185. Chapin, F.S., Zavaleta, E.S., Eviner, V.T., Naylor, R.L., Vitousek, P.M., Reynolds, H.L., Hooper, D.U., Lavorel, S., Sala, O.E., Hobbie, S.E., Mack, M.C., Diaz, S., 2000. Consequences of changing biodiversity. Nature 405, 234‐242. Chelgren, N.D., Adams, M.J., Bailey, L.L., Bury, R.B., 2011. Using multilevel spatial models to understand salamander site occupancy patterns after wildfire. Ecology 92, 408‐421. Cockle, K.L., Martin, K., Wesolowski, T., 2011. Woodpeckers, decay, and the future of cavity‐ nesting vertebrate communities worldwide. Frontiers in Ecology and the Environment 9, 377‐382. 107    Cohen, Y., 1988. Bayesian‐estimation of clutch size for scientific and management purposes. Journal of Wildlife Management 52, 787‐793. Daily, G.C., Ehrlich, P.R., Haddad, N.M., 1993. Double keystone bird in a keystone species complex. Proceedings of the National Academy of Sciences of the United States of America 90, 592‐594. Desrochers, A., Fortin, M.J., 2000. Understanding avian responses to forest boundaries: a case study with chickadee winter flocks. Oikos 91, 376‐384. Diggle, P.J., Tawn, J.A., Moyeed, R.A., 1998. Model‐based geostatistics. Journal of the Royal Statistical Society Series C‐Applied Statistics 47, 299‐326. Dinsmore, S.J., White, G.C., Knopf, F.L., 2002. Advanced techniques for modeling avian nest survival. Ecology 83, 3476‐3488. Dixon, R.D. 1995. Ecology of white‐headed woodpeckers in the central Oregon Cascades. Thesis, University of Idaho, Moscow, USA. Dorazio, R.M., Kery, M., Royle, J.A., Plattner, M., 2010. Models for inference in dynamic metacommunity systems. Ecology 91, 2466‐2475. Dorazio, R.M., Royle, J.A., 2005. Estimating size and composition of biological communities by modeling the occurrence of species. Journal of the American Statistical Association 100, 389‐398. Dorazio, R.M., Royle, J.A., Soderstrom, B., Glimskar, A., 2006. Estimating species richness and accumulation by modeling species occurrence and detectability. Ecology 87, 842‐854. Drever, M.C., Aitken, K.E.H., Norris, A.R., Martin, K., 2008. Woodpeckers as reliable indicators of bird richness, forest health and harvest. Biological Conservation 141, 624‐634. Drever, M.C., Martin, K., 2010. Response of woodpeckers to changes in forest health and harvest: Implications for conservation of avian biodiversity. Forest Ecology and Management 259, 958‐966. Dudley, J., Saab, V. 2003. A field protocol to monitor cavity‐nesting birds. Research Paper. U.S. Department of Agriculture, Forest Service, Station, R.M.R., 16. Edworthy, A.B., Drever, M.C., Martin, K., 2011. Woodpeckers increase in abundance but maintain fecundity in response to an outbreak of mountain pine bark beetles. Forest Ecology and Management 261, 203‐210. 108    Elchuk, C.L., Wiebe, K.L., 2003. Ephemeral food resources and high conspecific densities as factors explaining lack of feeding territories in Northern Flickers Colaptes auratus . Auk 120, 187‐193. Etterson, M.A., Ellis‐Felege, S.N., Evers, D., Gauthier, G., Grzybowski, J.A., Mattsson, B.J., Nagy, L.R., Olsen, B.J., Pease, C.M., Post van der Burg, M., Potvien, A., 2011. Modeling fecundity in birds: Conceptual overview, current models, and considerations for future developments. Ecological Modelling 222, 2178‐2190. Farnsworth, G.L., Pollock, K.H., Nichols, J.D., Simons, T.R., Hines, J.E., Sauer, J.R., 2002. A removal model for estimating detection probabilities from point‐count surveys. Auk 119, 414‐425. Folke, C., Holling, C.S., Perrings, C., 1996. Biological diversity, ecosystems, and the human scale. Ecological Applications 6, 1018‐1024. Franklin, J.F., 1989. Toward a new forestry. American Forests 95, 37‐44. Franklin, J.F., Spies, T.A., Van Pelt, R., Carey, A.B., Thornburgh, D.A., Berg, D.R., Lindenmayer, D.B., Harmon, M.E., Keeton, W.S., Shaw, D.C., Bible, K., Chen, J.Q., 2002. Disturbances and structural development of natural forest ecosystems with silvicultural implications, using Douglas‐fir forests as an example. Forest Ecology and Management 155, 399‐423. Fraser, G.S., Stutchbury, B.J.M., 2004. Area‐sensitive forest birds move extensively among forest patches. Biological Conservation 118, 377‐387. Garrett, K.L., Raphael, M.G., Dixon, R.D., 1996. White‐headed woodpecker Picoides albolarvatus . In: Poole, A., Gill, F. Eds. , The birds of North America. American Ornithologists' Union, Philadelphia, Pennsylvannia, USA, p. No. 252. Gelman, A., 2006. Prior distributions for variance parameters in hierarchical models. Bayesian Analysis 1, 515‐533. Gelman, A., Carlin, J.B., Stern, H.S., Rubin, D.B., 2003. Bayesian Data Analysis. 2nd edition. Chapman & Hall, Boca Raton, Florida, USA. Gelman, A., Hill, J., 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, New York, USA. Geman, S., Geman, D., 1984. Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. Ieee Transactions on Pattern Analysis and Machine Intelligence 6, 721‐741. Gu, W.D., Swihart, R.K., 2004. Absent or undetected? Effects of non‐detection of species occurrence on wildlife‐habitat models. Biological Conservation 116, 195‐203. 109    Hansen, A.J., Garman, S.L., Weigand, J.F., Urban, D.L., McComb, W.C., Raphael, M.G., 1995. Alternative silvicultural regimes in the Pacific Northwest ‐ simulations of ecological and economic effects. Ecological Applications 5, 535‐554. Hardie, D.C., Hutchings, J.A., 2010. Evolutionary ecology at the extremes of species' ranges. Environmental Reviews 18, 1‐20. Hessburg, P.F., Agee, J.K., Franklin, J.F., 2005. Dry forests and wildland fires of the inland Northwest USA: Contrasting the landscape ecology of the pre‐settlement and modem eras. Forest Ecology and Management 211, 117‐139. Hollenbeck, J.P., Saab, V.A., Frenzel, R.W., 2011. Habitat suitability and nest survival of white‐headed woodpeckers in unburned forests of Oregon. The Journal of Wildlife Management 75, 1061‐1071. Hooper, D.U., Chapin, F.S., Ewel, J.J., Hector, A., Inchausti, P., Lavorel, S., Lawton, J.H., Lodge, D.M., Loreau, M., Naeem, S., Schmid, B., Setala, H., Symstad, A.J., Vandermeer, J., Wardle, D.A., 2005. Effects of biodiversity on ecosystem functioning: A consensus of current knowledge. Ecological Monographs 75, 3‐35. Huebner, D.P., Hurteau, S.R., 2007. An economical wireless cavity‐nest viewer. Journal of Field Ornithology 78, 87‐92. Hunter, J.E., Bond, M.L., 2001. Residual trees: wildlife associations and recommendations. Wildlife Society Bulletin 29, 995‐999. Jones, C.G., Lawton, J.H., Shachak, M., 1994. Organisms as ecosystem engineers. Oikos 69, 373‐386. Kery, M., Royle, J.A., 2008. Hierarchical Bayes estimation of species richness and occupancy in spatially replicated surveys. Journal of Applied Ecology 45, 589‐598. Kozma, J.M., 2009. Nest‐site attributes and reproductive success of White‐headed and Hairy Woodpeckers along the east‐slope Cascades of Washington State. In: Rich, T.D., Arizmendi, C., Demarest, D., Thompson, C. Eds. , Tundra to tropics: connecting birds, habitats and people. Proceedings of the 4th International Partners in Flight Symposium, McAllen, TX, USA, pp. 52‐61. Kozma, J.M., 2011. Composition of forest stands used by white‐headed woodpeckers for nesting in Washington. Western North American Naturalist 71, 1‐9. Kozma, J.M., Kroll, A.J., in press. Woodpecker nest survival in burned and unburned managed ponderosa pine forests of the northwestern USA. Condor. Lawton, J.H., Jones, C.G., 1995. Linking species and ecosystems: organisms as ecosystem engineers. In: Jones, C.G., Lawton, J.H. Eds. , Linking species and ecosystems. Chapman and Hall, New York, NY, USA, pp. 141‐150. 110    Lawton, J.H., Naeem, S., Thompson, L.J., Hector, A., Crawley, M.J., 1998. Biodiversity and ecosystem function: getting the Ecotron experiment in its correct context. Functional Ecology 12, 848‐852. Ligon, J.D., 1973. Foraging behavior of white‐headed woodpecker in Idaho. Auk 90, 862‐ 869. Lindenmayer, D.B., Foster, D.R., Franklin, J.F., Hunter, M.L., Noss, R.F., Schmiegelow, F.A., Perry, D., 2004. Salvage harvesting policies after natural disturbance. Science 303, 1303‐1303. Lindenmayer, D.B., Franklin, J.F., 2002. Conserving forest biodiversity: A comprehensive multiscaled approach. Island Press, Washington, DC, USA. Lindenmayer, D.B., Noss, R.F., 2006. Salvage logging, ecosystem processes, and biodiversity conservation. Conservation Biology 20, 949‐958. Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P., Hector, A., Hooper, D.U., Huston, M.A., Raffaelli, D., Schmid, B., Tilman, D., Wardle, D.A., 2001. Biodiversity and ecosystem functioning: Current knowledge and future challenges. Science 294, 804‐ 808. MacKenzie, D.I., Nichols, J.D., Lachman, G.B., Droege, S., Royle, J.A., Langtimm, C.A., 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83, 2248‐2255. MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Hines, J.E., Bailey, L.L., 2006. Occupancy estimation and modeling: Inferring patterns and dynamics of species occurrence. Elsevier, San Diego, California, USA. MacKenzie, D.I., Nichols, J.D., Seamans, M.E., Gutierrez, R.J., 2009. Modeling species occurrence dynamics with multiple states and imperfect detection. Ecology 90, 823‐ 835. MacKenzie, D.I., Royle, J.A., 2005. Designing occupancy studies: general advice and allocating survey effort. Journal of Applied Ecology 42, 1105‐1114. Marcot, B.G., Vander Hayden, M., 2001. Key ecological functions of wildlife species. In: Johnson, D.H., O'Neil, T.A. Eds. , Wildlife‐habitat relationships in Oregon and Washington. Oregon State University, Corvalis, USA, pp. 168‐186. Martin, J., McIntyre, C.L., Hines, J.E., Nichols, J.D., Schmutz, J.A., MacCluskie, M.C., 2009. Dynamic multistate site occupancy models to evaluate hypotheses relevant to conservation of Golden Eagles in Denali National Park, Alaska. Biological Conservation 142, 2726‐2731. 111    Martin, K., Aitken, K.E.H., Wiebe, K.L., 2004. Nest sites and nest webs for cavity‐nesting communities in interior British Columbia, Canada: Nest characteristics and niche partitioning. Condor 106, 5‐19. Martin, K., Eadie, J.M., 1999. Nest webs: A community‐wide approach to the management and conservation of cavity‐nesting forest birds. Forest Ecology and Management 115, 243‐257. Martin, T.E., 1993. Evolutionary determinants of clutch size in cavity‐nesting birds ‐ nest predation or limited breeding opportunities. American Naturalist 142, 937‐946. Martin, T.E., 1995. Avian life‐history evolution in relation to nest sites, nest predation, and food. Ecological Monographs 65, 101‐127. Martin, T.E., Geupel, G.R., 1993. Nest‐monitoring plots: methods for locating nests and monitoring success. Journal of Field Ornithology 64, 507‐519. Mikusinski, G., Gromadzki, M., Chylarecki, P., 2001. Woodpeckers as indicators of forest bird diversity. Conservation Biology 15, 208‐217. Milne, K.A., Hejl, S.J., 1989. Nest‐site characteristics of white‐headed woodpeckers. Journal of Wildlife Management 53, 50‐55. Mordecai, R.S., Mattsson, B.J., Tzilkowski, C.J., Cooper, R.J., 2011. Addressing challenges when studying mobile or episodic species: hierarchical Bayes estimation of occupancy and use. Journal of Applied Ecology 48, 56‐66. Morrison, M.L., Raphael, M.G., 1993. Modeling the dynamics of snags. Ecological Applications 3, 322‐330. Naeem, S., 1998. Species redundancy and ecosystem reliability. Conservation Biology 12, 39‐45. Naeem, S., 2002. Ecosystem consequences of biodiversity loss: The evolution of a paradigm. Ecology 83, 1537‐1552. Naeem, S., Li, S.B., 1997. Biodiversity enhances ecosystem reliability. Nature 390, 507‐509. Nappi, A., Drapeau, P., 2009. Reproductive success of the black‐backed woodpecker Picoides arcticus in burned boreal forests: Are burns source habitats? Biological Conservation 142, 1381‐1391. NCASI. 2008. The biological basis for forest practices rules for retaining coarse woody debris in managed forests of the Pacific Northwest. Technical Bulletin No. 954. National Council for Air and Stream Improvement, Inc., Research Triangle Park, North Carolina, USA. 112    Neitro, W.A., Binkley, V.W., Cline, S.P., Mannan, R.W., Marcot, B.G., Taylor, D., Wagner, F.F., 1985. Snags wildlife trees . In: Brown, E.R. Ed. , Management of wildlife and fish habitats in forests of western Oregon and Washington. U.S. Department of Agriculture, Forest Service, Portland, OR, USA, pp. 129‐169. Newton, I., 1994. The role of nest sites in limiting the numbers of hole‐nesting birds ‐ a review. Biological Conservation 70, 265‐276. Nichols, J.D., Hines, J.E., MacKenzie, D.I., Seamans, M.E., Gutierrez, R.J., 2007. Occupancy estimation and modeling with multiple states and state uncertainty. Ecology 88, 1395‐1400. Noss, R.F., 1990. Indicators for monitoring biodiversity: a hierarchical approach. Conservation Biology 4, 355‐364. Oregon Department of Forestry. 2005. Oregon Forest Practices Act. Oregon Revised Statutes: ORS 527.610‐992. Oregon Department of Forestry, Salem, USA. Paradis, E., Baillie, S.R., Sutherland, W.J., Dudley, C., Crick, H.Q.P., Gregory, R.D., 2000. Large‐ scale spatial variation in the breeding performance of song thrushes Turdus philomelos and blackbirds T. merula in Britain. Journal of Applied Ecology 37, 73‐ 87. Petchey, O.L., Gaston, K.J., 2002. Functional diversity FD , species richness and community composition. Ecology Letters 5, 402‐411. Power, M.E., Tilman, D., Estes, J.A., Menge, B.A., Bond, W.J., Mills, L.S., Daily, G., Castilla, J.C., Lubchenco, J., Paine, R.T., 1996. Challenges in the quest for keystones. Bioscience 46, 609‐620. Preston, M.I., Harestad, A.S., 2007. Community and species responses by birds to group retention in a coastal temperate forest on Vancouver Island, British Columbia. Forest Ecology and Management 243, 156‐167. Ralph, C.J., Sauer, J.R., Droege, S. 1995. Monitoring bird populations by point counts. General Technical Report PSW‐149. U.S. Department of Agriculture, Forest Service, Pacific Southwest Research Station, 185. Raphael, M.G., White, M., 1984. Use of snags by cavity‐nesting birds in the Sierra Nevada. Wildlife Monographs, 1‐66. Redford, K.H., Richter, B.D., 1999. Conservation of biodiversity in a world of use. Conservation Biology 13, 1246‐1256. Ries, L., Fletcher, R.J., Battin, J., Sisk, T.D., 2004. Ecological responses to habitat edges: Mechanisms, models, and variability explained. Annual Review of Ecology Evolution and Systematics 35, 491‐522. 113    Rose, C.L., Marcot, B.G., Mellen, T.K., Ohmann, J.L., Waddell, K.L., Lindley, D.L., Schreiber, B., 2001. Decaying wood in Pacific Northwest forests: concepts and tools for habitat management. In: Johnson, D.H., O'Neil, T.A. Eds. , Wildlife‐habitat relationships in Oregon and Washington. Oregon State University, Corvalis, USA, pp. 580‐623. Rosenvald, R., Lõhmus, A., 2008. For what, when, and where is green‐tree retention better than clear‐cutting? A review of the biodiversity aspects. Forest Ecology and Management 255, 1‐15. Rotella, J.J., Dinsmore, S.J., Shaffer, T.L., 2004. Modeling nest‐survival data: A comparison of recently developed methods that can be implemented in MARK and SAS. Animal Biodiversity and Conservation 27, 187‐205. Royle, J.A., Dorazio, R.M., 2008. Hierarchical modeling and inference in ecology. Academic Press, Boston, USA. Royle, J.A., Kery, M., 2007. A Bayesian state‐space formulation of dynamic occupancy models. Ecology 88, 1813‐1823. Ruiz‐Gutierrez, V., Zipkin, E.F., Dhondt, A.A., 2010. Occupancy dynamics in a tropical bird community: unexpectedly high forest use by birds classified as non‐forest species. Journal of Applied Ecology 47, 621‐630. Russell, R.E., Royle, J.A., Saab, V.A., Lehmkuhl, J.F., Block, W.M., Sauer, J.R., 2009a. Modeling the effects of environmental disturbance on wildlife communities: avian responses to prescribed fire. Ecological Applications 19, 1253‐1263. Russell, R.E., Saab, V.A., Rotella, J.J., Dudley, J.G., 2009b. Detection probabilities of woodpecker nests in mixed conifer forests in Oregon. Wilson Journal of Ornithology 121, 82‐88. Saab, V.A., Russell, R.E., Rotella, J., Dudley, J.G., 2011. Modeling nest survival of cavity‐ nesting birds in relation to postfire salvage logging. Journal of Wildlife Management 75, 794‐804. Saracco, J.F., Siegel, R.B., Wilkerson, R.L., 2011. Occupancy modeling of black‐backed woodpeckers on burned Sierra Nevada forests. Ecosphere 2, Article 31. Schieck, J., Hobson, K.A., 2000. Bird communities associated with live residual tree patches within cut blocks and burned habitat in mixedwood boreal forests. Canadian Journal of Forest Research 30, 1281‐1295. Simberloff, D., 1998. Flagships, umbrellas, and keystones: Is single‐species management passe in the landscape era? Biological Conservation 83, 247‐257. Spiegelhalter, D.J., Thomas, A., Best, N.G., Lunn, D., 2003. WinBUGS Version 1.4 User Manual. In. MRC Biostatistics Unit, Cambridge, UK. 114    Sturtz, S., Ligges, U., Gelman, A., 2005. R2WinBUGS: A package for running WinBUGS from R. Journal of Statistical Software 12, 1‐16. Sustainable Forestry Initiative. 2011. The bigger picture. Sustainable Forestry Initiative 2011 Progress Report. Washington, DC, USA. Swanson, F.J., Franklin, J.F., 1992. New forestry principles from ecosystem analysis of Pacific‐Northwest forests. Ecological Applications 2, 262‐274. Thomas, J.W., Anderson, R.G., Maser, C., Bull, E.L., 1979. Snags. In: Thomas, J.W. Ed. , Wildlife habitat in managed forests: The Blue Mountains of Oregon and Washington. Agricultural Handbook 553. U.S. Department of Agriculture, Washington, D.C., pp. 60‐77. U.S. Forest Service. 2007. Existing Vegetation ‐ CALVEG. Digital map. USDA Forest Service, Pacific Southwest Region, Remote Sensing Lab, McClellan, California, USA. Virkkala, R., 2006. Why study woodpeckers? The significance of woodpeckers in forest ecosystems. Annales Zoologici Fennici 43, 82‐85. Walter, S.T., Maguire, C.C., 2005. Snags, cavity‐nesting birds, and silvicultural treatments in Western Oregon. Journal of Wildlife Management 69, 1578‐1591. Washington Forest Practices Board. 2002. Forest Practices Rules. Title 222 WAC Washington State Department of Natural Resources, Olympia, USA. Wiebe, K.L., Koenig, W.D., Martin, K., 2006. Evolution of clutch size in cavity‐excavating birds: The nest site limitation hypothesis revisited. American Naturalist 167, 343‐ 353. Wightman, C.S., Saab, V.A., Forristal, C., Mellen‐McLean, K., Markus, A., 2010. White‐headed woodpecker nesting ecology after wildfire. Journal of Wildlife Management 74, 1098‐1106. Wilson, E.O., 1999. The diversity of life. 2nd edition. W. W. Norton, New York, NY, USA. Yoccoz, N.G., Nichols, J.D., Boulinier, T., 2001. Monitoring of biological diversity in space and time. Trends in Ecology & Evolution 16, 446‐453. Zarnowitz, J.E., Manuwal, D.A., 1985. The effects of forest management on cavity‐nesting birds in northwestern Washington. Journal of Wildlife Management 49, 255‐263. Zipkin, E.F., Dewan, A., Royle, J.A., 2009. Impacts of forest fragmentation on species richness: a hierarchical approach to community modelling. Journal of Applied Ecology 46, 815‐822. 115    Zipkin, E.F., Royle, J.A., Dawson, D.K., Bates, S., 2010. Multi‐species occurrence models to evaluate the effects of conservation and management actions. Biological Conservation 143, 479‐484. 116