AN ABSTRACT OF THE DISSERTATION OF Michael C. Wimberly for the degree of Doctor of Philosophy in Forest Science presented on May 14, 1999. Title: Watershed-Scale Vegetation Patterns in a Late-Successional Forest Landscape in the Oregon Coast Range. Signature redactedfor privacy. Abstract approved: Thomas A. Spies Knowledge about vegetation patterns and ecological processes in unmanaged, late- successional watersheds is needed to provide a foundation for forest management strategies aimed at conserving native biodiversity. I examined influences of environmental variability and disturbance history on forest structure and composition in the Cummins Creek Wilderness, located on the central Oregon coast. Climatic and topographic variables explained the majority of hilislope community composition, while fire history explained most of the variability in hilislope forest structure. Forest structure and composition in riparianareas was related to a climatic gradient as well as position in the stream network. The abundance of two fire-sensitive species, Tsuga heterophylla (western hemlock) and Picea sitchensis (Sitka spruce), decreased with distance from old-growth patches, possibly reflectinga seed dispersal gradient that occurred following fires 80 to 140 yearsago. I developed predictive maps of understory conifer patterns using remote sensing, aerial photographs, digital elevation models and streammaps. I predicted P. sitchensis regeneration based on distance from the coast and topography, and T heterophylla regeneration based on crown size, percent hardwood composition, topography, and distance from old-growth patches. Although I found statistically significant relationships between understory patterns and GIS predictor variables, the models explained only low to moderate amounts of the overall variability. Landscape-scale simulations ofT. heterophylla showed that population expansion through gap-phase recruitment was limited by short seed dispersal distances in closed-canopy forests, the requirement for canopy gap disturbances to facilitate overstory recruitment, and the lag between recruitment and reproduction. Although fine-scale habitat features can influence the amount of regeneration in a gap when seed sources are present, the fire regime may ultimately control the abundance of 1'. heterophylla at the landscape scale through dispersal limitations. Brief increases in fire frequency can cause a sustained decrease in the amount ofT. heterophylla on the landscape once fire frequency is reduced below a threshold value. Our results emphasize the complexity and diversity of forest vegetation at the watershed scale. Environmental variability, disturbance history, and dispersal limitations have all played a role in creating the current landscape patterns in the Cummins Creek Wilderness. © Copyright Michael C. Wimberly May 14, 1999 A!! Rights Reserved Watershed-Scale Vegetation Patterns in a Late-Successional Forest Landscape in the Oregon Coast Range by Michael C. Wimberly A DISSERTATION submitted to Oregon State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy Presented May 14, 1999 Commencement June, 2000 Doctor of Philosophy dissertation of Michael C. Wimberly presented on May 14, 1999 Approved: Signature redacted for privacy. Maj& Professor, repreynting Fortst Science Signature redacted for privacy. Chair of De eof Forest Science Signature redacted for privacy. Dean of Gra(uate School ' I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my dissertation to any reader upon request. Signature redacted for privacy. Michael C. Wimberly, AMior ACKNOWLEDGEMENTS This work would not have been possible without assistance from many people. Tom Spies, my major professor, allowed me the freedom to develop and pursue my own research ideas, but also provided the right amount of guidance, insight, and constructive criticism. My other committee members, Julia Jones, Bruce McCune, and John Tappeiner, all supplied feedback and advice during the formulation of my study plan and the development of my dissertation. Bill Proebsting served as an able grad rep. Colleen Grenz, Nick Maetske, and Leslie Trabant were intrepid field assistants through many long days of of shrubs, bugs, rain, and steep hilislopes. Rebecca Hess and Rob Pabst also pitched in to help with the field work. I am also thankful for the camaderie of many other colleagues in the Forest Science department. Finally, the lion's share of my gratitute is reserved for wife, Anne, who has provided love and support throughout my long weeks in the field and 16-hour days of dissertation editing. I hope I can find a way to repay her someday. This research was funded through a graduate training fellowship in landscape studies from the National Science Foundation, a Saubert fellowship from the Department of Forest Science at Oregon State University, and research support from the USDA Forest Service Pacific Northwest Research Station. TABLE OF CONTENTS Page CHAPTER 1: General Introduction 1 Background 1 Overview of the Study 7 CHAPTER 2: Assessing the Influences of Disturbance History and Environmental Variability on Forest Structure and Composition in Coastal Oregon Watersheds 14 Abstract 15 Introduction 16 Study Area 22 Methods 25 Vegetation Data 25 Environmental Data 27 Disturbance Data 28 Data Analyses 30 Results 34 Disturbance History 37 Differences Between Riparian and Hilislope Forests 34 Species Composition Patterns 41 Forest Structure Patterns 50 Relative Importance of Disturbance and Environment 56 Discussion 58 Recent Fire History 58 Riparian Forests 60 Landscape Patterns 62 Implications for Management and Conservation 66 CHAPTER 3: Predicting Landscape Patterns of Understory Conifers Using Field Plots, Remote Sensing, and Digital Elevation Models 69 Abstract 70 Introduction 71 Study Area 76 Methods 77 Field Data Collection 77 TABLE OF CONTENTS (CONTINUED) Page GIS Data Layers 80 Predictive Models 84 Correlation with Stand Variables 86 Results 87 Discussion 101 Validity of the Modeling Approach 101 Remnants and the Role of Disturbance 103 Riparian Areas 105 Fate of Hardwood Patches 106 Conclusions 108 CHAPTER 4: Modeling the Temporal and Spatial Dynamics of a Late-Successional, Fire-Sensitive Tree Species 110 Abstract 111 Introduction 111 Conceptual Framework 114 Modeling Approach 118 Methods 119 Model Description 119 Model Analysis 125 Results 127 Discussion 134 Conclusions 143 CHAPTER 5: Conclusions 145 Overview 145 Conclusions 149 BIBLIOGRAPHY 152 APPENDIX 165 LIST OF FIGURES Figure Page 1.1 Conceptual model of watershed-scale vegetation patterns 8 1.2 Location of the Cumm ins Creek Wilderness 10 2.1 Map of field sites in the Cummins Creek Wilderness 23 2.2 Age class distribution by topographic position 35 2.3 Redundancy analysis biplot for hilislope species composition 43 2.4 Basal area per hectare of a) T heterophylla and b) P. sitchensis as a function of distance from the nearest old-growth patch. 46 2.5 Redundancy analysis biplot for riparian species composition. 48 2.6 Overstory species composition as a function of distance from the coast for a) riparian areas and b) hillslopes. 49 2.7 Redundancy analysis biplot for hillslope forest structure. 51 2.8 Redundancy analysis biplot for riparian forest structure. 55 2.9 Percentage of the total variation explained (TVE) accounted for by disturbance and environment variables in the redundancy analysis ordinations. 57 3.1 Maps of field site locations in the Cummins Creek Wilderness with raw data values 78 3.2 Maps of independent variables used to develop the predictive maps, including a) distance from remnant old-growth patches, b) percent broadleafcover and c) the riparian network 83 3.3 Percentage of sites with P. sitchensis regeneration as a function of a) distance from the coast and b) slope position 89 3.4 P. sitchensis regeneration density as a function of distance from thecoast for riparian areas versus hillslopes. 89 3.5 Maps of predicted conifer patterns generated usinga fixed cutoff to predict presence/absence and the mean regression response to predict density. a) P. sitchensis advance regeneration, b) T. heterophylla advance regeneration and c) T heterophylla understory trees. Shadedareas represent cut stands that were excluded from our analysis. 83 LIST OF FIGURES (CONTINUED) Figure Page 3.6 Proportion of sites with T. heterophylla regeneration as a function of a) distance from the nearest old-growth remnant patch, b) percent broadleaf cover, c) overstory conifer crown diameter and d) percent slope. 92 3.7 T. heterophylla advance regeneration density as a function of a) distance from the coast for low-order riparian areas versus other landfonn types, b) slope position and c) landform 93 3.8 Proportion of sites with understory T. heterophylla trees as a function of a) distance from remnants and b) percent broadleaf cover 95 3.9 T heterophylla advance regeneration density as a function of a) percent broadleaf cover, b) distance from old-growth remnant patches and c) landform 96 3.10 Maps of predicted conifer patterns generated using probabilistic predictions of presence/absence and incorporating random error into the prediction of density. a) P. sitchensis advance regeneration, b) T heterophylla advance regeneration c) T heterophylla understory trees 97 4.1 State and transition diagram for the T. heterophylla patch dynamics model 120 4.2 Probability of T heterophylla establishment