Open Dlaflowerthesisfinal.Pdf

The Pennsylvania State University

The Graduate School

Intercollege Graduate Degree Program in Ecology

PROJECTING THE EFFECTS OF CLIMATE CHANGE AND MANAGEMENT IN

FORESTS OF THE PUGET SOUND LOWLANDS

A Thesis in

Ecology

Danelle Laflower

Submitted in Partial Fulfillment of the Requirements for the Degree of

Master of Science

August 2015

The thesis of Danelle Laflower was reviewed and approved* by the following:

Matthew D. Hurteau Assistant Professor of Forest Resources, Department of Ecosystem Science and Management Thesis Advisor

Margot W. Kaye Associate Professor of Forest Ecology, Department of Ecosystem Science and Management

Katriona Shea Alumni Professor in the Biological Sciences, Department of Biology and Center for Infectious Disease Dynamics

David Eissenstat Professor of Woody Plant Physiology Chair of Ecology Intercollege Graduate Degree Program

*Signatures are on file in the Graduate School

iii

ABSTRACT

Tree species have differential responses to changing climate, which may affect competitive interactions and alter species composition and forest carbon dynamics. In energy-limited systems, warming temperatures may lengthen growing season and increase water demand. In water-limited systems, increased evapotranspiration associated with warmer temperatures may limit species diversity and productivity. Management may help maintain productivity under changing climate by reducing competition for resources, increasing heterogeneity, and decreasing disturbance risk. I sought to quantify how projected changes in climate may alter the composition and productivity of Puget Lowland forests, and how management could mitigate changes. I also sought to determine the effects of management on, and carbon tradeoffs of, restoring oak woodland habitats. I used a simulation approach to model four landscape-scale treatments (control, burn only, thin only, thin and burn) using current climate and projected climate for two general circulation models (CCSM 4 and CNRM-CM5), driven by two emission scenarios (moderate (RCP 4.5) and high (RCP 8.5) emissions) at Joint Base Lewis-McChord, Washington. At the landscape-scale, I found little difference in carbon dynamics and species composition under the moderate emission scenarios. However, by late-century under the high emission scenario, I found substantial declines in productivity due to a decrease in late-successional species growth. The major effect of management was to increase net ecosystem carbon balance after 2030, relative to the control. Relative to 2012, I found the frequency of sites that were likely to contain Oregon white oak decreased over time under all treatments, regardless of climate. To counteract the decline, I added an intensively managed 708 ha oak restoration area (2% of the landscape) to the landscape-scale thin and burn treatment. Relative to the thin and burn treatment, the intensive management increased oak presence and incurred small initial carbon costs (<400 g C m-2) that decreased over time. My research suggests that under the late-century high emission scenario, management cannot alter the climate- driven decline of late-successional species, but intensive management can alter forest structure, which may increase species richness and help alleviate declines in productivity.

TABLE OF CONTENTS

List of Figures ...... v

List of Tables ...... vii

Acknowledgements ...... ix

Projecting the effects of climate change and management in forests of the Puget Sound Lowlands ...... 1

Introduction ...... 1 Methods ...... 4 Study Area ...... 4 Field Data ...... 6 Simulation Approach...... 7 Model Parameterization ...... 8 Climate Data ...... 16 Simulation Experiment and Data Analysis ...... 16 Data and Model Limitations ...... 18 Results ...... 19 Climate Effects on Forest C Dynamics and Species Composition ...... 19 Management Effects on Forest C Dynamics and Species Composition ...... 25 Oak Restoration ...... 31 Discussion ...... 37 Productivity and Carbon Stocks ...... 37 Successional Trajectory ...... 38 Management Effects on Forest Productivity and Composition ...... 39 Managing Under Uncertainty ...... 40 Conclusion ...... 42 References ...... 43 Appendix A Annual projected temperature data...... 51 Appendix B Annual projected precipitation data...... 52 Appendix C LANDIS-II decadal summer (June, July, August) precipitation...... 53 Appendix D LANDIS-II decadal temperature projections...... 54

LIST OF FIGURES

Figure 1. Joint Base Lewis-McChord and surrounding area land cover. Study area location within Washington in red...... 6

Figure 2. Comparison of aboveground biomass (AGB) of field plot data (n=346) and LANDIS-II output of all forested pixels (n=5533). Boxes represent interquartile range...... 13

Figure 3. Net Ecosystem Carbon Balance (NECB) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean NECB and shading the 95% confidence intervals from 25 replicate simulations...... 21

Figure 4. Net Ecosystem Carbon Balance (NECB) under baseline climate and an average result from two general circulation model projections (CCSM and CNRM) driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean NECB and shading the 95% confidence interval estimates from 25 replicate simulations...... 22

Figure 5. Year 2100 species richness comparison of baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios for frequency of cells for a particular species richness. Lines are the mean species richness and shading the 95% confidence intervals from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 23

Figure 6. Landscape-scale aboveground carbon (AGC) stocks for Douglas-fir and pooled late successional species (western hemlock and western redcedar) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean AGC and shading the 95% confidence intervals from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 24

Figure 7. Total Ecosystem Carbon (TEC) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean TEC and shading the 95% confidence interval estimates from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 25

Figure 8. Net Ecosystem Carbon Balance (NECB) of simulated forest management treatments (control, burn only, thin only, thin and burn) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean NECB and shading the 95% confidence intervals from 25 replicate simulations...... 26

Figure 9. Total Ecosystem Carbon (TEC) of simulated forest management treatments (control, burn only, thin only, thin and burn) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean TEC and shading the 95% confidence intervals from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 28

Figure 10. Aboveground carbon (AGC) stocks of Douglas-fir and pooled late successional species (western hemlock and western redcedar) for simulated forest management treatments (control, burn only, thin only, thin and burn) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean AGC and shading the 95% confidence intervals from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 30

Figure 11. Joint Base Lewis-McChord grid cells of year 2100 oak probability surfaces of simulated forest management treatments (control, burn only, thin only, thin and burn) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Darker areas have a higher probability of oak occurrence. For each treatment n=25 replicate simulations...... 33

Figure 12. Year 2100 empirical cumulative distribution of simulated forest management treatments (control, burn only, thin only, thin and burn) under baseline climate for probability of oak existence. For each treatment n=7386, the number of pixels in simulation...... 34

Figure 13. Year 2100 oak probability surface in Oak Restoration Area of simulated management treatments (A. thin and burn, and B. oak restoration thin and burn) under baseline climate. Darker green pixels have a higher probability of oak occurrence. Black areas are not managed by Joint Base Lewis-McChord and were excluded. For each treatment n=25 replicate simulations...... 35

Figure 14. Total ecosystem carbon (TEC) of simulated forest management treatments (thin and burn, and thin and burn with oak restoration) under baseline climate and climate projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. Lines are the mean TEC and shading the 95% confidence intervals from 25 replicate simulations. Confidence intervals appear small due to the scale of the Y axis relative to the size of the confidence interval...... 36

vii

LIST OF TABLES

Table 1. Species parameters for LANDIS-II. Columns are as follows: Longevity is the maximum tree age in years, Sexual Maturity is the age of earliest seed production, Shade Tol. and Fire Tol. are ranges between 1-5, with 1 being the least tolerant and 5 being the most tolerant, Effective Disp. and Max Disp. are mean and maximum dispersal distance in meters, Sprout Prob. is the probability of the species stump sprouting, Min Sprout and Max Sprout are the minimum and maximum age that sprouting may occur, and Sprouting denotes species that can reproduce by sprouting. .. 10

Table 2. Century Succession species parameters for LANDIS-II. Columns are as follows: N-fix denotes nitrogen-fixing species, minimum and maximum growing degree days are GDDmin and GDDmax, Min T is minimum temperature for establishment in °C, Max Drought is proportion of growing season above which a species cannot establish if available water falls below zero, Leaf Long. is number of years for leaf persistence, Epicorm Resprout denotes species that have epicormic resprouting ability; Lignin of Leaf, Fine root, Wood, Coarse Root are the lignin percentages for each component; C:N of Coarse Root, Leaf, Fine Root, Wood, Coarse Root, Litter are the carbon to nitrogen ratios of each component...... 11

Table 3. Initial Century Succession ecoregion parameters for soil organic matter. Columns are as follows: Eco denotes ecoregion; SOM1, 2, 3 denote soil organic matter fast, slow, and passive pools of carbon (C) and nitrogen (N) in g m-2; Minrl N is available mineral N g m-2...... 11

Table 4. Fire behavior parameters for the Dynamic Fire extension. Columns are as follows: Region is fire ecoregion; Fire size parameters, Mu (mean fire size (ha)), Sigma (standard deviation of mean fire size (ha)), and Max (maximum size (ha)); Open Fuel indicates the value that represents a no tree fuel index; Num Igns is the number of expected yearly ignitions; MFRI is the mean fire return interval; foliar moisture content ((FMC) high (FMCH), low (FMCL)) help determine fire size as a function of seasonal fire weather (Spring (Sp), Summer (Su), Fall (Fa)); and proportion of fires that occur during the seasonal high period (HProp)...... 14

Table 5. Kolmogrov-Smirnov one-sided test results for year 2100, for oak probability of simulated forest management treatments (burn only, thin only, thin and burn) against control under baseline climate. H0: No difference in distribution. Ha: the cumulative distribution function of the control lies above that of the active treatment. For each treatment n=7386, the number of pixels in simulation...... 32

Table 6. Kolmogrov-Smirnov one-sided test results for comparison of year 2100 probability of oak occurrence for each climate scenario control against the baseline climate control. The climate scenarios are from two general circulation models (CCSM and CNRM) driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. H0: No difference in distribution. Ha: the cumulative distribution function of the control lies above that of the increased emission scenario. For each treatment n=7386, number of pixels in simulation...... 34

viii

Table 7. Kolmogrov-Smirnov one-sided test results for comparison of year 2100 probability of oak occurrence for three treatments (burn only, thin only, thin and burn) against the control under each climate scenario. Climate scenarios include projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. H0: No difference in distribution. Ha: the cumulative distribution function of the control lies above that of the increased emission scenario. For each treatment n=7386, number of pixels in simulation...... 35

Table 8. Year 2100 mean oak carbon stocks and mean cell-level species richness within the 708 ha oak restoration area. Values in parentheses are standard error of the mean for each treatment (n=25 replicate simulations)...... 36

ACKNOWLEDGEMENTS

I would like to thank my advisor, Dr. Matthew Hurteau, for this opportunity and his guidance. I have learned so much from his expertise and thoroughness. I would like to thank my committee members, Dr. Margot Kaye and Dr. Katriona Shea for their comments and suggestions. I would like to acknowledge my funding sources, the Strategic Environmental

Research and Development Program and the Penn State Ecology Program. I would like to thank

Dr. Katie Martin and Dr. Megan Creutzberg for their advice and help with the initial parameterization of the LANDIS-II model. I would like to thank Shuang Liang for her suggestions, clarifications, and encouragement. I would also like to thank everyone else who made this possible and who brightened my day.

Projecting the effects of climate change and management in forests of the Puget Sound Lowlands

Introduction

Climate and disturbance influence forest productivity and species composition, both of which drive carbon (C) cycle dynamics in forests (Littell et al. 2010, Walther et al. 2002).

Temperature increases associated with changing climate can alter the abiotic environment by lengthening the growing season and increasing water demand; causing changes that span from increasing productivity (Anderson-Teixeira et al. 2013) to species’ range shifts (Rehfeldt et al.

2006, Littell et al. 2008, Coops and Waring 2011). Disturbance can alter resource availability, increase spatial heterogeneity, and alter successional trajectories (Harmon et al. 1986, Franklin et al. 2002, Flower and Gonzalez-Meler in press). Additionally, changing climate and disturbance may differentially affect species-specific growth and mortality rates, causing changes to forest C dynamics (Kashian et al. 2006, Littell et al. 2010, Flower and Gonzalez-Meler in press). In managed forests, human intervention can alter resource availability, species composition, and demographic patterns, such that management may mitigate some or all of the effects of changing climate and disturbance (Hurteau and North 2010, Spies et al. 2010, Kerhoulas et al. 2013). An improved understanding of how changing climate will differentially affect species and interspecific interactions is a necessary foundation for managing landscapes under novel conditions (Millar et al. 2007, Carroll et al. 2010). Furthermore, understanding community responses to climate allows us to project changes in C and species dynamics.

2 Abiotic factors, such as temperature and precipitation, exert control on forest productivity

(Littell et al. 2008, Keith et al. 2009) and influence forest composition by constraining species within specific climate envelopes (Hawkins 2001, Boucher-Lalonde et al. 2012). In energy- limited systems, where low temperature limits photosynthetic activity, increasing temperature generally has a positive influence on productivity because the growing season is extended

(Beedlow et al. 2013, Anderson-Teixeira et al. 2013). In water-limited systems, the drying effect of increasing temperature generally decreases productivity, and extreme cases can cause widespread tree mortality events (Littell et al. 2010, Williams et al. 2010). Specific climatic controls on productivity can change seasonally. In systems with a non-uniform precipitation distribution, an increase in temperature during a dry period can decrease productivity, while increasing temperature during a wet period can increase productivity. The net effect of changing climate is determined, in part, by species-specific responses to these changes in resource limitation.

Maintaining a heterogeneous assemblage of species that span a wide range of responses to climate and disturbance tolerances may help maintain productivity in changing environments

(Tilman and Downing 1994). Management activities can create a mosaic of conditions on the landscape by altering both forest structure and species composition. Selectively thinning forests can reduce competition for water, nutrients, and space (Bréda et al. 1994, Roberts and Harrington

2008, Kerhoulas et al. 2013) and decrease fire risk and predisposition to insect and disease outbreaks (Chmura et al. 2011). Manipulating forest structure can increase the C sequestration rate of the remaining trees (Latham and Tappeiner 2002, Hurteau and North 2010, Martin et al.

2015) and buffer against the effects of species decline due to maladaptation to climate (Rehfeldt, et al. 2006). Quantifying forest response to the interaction between management and projected climate is necessary for informed decision-making when developing forest adaptation strategies to maintain ecosystem function and provision of ecosystem services.

3 The US Pacific Northwest (PNW) currently has some of the most C dense and productive forests in the United States (Smithwick et al. 2002, Keith et al. 2009, Hudiburg et al. 2009).

However, increasing temperature and decreasing growing season precipitation are likely to drive changes in these forests (Rupp et al. 2013, Mote et al. 2013). In the already dry summers, increasing temperature could further intensify water limitation, especially in forests that occur on the excessively well-drained glacial outwash soils of the Puget Lowlands region (Crawford and

Hall 1997, Littell et al. 2008). Douglas-fir (Pseudotsuga menziesii), a long-lived conifer of intermediate drought-tolerance, dominates much of the Puget Lowland forests, and climate change is projected to have a negative effect on this species, as moisture limitation has the largest influence on its presence (Littell et al. 2008). A reduction in Douglas-fir productivity, due to moisture limitation, may translate to a reduction in overall regional productivity, since the other common large, long-lived species, western hemlock (Tsuga heterophylla) and western redcedar

(Thuja plicata), are even less drought-tolerant (Urban et al. 1993, Burton and Cumming 1995).

However, Oregon white oak (Quercus garryana) and ponderosa pine (Pinus ponderosa) may be favored under warmer, drier conditions. These drought-tolerant species occur on the well-drained glacial outwash soils and are out-competed by encroaching Douglas-fir. Although forest models for the PNW project shifts in individual species ranges, declines in Douglas-fir, and increases in wildfire (Littell et al. 2010), there is a lack of information regarding how climate- and disturbance-driven changes in species will alter forest-level composition and productivity.

Given the high C stocks and productivity of Pacific Northwestern US forests, I sought to quantify how projected changes in climate and management would alter the composition and productivity of Puget Lowland forests. I hypothesized that at the landscape scale: 1) increasing growing season temperature and decreasing growing season precipitation would decrease forest productivity and favor less dense forests that contain drought-tolerant species such as ponderosa pine and Oregon white oak; species that occupy the historical woodland habitats in the region;

4 and 2) under projected climate, management to reduce resource competition would maintain productivity closer to current levels. I also hypothesized that under warmer and drier conditions, management actions that decreased competition for oaks would shift the declining trajectory of the species upon the landscape and that this management would not substantially decrease landscape-scale C sequestration, due to the limited extent of current oak habitats. I used a simulation approach to test these hypotheses by modeling forest growth and succession under historical climate and projected climate under two emission scenarios from two general circulation models through the late 21st century.

Methods

Study Area

Joint Base Lewis-McChord (JBLM) is a 36,182 ha military installation located in the

Puget Lowlands of western Washington. Natural areas include dense conifer forests, young mixed-conifer forests, Oregon white oak woodlands, and open grasslands (Figure 1). Conifer forests include historical forests and afforested prairies. Hardwood forests comprise only about

5% of the installation (Griffin 2007). The oak-conifer ecotone between forest and prairie supports one of three geographically isolated populations of the Washington State-listed western gray squirrel (Sciurus griseus) (Washington Department of Fish and Wildlife 2013). The base generally consists of moderate to rolling topography with elevation ranging from 0-201 m (Carey et al. 1999). The climate is mild Mediterranean, with a mean annual temperature of 9.7 °C and only 14% of the 1430 mm of mean annual precipitation falling during summer months (National

Climate Data Center (NCDC) GHCND:USC00454486). Soils on the installation are predominantly (77% of the area) excessively well-drained prairie soils (Spanaway series with

5 inclusions of Spana and Nisqually (NRCS 2013b)). The prairie soils have moderately rapid permeability with very low available water-holding capacity and support grasslands, woodlands, savanna, and Douglas-fir colonization forest habitats (Zulauf 1979). Although Douglas-fir is the dominant species, Oregon white oak and ponderosa pine are typically present on drier sites. The principal non-prairie soil is the McChord-Everett complex, a moderately well-drained soil complex that supports the historical moist forests that are dominated by Douglas-fir, with late- successional stands of western hemlock and western redcedar. Broad-leaved trees, such as bigleaf maple (Acer macrophyllum), vine maple (Acer circinatum), and Oregon ash (Fraxinus latifolia), are also present in this forest type. Glacial retreat (7-10,000 YBP) and a warmer, drier climate facilitated the establishment of the region’s grasslands (Leopold and Boyd 1999).

Although cooling temperatures favored the establishment of less drought-tolerant species, Hansen

(1947) suggested that Oregon white oak and ponderosa pine maintained a presence on the landscape due to dry summers and excessively well-drained soil. Fire is also a limitation to

Douglas-fir encroachment. Records from the 1800s documented a history of periodic burning by

Native Americans (Agee 1996). Periodic burning was largely abandoned with the advent of

European colonization in the mid-19th century (Foster and Shaff 2003). Lack of fire allowed

Douglas-fir afforestation and led to the conversion of 8221 ha of JBLM’s historical prairie to colonization forest. Much of the installation was extensively harvested prior to Department of

Defense acquisition in 1917. The Army conducted both clear-cutting and selective harvest from

1947-1952, and shifted to thinning and selection harvests in the 1960s. Current harvesting practices on the installation include variable density thinning designed to meet military training and other land management objectives.

Figure 1. Joint Base Lewis-McChord and surrounding area land cover. Study area location within Washington in red.

Field Data

Field data were collected between May-August 2012 on 346 plots distributed across the installation to obtain estimates of aboveground biomass for model validation. Sampling was stratified to capture the range of forest conditions prevalent on the installation, and areas were sampled based on accessibility given military training schedules. Prior to sampling available areas, a 100 m grid was established to locate plots. The plots consisted of a circular 1/5 ha nested design to measure all trees > 80 cm diameter at breast height (DBH). Trees > 50 cm DBH were measured in a 1/10 ha subplot and trees > 5 cm DBH were measured in a 1/50 ha subplot.

Species, DBH, and height for both live and dead trees were collected within each subplot.

7 Regeneration was tallied by height in a 2 m radius subplot at plot center. Surface fuels and coarse woody debris were measured using three 15 m modified Brown’s fuel’s transects originating from plot center (Brown 1974). Brown’s method employs a planar intersect technique where downed woody material volume is estimated from the count of woody pieces within set diameter classes as they intersect vertical sampling planes. Coarse woody debris measurements included a four decay class system, and length and end diameter measurements.

Simulation Approach

To capture forest response to changes in climate, wildfire, and management over time, I used the LANDIS-II model (Scheller et al. 2007). I chose LANDIS-II because it is species- specific, includes above- and belowground ecosystem processes, and spatially projects forest succession at the landscape-scale. This stochastic, forest disturbance and succession model uses a species-specific, age cohort-based approach to simulating forest succession, where species cohorts are represented by biomass in age classes (Scheller and Mladenoff 2004, Scheller et al.

2007). Layers of interacting grid cells containing initial species age cohorts and abiotic attribute data represent the study area. Within grid cells, species grow, compete, reproduce, and die according to user-defined life history traits (e.g. shade and fire tolerance, dispersal, sprouting, longevity), abiotic conditions, and disturbances (Scheller and Mladenoff 2004, Scheller et al.

2007). Seed dispersal and disturbances occur within and across grid cells (Scheller and

Mladenoff 2004).

In addition to the base LANDIS-II model, I used the Century Succession, Dynamic Fire and Fuels, and Leaf Biomass Harvest extensions to project C dynamics and species composition resulting from the effects of climate, management, and fire. The Century Succession extension

(Century) is based on the original CENTURY Soil model (Metherell et al. 1993, Parton et al.

8 1993). Century simulates above- and belowground C and nitrogen (N) dynamics as influenced by soil characteristics, climate, and species-specific parameters (e.g. growing degree days, C:N ratios, percent lignin, etc.); and projects productivity, C storage, and changes in forest composition (Scheller et al. 2011). I used the Dynamic Fire and Fuels extensions to simulate wildfire. The fire extension models frequency, spread, and tree mortality based on local fire weather data and fuel types derived from the Canadian Forest Fire Prediction System (Sturtevant et al. 2009, Van Wagner et al. 1992). The fuels extension assigns and annually updates fuel types based on species composition, age, and post-disturbance information (Sturtevant et al. 2009).

The Leaf Biomass Harvest extension can implement multiple overlapping, user-defined prescriptions at variable time steps, as prescription criteria are met (Gustafson et al. 2000). I used the Leaf Biomass Harvest extension to simulate both thinning and prescribed burning treatments, as the fire extension cannot implement both wildfire and prescribed fire in the same simulation.

Model Parameterization

LANDIS-II requires the subdivision of the study area into ecoregions that have similar climate and soils. Given the small elevation gradient and that climate is relatively consistent across the installation, I used soil type to classify the area into two ecoregions, based on prairie and non-prairie soils, using the Soil Survey Geographic Database (SSURGO; NRCS 2013b). I divided the study area using a 200 m grid (4 ha grid cells) to balance the resolution at which management decisions are made with computational demands. I used the Landscape Ecology,

Modeling, Mapping & Analysis (LEMMA) Laboratory’s Washington Coast and Cascades gradient nearest neighbor (GNN) interpolated map (221) and database (Ohmann and Gregory

2002, LEMMA 2006) to establish JBLM’s initial-communities layer of species with their age cohorts. The GNN technique incorporates regional grids of Forest Inventory Analysis (FIA) and

9 USFS Current Vegetation Survey (CVS) plot data, spatially-explicit environmental data (such as geology, topography, climate), and Landsat Thematic Mapper (TM) satellite imagery to predict forest composition at the landscape scale (Ohmann and Gregory 2002). The LEMMA GNN product has 30 m resolution. I used Arcmap 10.1 (ESRI 2012) to resample the GNN-based initial communities layer to decrease pixel resolution from 30 m2 to 200 m2 using a nearest neighbor resampling technique, and further reduced the number of unique communities by binning similar species composition and cohort ages. The aggregated layer included all listed GNN map species that occurred in at least 1% of the grid cells and occupied at least 10% of the basal area in a given

GNN grid cell. The initial-communities map included 21 unique, forested or able to be forested

(e.g. prairies, shrub swamps), communities and nine tree species (Table 1). These nine species represented 97% of the individual trees sampled in the field (n=4612).

The Century Succession extension requires species-specific life history characteristics, soil, and climate data. I parameterized life history and functional type data for the nine species in the initial communities layer using the scientific literature and other sources such as the

CENTURY user guide (Lee and Weber 1983, Burns and Honkala 1990a, Burns and Honkala

1990b, Habeck 1992, Metherell et al. 1993, Burton and Cumming 1995, Schowalter et al. 1998,

Thompson et al. 1999a, Thompson et al. 1999b, Balster and Marshall 2000, White et al. 2000,

Chen et al. 2001, Apple et al. 2002, Chen et al. 2002, Valachovic et al. 2004, Harlow et al. 2005,

Harrington 2006, Moore et al. 2006, Gucker 2007, Fryer 2011, Cross and Perakis 2011, Hamdan and Schmidt 2012, NRCS 2013a) (Tables 1 and 2). The soil layer, including initial C and N values, was developed using NRCS SSURGO data (NRCS 2013b). I partitioned the soil C into pools and estimated N as a function of soil C using the methodology from the Century Soil Model

(Metherell et al. 1993) (Table 3). I adjusted N deposition to reflect the 2012 N deposition for

LaGrande, Washington (NADP 2012) and biological fixation (Heath et al. 1987). Following

Loudermilk et al. (2013) and Martin et al. (2015), I calibrated soil decay parameters so that at the

10 initial time-step of the simulation, soil carbon output was similar to SSURGO values. For baseline climate, I calculated mean and standard deviation of monthly temperature and precipitation for each decade, from 51 years (1962-2012) of climate data from the Landsburg,

WA weather station (NCDC GHCND:USC00454486). LANDIS-II draws temperature and precipitation from normal distributions created using monthly mean and standard deviations

(Scheller et al. 2011).

Sexual Shade Fire Effective Max Sprout Min Max Species Longevity Maturity Tol. Tol. Disp. Disp. Prob. Sprout Sprout Sprouting Acer macrophyllum 150 20 2 2 15 120 0.7 0 100 resprout Alnus rubra 100 10 2 2 50 100 0.7 0 10 resprout Quercus garryana 500 20 2 4 6 400 0.7 0 200 resprout Pseudotsuga menziesii 750 25 3 3 100 1500 0 0 0 none Pinus ponderosa 600 16 2 5 37 120 0 0 0 none Thuja plicata 1000 25 5 2 100 122 0.5 0 200 none Tsuga heterophylla 400 30 5 1 100 1600 0.3 0 2 none Abies grandis 300 20 4 2 54 100 0 0 0 none Fraxis latifolia 150 30 3 2 5 300 0.7 0 100 resprout

11 Table 2. Century Succession species parameters for LANDIS-II. Columns are as follows: N-fix denotes nitrogen-fixing species, minimum and maximum growing degree days are GDDmin and GDDmax, Min T is minimum temperature for establishment in °C, Max Drought is proportion of growing season above which a species cannot establish if available water falls below zero, Leaf Long. is number of years for leaf persistence, Epicorm Resprout denotes species that have epicormic resprouting ability; Lignin of Leaf, Fine root, Wood, Coarse Root are the lignin percentages for each component; C:N of Coarse Root, Leaf, Fine Root, Wood, Coarse Root, Litter are the carbon to nitrogen ratios of each component.

GDD GDD Min Max Leaf Epicorm Species N-fix Min Max T Drought Long Resprout Acer macrophyllum N 900 3100 -25 0.7 1 Y Alnus rubra Y 600 2200 -24 0.8 1 Y Quercus garryana N 1400 2600 -34 0.9 1 Y Pseudotsuga menziesii N 500 2500 -37 0.8 4.8 N Pinus ponderosa N 800 3900 -41 0.9 4.5 N Thuja plicata N 500 2000 -36 0.7 8.9 N Tsuga heterophylla N 500 1900 -31 0.7 1.6 N Abies grandis N 500 2450 -9 0.8 6 N Fraxis latifolia N 150 2400 -22 0.7 1 N

Fine Coarse Fine Coarse Leaf Root Wood Root Leaf Root Wood Root Litter Species Lignin Lignin Lignin Lignin C:N C:N C:N C:N C:N Acer macrophyllum 0.192 0.224 0.25 0.26 20 30 440 90 62 Alnus rubra 0.117 0.151 0.25 0.19 23 25 50 50 24 Quercus garryana 0.176 0.22 0.14 0.26 32 63 63 62 33 Pseudotsuga menziesii 0.155 0.296 0.269 0.32 42 52 455 214 68 Pinus ponderosa 0.28 0.233 0.35 0.28 43 47 380 284 85 Thuja plicata 0.18 0.205 0.293 0.25 53 29 80 38 100 Tsuga heterophylla 0.191 0.216 0.288 0.25 46 50 380 313 37 Abies grandis 0.25 0.22 0.35 0.35 42 27 500 170 77 Fraxis latifolia 0.122 0.159 0.25 0.2 24 38 400 90 55

SOM1 C SOM1 N SOM1 SOM1 SOM2 SOM2 SOM3 SOM3 Minrl Eco Surface Surface C N C N C N N eco1 267 7 226.8 18.9 4158 207.9 3175.2 317.5 0.306 eco2 2064 52 137.2 11.4 2514.6 125.7 1920.2 192 0.476

I used the field data and genus-specific allometric equations from Jenkins et al. (2003) to estimate aboveground biomass for comparison with simulation data. To simulate 2012

12 aboveground biomass I subtracted 100 years from current species cohort ages for the initial communities layer. All cohorts less than 100 years were eliminated. To re-create a landscape comparable to 2012 conditions, the simulation included three prescriptions: a prescription that prevented establishment from all areas that were treeless in 2012; a regeneration prescription that retained only young cohorts of species present in “regrowth” areas in 2012; and a 100-year rotation thinning prescription that removed 67% of 70-180 year old Douglas-fir over 0.5-1% of the landscape annually. I excluded riparian areas and old growth stands from all prescriptions.

Using baseline climate, I ran the simulation from 1912 to 2012. From this output, I extracted

2012 aboveground biomass (AGB) values from the 5533 pixels that contained forests. Field data biomass ranged from 1510 to 82,414 g m-2, with a mean of 36,988 g m-2. Simulated biomass ranged from 416 to 65,535 g m-2, with a mean of 41,100 g m-2. Differences in initial species composition between field data and the GNN map, coupled with the stochastic nature of the model, prevent a direct comparison of AGB between methods. However, the simulated data exhibited much of the same variability as the field data, indicating that the model parameterization captured the influence of stand age and composition as well as climate and site variables on forest productivity (Figure 2).

Figure 2. Comparison of aboveground biomass (AGB) of field plot data (n=346) and LANDIS-II output of all forested pixels (n=5533). Boxes represent interquartile range.

I parameterized the Dynamic Fire extension by creating three fire regions: prairie, forests on prairie-soil, and forests on non-prairie-soil (soils described in Study Area), as defined by ecoregion and cover type. I used wildfire occurrence data from JBLM prairie fires (2007-2011) and USFS forest fires within 100 km of JBLM (1995-2005) to estimate parameters for fire size

(Table 4). To estimate frequency, I reclassified the LANDFIRE mean fire return interval data product to represent my three fire regions (Table 4, LANDFIRE 2013). Even though wildfire only plays a small part in this system (forest wildfire return interval is >75 years), I included wildfire in all climate-treatment scenarios. To parameterize fire weather (wind speed, direction, etc.) and fuel conditions, I obtained data from the Enumclaw, WA Remote Automatic Weather

Station (RAWS; 451702) and processed it with Fire Family Plus 4.1 (Bradshaw and McCormick

2000). To refine fire spread patterns, I created spatial layers for slope and aspect using the

14 Spatial Analyst extension in ArcMap (ESRI 2012) from Washington State GIS digital elevation model raster layers (http://wagda.lib.washington.edu/data/geography/wa_state/#elevation). To define fuel types, I binned general forest types that exhibited similar behavior following

Sturtevant et al. (2009) and calibrated them according to the methodology outlined in Scheller et al. (2011).

Sp Sp Sp Su Su Su Fa Fa Fa Region FMCL FMCH HProp FMCL FMCH HProp FMCL FMCH HProp Prairie Soil Forest 85 120 1 120 120 1 120 120 1 Non-prairie Soil Forest 85 120 1 120 120 1 120 120 1 Prairies 85 120 1 120 120 1 120 120 1

I parameterized timber harvest in the Leaf Biomass Harvest extension based on JBLM’s current land management practices, which sets annual harvest at approximately 40% of net annual growth (45000-54000 m3 biomass annually) (Griffin 2007). To remove this amount of biomass, I simulated an annual harvest that targeted 67% of the 70-180 year old Douglas-fir cohorts on 1% of the area on prairie soils and 0.5% of the area on non-prairie soils. To account for unintended mortality during harvesting operations, I removed 2% of all other species. I excluded stands that contained cohorts over 300 years, riparian communities, or wooded wetlands to incorporate

JBLM’s priority management goals. To increase landscape heterogeneity and maintain regeneration sources, I limited harvest to once during the simulation period in any given grid cell,

15 and did not harvest on sites that had been burned by wildfire within the last 10 years or that were adjacent to sites harvested within the last 40 years.

I used Leaf Biomass Harvest to simulate prescribed burning by implementing a thin- from-below prescription that preferentially removed young, fire-sensitive cohorts following

Syphard et al. (2011). To simulate fire-induced mortality, treatment removed 85% of 1-26 year old Douglas-fir cohorts, 90% of 1-5 year old Oregon white oak cohorts, 5% of 1-24 year old ponderosa pine cohorts, 90% of 1-15 year-old cohorts for all other species, and a smaller fraction of older cohorts. Prescribed burning treatments were limited to prairies and Douglas-fir forests on prairie soils. I did not simulate prescribed burning in areas where the initial communities contained late-successional conifers, cohorts over 300 years old, or were in riparian buffer areas.

For eligible sites that had not experienced a disturbance for 10 years, I implemented prescribed fire treatments by ranking grid cells based on the Dynamic Fuels extension fire-hazard index fuel type, with the highest ranked cells being treated first. To reflect increased canopy base height following prescribed burn treatments, I modified fuel types for a 10-year duration after application, following Martin et al. (2015). I used a 5-year fire return interval for prairie and a

20-year fire return interval for prairie colonization forests, such that 20% of qualifying prairie and

5% of qualifying forest were available to burn annually. Because I used the harvest extension to simulate prescribed burning, areas that received a prescribed burn were excluded from future thinning treatments for the length of the simulation.

To test my hypothesis that intensive management to increase Oregon white oak abundance on the landscape would be successful at shifting the declining trajectory of the species while only incurring small C costs, I used the Leaf Biomass Harvest extension to simulate an oak restoration treatment. I created a 708 ha management area that occurred on prairie soil, did not include riparian areas, and contained oaks. Within this area, I conducted intensive management that replicated thinning and burning to reduce conifer competition to a level that would allow

16 oaks to increase. My treatment thinned 85% of the non-pine conifers between 20-230 years old and included prescribed burning with a 10-year fire return interval. The entire management area was eligible for treatment annually, and management occurred on stands that were at least 11 years old and had not been harvested or burned within ten years.

Climate Data

To investigate the effects of projected changes in climate on forest dynamics, I used

NASA Earth Exchange US Downscaled Climate Projections for Pierce County, Washington

(Thrasher et al. 2013). These climate projections, downscaled from two Coupled Model

Intercomparison Project Phase 5 general circulation models (CNRM-CM5 (CNRM) National

Centre of Meteorological Research, France and CCSM4 (CCSM) National Center of Atmospheric

Research, USA), were found to best simulate regional climate (Rupp et al. 2013) (Appendices A and B). To bracket the range of potential future climate, I selected two emission scenarios

(Representative Concentration Pathways, RCPs). RCP 4.5 is a moderate scenario where greenhouse gas emissions (GHG) stabilize at 650 ppm carbon dioxide (CO2) by 2100, and RCP

8.5 is a high, or business-as-usual, scenario where GHG emissions do not stabilize by 2100 (Moss et al. 2008).

Simulation Experiment and Data Analysis

I used a full factorial design to quantify the effects of management and changing climate on C dynamics and species composition. I had a control (no management) scenario to serve as reference, and three active treatments of increasing intensity to determine level of management needed. The active management scenarios included burn only, thin only, and thin and burn

17 treatments. I implemented the burn only treatment to decrease surface fuels and reduce understory layer, the thin only treatment to decrease competition for light and other resources, and the thin and burn treatment to reduce understory and overstory competition. I ran all four treatments under current climate (baseline) and four projected climate scenarios (CNRM 4.5, 8.5;

CCSM 4.5, 8.5). I ran 25 replicates of each of the 20 treatment-climate scenarios to capture the stochastic variability of individual simulations.

To quantify treatment and climate effects on carbon pools and fluxes, I calculated net ecosystem carbon balance (NECB, net primary productivity minus heterotrophic respiration and fire emissions) and total ecosystem carbon (TEC). To quantify changes in successional dynamics, I compared aboveground carbon (AGC) for the mid-seral Douglas-fir and late- successional tree species (pooled western hemlock and western redcedar C). For each of the 20 scenarios, I calculated landscape-scale mean and 95% confidence intervals from the replicates for between-scenario comparisons of NECB, TEC, and species-specific AGC. To quantify changes in species composition, I calculated year 2100, landscape-scale mean grid-cell-level richness and

95% confidence intervals to create frequency distributions for cell-level species richness under each climate-treatment scenario.

To quantify differences in the probability of oak presence under baseline climate for each of my four treatments, I compared the empirical cumulative distribution for probability of oak on the landscape using the Kolmogorov-Smirnov non-parametric test (K-S test) (Kirkman 1996). I also used the K-S test to compare the probability of oak presence between the controls of each climate scenario against the control of the baseline climate scenario to test if climate was influential for determining the probability of oak occurrence. To evaluate the effects of treatment on the probability of oak presence under projected climate, I ran K-S tests comparing each treatment to the control within each climate scenario. I also ran K-S tests comparing the thin and burn treatment to the thin only treatment for each climate scenario. A one-sided, “greater” K-S

18 test calculates a D-statistic that determines whether the x group (e.g. control) distribution function lies above and to the left of the distribution function of the y group (e.g. treatment) [(D^+ = max[F_x(u) - F_y(u)])] (R Core Team 2013). To visualize effects of climate and management on probability of oak presence, I created year 2100 oak probability surfaces by converting cell-level oak biomass to presence/absence and calculated probability of oak occurrence from all 25 replicates of each simulation. To quantify differences in landscape-scale carbon sequestration when managing to favor oaks, I compared TEC for each thin and burn treatment-climate scenario, with and without the 708 ha oak management block being managed to favor oak. I calculated mean TEC and 95% confidence intervals for the replicate simulations.

I used ArcMap 10.1 (ESRI 2012) and R 3.0.01 (R Core Team 2013) with the ggplot2, plyr, and raster packages (Wickham 2009, Wickham 2011, Hijmans 2014) to process the simulation data and produce figures.

Data and Model Limitations

Key model limitations include tree species and spatial distribution of age cohorts at model initialization, model parameterization, and the limitations of simulating prescribed burning with the Leaf Biomass Harvest extension. The GNN map product provides a reasonable portrayal of forest composition at the landscape-scale, but the result in any given cell may not represent actual forest composition for any particular location. My model results are based on available parameterization data. Parameterization data was generally available for Douglas-fir, because it is an economically important species, however, where data were not available, parameterization came from other locations or similar species. Further empirical research will reduce parameter uncertainty. Although simulating prescribed burning through a thinning treatment removes 75% of the harvested woody vegetation and leaves 25% on site, it does not directly affect soil

19 conditions, leaf litter, or nitrogen (N) content. Furthermore, simulating fire this way does not expose mineral soil, which may result in lower regeneration for species that establish best on the bare mineral soil exposed by fire. To account for C lost as emissions from the simulated prescribed burn, at each time-step I subtracted all C removed from NECB and allowed the unburned fraction to return to the dead wood pool. This approach likely causes larger reductions in NECB than would occur with prescribed burning because combustion efficiencies are less than

100% in a prescribed burn.

Results

Climate Effects on Forest C Dynamics and Species Composition

Given the on-going prairie afforestation at JBLM, net ecosystem carbon balance (NECB) increased for the first 20 years of the simulation, regardless of climate scenario (Figure 3). As the forest matured, respiration increased and NECB began to decline under all climate scenarios. The effects of changing climate on carbon flux began at mid-century, with a substantial deviation from baseline NECB occurring under the high emission projections in late-century (Figure 3).

Differences in emission scenario and the late-century deviation in NECB are more apparent using an average of the two climate model simulations for each emission scenario (Figure 4). The late- century decline in NECB was due to decreasing NPP from both warmer temperatures and decreased summer precipitation (Appendices C and D). Under the CNRM high and the CCSM moderate emission scenarios, a large NECB decline occurred in 2060. Warm temperatures, and a dry decade followed by a wet year, caused a pulse of increased respiration (Appendices C and D).

Infrequent drying-wetting cycles can trigger decomposition in soil (Chapin et al. 2011).

20 As a result of the fire parameterization, wildfire burned less than 0.1% of forest annually in the control simulations under all climate scenarios. While the occurrence of wildfire does result in decreased NECB for the area burned, the small area burned each year had little influence on landscape scale annual NECB projections.

The influence of projected climate on species richness was driven by emission scenario, and late-century species richness frequency distributions for the baseline climate and moderate emission scenarios were similar (Figure 5). However, under both high-emission scenarios, species richness frequency distributions shifted toward more of the landscape having lower richness (Figure 5). The late-century decline in NECB under the high emission scenario was, in large part, driven by species-specific responses to changing climate. In addition to slowing afforestation, the high emission scenarios altered the transition to late-successional species. The typical trajectory for this system, under baseline climate, is a gradual decline in Douglas-fir regeneration as late-successional tree species (western hemlock and western redcedar) establish

(Figure 6). This successional pattern held under the moderate emission scenario. However,

23 under the high emission scenarios, late-century Douglas-fir AGC became asymptotic as the rate of increase of late-successional species AGC declined. The cumulative effects of late-century

AGC decline under the high emission scenarios were evident in the TEC projections (Figure 7).

Management Effects on Forest C Dynamics and Species Composition

My objective was to quantify changes in productivity and species diversity under common management regimes. By 2030 under baseline climate, all three treatments had higher

NECB than the control. Before 2030, only the thin only treatment had NECB comparable to control. Under baseline climate, the thin and burn and burn only treatments sustained NECB at a higher rate than the control, once respiration became more influential under the control scenario by approximately 2050 (Figure 8). The prescribed burn and thin and burn treatments caused an

26 early century decrease in NECB (Figure 8), because of the initial fuel loads and the C flux to the atmosphere from burning.

27 Although the control NECB had the largest decline under baseline climate, control TEC was the largest of the four management scenarios (Figure 9) because the only C losses were from wildfire. Total ecosystem carbon decreased with increasing management intensity. As expected, the thin and burn caused a larger TEC decrease than either thinning or burning alone (Figure 9).

The TEC decline in both thinning treatments was driven primarily by the harvest of Douglas-fir.

The TEC plateau that occurred under the thin only treatment was a consequence of prescription design. In the beginning of the simulation, the thin only treatment harvested the target amount of biomass by removing a set percentage of Douglas-fir cohorts. However, between 2050 and 2080, the same percentage of target cohorts had a higher total biomass, resulting in larger harvests that reduced TEC gains.

29 When I included projected climate in the simulations, treatments had higher NECB than the control after 2030, following the same general pattern as baseline climate NECB (Figure 8).

However, the influence of changing climate caused both increased inter-annual NECB variation and a late-century decline, relative to the baseline climate scenario. The influence of rising temperature and fluctuating precipitation is evident in year 2060 under the CCSM moderate emission scenario and CNRM high emission scenarios (Figures 8c and d). The overall decline of

NECB resulted from warmer, drier summers reducing productivity and causing lower TEC across management scenarios. Under the high emission scenarios, TEC became asymptotic during late century for all treatments.

Under baseline climate, late-successional species AGC continued to increase through the simulation period. I found slightly lower rates of C increase for late-successional species in simulations that included prescribed burning, which was an expected result given that these species are fire-intolerant. However, projected climate under the high emission scenario caused a decline in the rate of late-successional species AGC increase, regardless of management (Figure

10). Under the high emission scenarios, management had a negligible effect on increasing species richness, in part because the treatments did not increase late-successional species AGC.

31 Oak Restoration

Under baseline climate I found that the thin only and thin and burn treatments had a larger proportion of cells with higher year 2100 probability of oak occurrence than the control

(thin only p<0.001, thin and burn p<0.001) (Table 5, Figures 11 and 12). I found no significant difference between the thin and burn and thin only treatments, because oak regeneration is light- limited and thinning removes enough biomass to allow oak establishment (one-sided K-S test,

D=0.0034, p=0.919). This suggests that adding the burning treatment does not increase the probability of oak presence. Although I expected a warmer, drier climate to favor oaks; under the control scenario, I found no significant difference between the baseline climate control distribution and the projected climate control distributions, regardless of emission scenario (Table

6, Figure 11). I found that treatments had the same effect on the year 2100 probability of oak occurrence, regardless of climate scenario. I tested each active treatment against the control for each climate scenario and found significantly greater probability of occurrence for the thin only and thin and burn, than for the control and burn only (Table 7, Figure 11). Similar to baseline climate, there was no significant difference between probability of oak presence for the thin and burn treatment and the thin only treatment for all increased emission scenarios (one-sided K-S test, D<0.0026, p>0.952 in all scenarios). Within the intensive oak management area, the harvest prescription removed 85% of overstory conifers and reduced conifer regeneration, resulting in a substantial increase in the probability of oak presence and in overall species richness (Figure 13,

Table 8). Depending upon the climate scenario, late-century projections in the oak management area showed a 5- to 6-fold increase in oak AGC, and mean grid cell richness increased by at least one, relative to the landscape-scale thin and burn treatment that did not include oak restoration

(Table 8). Intensive treatments within the 708 ha site decreased landscape-scale TEC between

32 355-380 g C m-2 compared to the thin and burn treatment, but a faster overall growth rate minimized the difference over time (Figure 14).

35 Table 7. Kolmogrov-Smirnov one-sided test results for comparison of year 2100 probability of oak occurrence for three treatments (burn only, thin only, thin and burn) against the control under each climate scenario. Climate scenarios include projections from two general circulation models (CCSM and CNRM), driven by moderate (RCP 4.5) and high (RCP 8.5) emission scenarios. H0: No difference in distribution. Ha: the cumulative distribution function of the control lies above that of the increased emission scenario. For each treatment n=7386, number of pixels in simulation. Climate Treatment D Statistic P-value CNRM 4.5 Thin only 0.0927 <0.001 Thin and burn 0.0737 <0.001 Burn only 0.019 0.070

CCSM4.5 Thin only 0.0961 <0.001 Thin and burn 0.0639 <0.001 Burn only 0.0198 0.056

CNRM8.5 Thin only 0.0699 <0.001 Thin and burn 0.0491 <0.001 Burn only 0.0149 0.194

CCSM8.5 Thin only 0.0649 <0.001 Thin and burn 0.0426 <0.001 Burn only 0.0175 0.105

36 Table 8. Year 2100 mean oak carbon stocks and mean cell-level species richness within the 708 ha oak restoration area. Values in parentheses are standard error of the mean for each treatment (n=25 replicate simulations). Mean oak Mean cell-level Treatment Climate Carbon (g m-2) Species Richness Thin and Burn Baseline 9.34 (0.94) 3.37 (0.02) Oak Restoration Baseline 54.44 (1.42) 5.19 (0.02) Oak Restoration CNRM 4.5 51.34 (1.28) 5.09 (0.02) Oak Restoration CCSM 4.5 51.42 (1.06) 5.07 (0.02) Oak Restoration CNRM 8.5 52.3 (1.48) 4.51 (0.02) Oak Restoration CCSM 8.5 49.06 (1.0) 4.53 (0.02)

37 Discussion

Productivity and Carbon Stocks

Pacific Northwest forests west of the Cascade Range are generally considered to be energy-limited, as competition for light is a primary constraint on productivity (Littell et al.

2010). However, rising temperature in this Mediterranean climate could cause increased water stress during summer months, leading to a water limitation. I had hypothesized that a warmer, drier climate would decrease productivity and favor drought-tolerant species. I found that climate changes associated with the moderate emission scenario did not substantially deviate from the baseline climate, resulting in relatively small changes in NECB (Figure 4). However, the warmer and drier high emission scenario caused a substantial late-century NECB decline. The climate limitations on Douglas-fir slowed prairie colonization, such that prairies had not been completely afforested by late-century under the high emission scenario (Figure 5, more cells with zero species). My finding supports Littell et al. (2010) that used a climate envelope approach and found that Douglas-fir in the Puget Lowlands is at risk of unsuitable climate by mid-21st century.

However, the slower colonization of prairies facilitated a longer colonization period, and helped maintain growth during favorable climate years. At JBLM, prairie afforestation occurs patchily, over decades, and results in multi-aged Douglas-fir prairie colonization forest stands (Foster and

Shaff 2003). Climate limitations will likely decrease the establishment rate.

Under all climate scenarios, Douglas-fir and the late-successional species (western hemlock and western redcedar) proportion of total aboveground C increase slightly over time, from 95% to 97%. This suggests that more drought-tolerant species had little opportunity to compete for light, which limited their establishment and growth at rates sufficient to compensate for the climate-driven declines in productivity of the drought-intolerant species.

38 Successional Trajectory

In this energy-limited system, light availability and species-specific shade tolerance are important factors determining successional transitions over time. However, changing climate and its differential influence on species may become a determining factor for future plant community assembly in this ecosystem. The rate of change in climate water deficit (potential evapotranspiration - actual evapotranspiration, Lutz et al. 2010) has been increasing in the Pacific

Northwest (Dobrowski et al. 2013) and will continue to increase under the high emission scenario. In an empirical analysis conducted by Rehfeldt et al. (2006), ratios between summer and annual precipitation, and summer and winter temperature were the most important climate variables for determining community composition in the western US. My results from the high emission scenario suggest that water availability could become the primary factor controlling species establishment and resultant successional development. I found that, in late-century, under the warmer and drier climate of the high emission scenario, 40% of the installation had 0-2 species per cell, whereas under the baseline climate only 21% of the installation had 0-2 species per cell (Figure 5). This was partially driven by decreasing summer precipitation and increasing temperature, which caused declines in late-successional species growth (Figure 6). My findings are supported by empirical research investigating the role of moisture limitation in successional development. Previous research has found that western hemlock is particularly susceptible to drought-induced mortality during regeneration (Christy and Mack 1984), making the moisture- holding capacity of the substrate and the amount of incoming solar radiation important factors influencing western hemlock establishment (Gray and Spies 1997, Dodson et al. 2014). Although western hemlock is often the climax species in infrequent fire systems, when moisture limitation impedes western hemlock, Douglas-fir may become the climax species (Franklin and Hemstrom

1981). Given the moisture controls on late-successional species regeneration and the low water-

39 holding capacity of the prairie soils at JBLM, these areas are likely to be the first to experience a shift to a Douglas-fir climax community. This transition would likely result in a less productive forest because of climate and shade limitations on Douglas-fir regeneration (Van Tuyl et al.

2005).

Management Effects on Forest Productivity and Composition

Changing climate had a stronger effect than management on productivity, but treatments alleviated some of the NECB decline brought on by the late-century high emission scenarios.

The thinning treatment had the largest overall NECB because it reduced competition, leading to increased growth in the remaining trees. Under the thin and burn treatment, increased growth following thinning was offset by combustion from the prescribed burn. However, the large amount of biomass removed in the late-century thinning treatments reduced TEC accumulation.

My landscape-scale treatments had a negligible effect on landscape-scale species richness because only 0.5% was eligible for thinning and 5% for burning annually. If the objective is to increase species richness however, more intensive treatments can be effective. The frequent, heavy thinning under the oak restoration treatment excluded the oldest conifer cohorts and increased available light to maintain more shade-intolerant species (Table 8).

The biggest challenge to maintaining or increasing oak habitat in this system comes from competition with conifers. Gould and Harrington (2008) found that intensive management was required to reduce biomass enough to allow oak establishment and limit fire-intolerant conifer encroachment. In my 708 ha oak restoration area, initial conifer harvest followed by subsequent burning and additional harvests reduced light competition to a level that allowed oak regeneration. These treatments resulted in an initial TEC decrease and decadally thereafter, with a regime designed to maintain a higher light environment (Figure 14). I simulated prescribed

40 burning using a 10-year fire return interval, at the high end of the range suggested by Peter and

Harrington (2014) because prescribed fires can top-kill young oaks (Tveten and Fonda 1999).

My results demonstrate that implementing an intensive thin and burn treatment can increase the probability of oak presence (Figure 13). Although including prescribed burning did not increase probability of oak presence at the landscape-scale, I included it in the oak restoration treatment because it limits conifer encroachment and kills the invasive shrub, scotch broom (Cytisus scoparius L.). Increasing oak woodlands could improve western gray squirrel habitat with minimal landscape-scale C costs when small areas of the landscape are intensively managed

(Figure 14).

Managing Under Uncertainty

Given the uncertainty surrounding the sign and strength of the global forest C sink,

Bellassen and Luyssaert (2014) have suggested that the lowest risk path forward is a strategy that increases the sequestration rate and retains C in forests. I found that through 2030, both high

NECB and TEC can be achieved without management. However, the rate of C sequestration declines as the forest matures and the decline accelerates under the high emission scenario.

Because species have differential physiological and phenological responses to changing climate, diverse forests may make the forest more robust to climate-driven productivity declines and as such, acts as a risk-minimizing tactic that can be effective when the future is uncertain (Crowe and Parker 2008). My results suggest that under the high emission scenarios’ climate, JBLM’s forests will tend toward a Douglas-fir climax state, regardless of management. When the majority of forest biomass is comprised of one species, increasing summer moisture limitation under the CNRM high emission scenario (Figure 8d) has the potential to significantly impact the forest C sink strength. Diversifying species composition and structure in this forest may require

41 active management strategies that increase species richness and favor more drought-tolerant species. For this reason, efforts to increase Oregon white oak to provide western gray squirrel habitat may be useful as a strategy for hedging against climate-driven productivity decline.

My results should be considered within the context of the limitations of my approach, and the abiotic and biotic factors that influence this system that are not included in this study. I did not include increasing atmospheric CO2 in my simulations. Increasing CO2 concentrations allow leaf stomata to remain closed for longer periods, which decreases transpiration and may improve plant water-use efficiency, thereby maintaining photosynthetic rates during drier periods (Norby et al. 2005, Keenan et al. 2013). However, over time nitrogen limitation has been found to reduce the CO2 fertilization effect (Norby et al. 2010). Thus, while not including the effects of rising atmospheric CO2 in my simulations may have resulted in an underestimate of NECB during late- century under the high emission scenario; increased growth is unlikely to be sustained because nitrogen inputs in this system are relatively small. Another potential limitation of my study is that I held wildfire probability constant throughout the simulations and did not include insect and pathogen disturbance. Previous work by Littell et al. (2010) found that area burned in

Washington could experience as much as a three-fold increase by the late-21st century. However, they found that area burned in the Puget Lowlands was too low to adequately model. If area burned does increase at JBLM, I would expect a transition to more early-successional forest types and a potential opportunity for increased establishment of more drought-tolerant species. Insect and pathogen disturbances could be exacerbated by warmer, drier climate because drought- stressed individuals can be more susceptible to attack (Sturrock et al. 2011). Schmitz and Gibson

(1996) found that Douglas-fir beetle outbreaks were coincident with periods of drought. If the frequency of insect outbreaks does increase with changing climate, this could drive substantial declines in NECB. Finally, I did not account for the effects of competition from the invasive shrub Scotch broom (Cytisus scoparius L.), which occurs on approximately 5% of the base.

42 Scotch broom can form dense monocultures and reduce Douglas-fir establishment (Peterson and

Prasad 1998) and this could negatively affect post-disturbance regeneration, causing declines in

NECB.

Conclusion

Understanding how changing climate and forest composition interact and influence carbon dynamics requires a framework that addresses biotic and abiotic processes in a species- specific, spatially-dynamic framework over large scales. Using the LANDIS-II environment, I found that relatively moderate climatic changes through mid-century are unlikely to cause substantial changes in forest composition or carbon dynamics at JBLM, unlike other ecosystems where extreme drought events are already causing community composition changes (Breshears et al. 2005). However, by late-century under the high emission scenario, the increase in evaporative stress from warmer, drier summers may alter the successional trajectory from mesic conifer to a

Douglas-fir climax community, especially on the well-drained prairie soils. Drier conditions and

Douglas-fir’s inability to regenerate in its own shade could slow productivity, favoring a landscape more similar to historical conditions. In the context of future climate uncertainty and the current need to provide habitat for listed species, a strategy that produces heterogeneous ecological conditions presents the best opportunity for building an ecosystem that continues to sequester C.

43 References

Agee, J.K. 1996. Achieving conservation biology objectives with fire in the Pacific Northwest. Weed Technology 10:417-421. Anderson-Teixeira, K.J., A.D. Miller, J.E. Mohan, T.W. Hudiburg, B.D. Duval, and E.H. DeLucia. 2013. Altered dynamics of forest recovery under a changing climate. Global Change Biology 19:2001-2021. Apple, M., K. Tiekotter, M. Snow, J. Young, A. Soeldner, D. Phillips, D. Tingey, and B.J. Bond. 2002. Needle anatomy changes with increasing tree age in Douglas-fir. Tree Physiology 22:129-136. Balster, N.J., and J.D. Marshall. 2000. Decreased needle longevity of fertilized Douglas-fir and grand fir in the northern Rockies. Tree Physiology 20:1191-1197. Beedlow, P.A., E.H. Lee, D.T. Tingey, R.S. Waschmann, and C.A. Burdick. 2013. The importance of seasonal temperature and moisture patterns on growth of Douglas-fir in western Oregon, USA. Agricultural and Forest Meteorology 169:174-185. Bellassen, V., and S. Luyssaert. 2014. Managing forests in uncertain times. Nature 506:153- 155. Boucher-Lalonde, V., A. Morin, and D.J. Currie. 2012. How are tree species distributed in climatic space? A simple and general pattern. Global Ecology and Biogeography 21:1157-1166. Bradshaw, L., and E. McCormick. 2000. FireFamily Plus user's guide, Version 2.0. Gen. Tech. Rep. RMRS-GTR-67WWW. Ogden, UT: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. Bréda, N., A. Granier, and G. Aussenack. 1994. Effects of thinning on soil and tree water relations, transpiration and growth in an oak forest (Quercus petraea (Matt.) Liebl.). Tree Physiology 15:295-306. Breshears, D.D., N.S. Cobb, P.M. Rich, K.P. Price, C.D. Allen, R.G. Balice, W.H. Romme, J.H. Kastens, M. L. Floyd, J. Belnap, J.J. Anderson, O.B. Myers, and C.W. Meyer. 2005. Regional vegetation die-off in response to global-change-type drought. Proceedings of the National Academy of Sciences of the United States of America 102:15144-15148. Brown, J.K. 1974. Handbook for inventorying downed woody material. USDA Forest Service General Technical Report INT-16. Ogden, UT. Burns R.M., and B.H. Honkala. 1990a. Silvics of North America. Vol.1. Conifers. Agricultural Handbook no. 654. U.S. Department of Agriculture. Washington, D.C., USA. Burns R.M., and B.H. Honkala. 1990b. Silvics of North America. Vol.2. Hardwoods. Agricultural Handbook no. 654. U.S. Department of Agriculture. Washington, D.C., USA. Burton, P.J., and S.G. Cumming. 1995. Potential effects of climatic change on some western Canadian forests, based on phenological enhancements to a patch model of forest succession. Water, Air and Soil Pollution 82:401-414. Carey, A.B., D.R. Thysell, and A.W. Brodie. 1999. The forest ecosystem study: background, rationale, implementation, baseline conditions, and silvicultural assessment. USFS PNW-GTR-457. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon, USA.

44 Carroll, C., J.R. Dunk, and A. Moilanen. 2010. Optimizing resiliency of reserve networks to climate change: multispecies conservation planning in the Pacific Northwest, USA. Global Change Biology 16:891-904. Chapin, F.S., P.A. Matson, and P.M. Vitousek. 2011. Principles of terrestrial ecosystem ecology 2nd edition. Springer New York. p. 200. Chen, H., M.E. Harmon, and R.P. Griffiths. 2001. Decomposition and nitrogen release from decomposing woody roots in coniferous forests of the Pacific Northwest: a chronosequence approach. Canadian Journal of Forest Research 31:246-260. Chen, H., M.E. Harmon, J. Sexton, and B. Fasth. 2002. Fine-root decomposition and N dynamics in coniferous forests of the Pacific Northwest, U.S.A. Canadian Journal of Forest Research 32:320-331. Chmura D.J., P.D. Anderson, G.T. Howe, C.A. Harrington, J.E. Halofsky, D.L. Peterson, D.C. Shaw, and J.B. St.Clair. 2011. Forest responses to climate change in the northwestern United States: ecophysical foundations for adaptive management. Forest Ecology and Management 261:1121-1142. Christy, E.J., and R.N. Mack. 1984. Variation in demography of juvenile Tsuga heterophylla across the substratum mosaic. Journal of Ecology 72:75-91. Coops, N.C., and R.H. Waring. 2011. Estimating the vulnerability of fifteen tree species under changing climate in northwest North America. Ecological Modelling 222:2119-2129. Crawford, R., and H. Hall. 1997. Changes in the south Puget Prairie Landscape. Chapter 2 in P. Dunn. and K. Ewing, editors. Ecology and Conservation of the South Puget Sound Prairie Landscape. The Nature Conservancy, Seattle, Washington. Creutzberg, M.K., R.M. Scheller, M.S. Lucash, L.B. Evers, S.D. Leduc, and M.G. Johnson. In press. Bioenergy harvest, climate change, and forest carbon in the Oregon Coast Range. Global Change Biology Bioenergy. Cross, A., and S.S. Perakis. 2011. Tree species and soil nutrient profiles in old-growth forests of the Oregon coast range. Canadian Journal of Forest Research 41:195-210. Crowe, K.A., and W.H. Parker. 2008. Using portfolio theory to guide reforestation and restoration under climate change scenarios. Climatic Change 89:355-370. Dobrowski, S.Z., J. Abatzoglou, A.K. Swanson, J.A. Greenberg, A.R. Mynsberge, Z.A. Holden, and M.K. Schwartz. 2013. The climate velocity of the contiguous United States during the 20th century. Global Change Biology 19:241-251. Dodson, E.K., J.I. Burton, and K.J. Puettmann. 2014. Multiscale controls on natural regeneration dynamics after partial overstory removal in Douglas-fir forests in western Oregon, USA. Forest Science 60(2):1-8. ESRI. 2012. ArcMap desktop. Release 10.1. Environmental Systems Research Institute, Redlands, CA, USA. Flower, C.E., and M.A. Gonzalez-Meler. In press. Responses of temperate forest productivity to insect and pathogen disturbances. Annual Review of Plant Biology 66(12):1-23. Foster, J.R., and S.E. Shaff. 2003. Forest colonization of Puget Lowland grasslands at Fort Lewis, Washington. Northwest Science 77(4):283-296. Franklin, J.F., and M.A. Hemstrom. 1981. Aspects of succession in the coniferous forests of the Pacific Northwest. Chapter 14 in D.C. West, H.H. Shugart, and D.B. Botkin, editors. Forest succession: concepts and application. Springer-Verlag. New York, New York, USA.

45 Franklin, J.F., T.A. Spies, R. van Pelt, A.B. Carey, D.A. Thornburgh, D.R. Berg, D.B. Lindenmayer, M.E. Harmon, W.S. Keeton, D.C. Shaw, K. Bible, and J. Chen. 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. Fryer, J L. 2011. Acer macrophyllum. In: Fire Effects Information System, U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory. Available at http://www.fs.fed.us/database/feis/ . Accessed 22 July 2013. Gould, P.J., and C.A. Harrington. 2008. Evaluation of landscape alternatives for managing oak at Tenalquot Prairie, Washington. PNW-GTR-745. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon. Gray, A.N., and T.A. Spies. 1997. Microsite controls on tree seedling establishment in conifer forest canopy gaps. Ecology 78:2458-2473. Griffin, S. 2007. Forest management and stump-to-forest gate chain-of-custody certification evaluation report for the forestlands managed by the forestry branch, Fort Lewis Military Installation, Department of Defense. Scientific Certification Systems, Emeryville, CA. Gucker, Corey L. 2007. Quercus garryana. In: Fire Effects Information System. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory (Producer). Available at http://www.fs.fed.us/database/feis/ Accessed 22 July 2013. Gustafson, E.J., S.R. Shifley, D.J. Mladenoff, K.K. Nimerfro, and H.S. He. 2000. Spatial simulation of forest succession and timber harvesting using LANDIS. Canadian Journal of Forest Research 30:32-43. Habeck, R. J. 1992. Pinus ponderosa var. ponderosa. In: Fire Effects Information System. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory. Available at http://www.fe.fed.us/database/feis/ . Accessed 22 July 2013. Hamdan, K., and M. Schmidt. 2012. The influence of bigleaf maple on chemical properties of throughfall, stemflow, and forest floor in coniferous forest in the Pacific Northwest. Canadian Journal of Forest Research 42:868-878. Hansen, H.P. 1947. Climate versus fire and soil as factors in postglacial forest succession in the Puget Lowland of Washington. American Journal of Science 245:265-286. Harlow, B.A., R.A. Duursma, and J.D. Marshall. 2005. Leaf longevity of western red cedar (Thuja plicata) increases with depth in the canopy. Tree Physiology 25:557-562. Harmon, M.E., J.F. Franklin, F.J. Swanson, P. Sollins, S.V. Gregory, J.D. Lattin, N.H. Anderson, S.P. Cline, N.G. Aumen, J.R. Sedell, G.W. Lienkaemper, K. Cromack Jr. and K.W. Cummins. 1986. Ecology of coarse woody debris in temperate ecosystems. Advances in Ecological Research 15: 133-302. Harrington, C.A. 2006. Biology and ecology of red alder. Chapter 2 in R.L. Deal and C.A. Harrington, editors. Red Alder: A State of Knowledge. PNW-GTR-669. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland Oregon. Hawkins, B.A. 2001. Ecology’s oldest pattern? TRENDS in Ecology and Evolution 16(8):470. Heath, B., P. Sollins, D.A. Perry, and K. Cromack, Jr. 1987. Asymbiotic nitrogen fixation in litter from Pacific Northwest forests. Canadian Journal of Forest Research 18:68-74.

46 Hijmans, R. J. 2014. Raster: Geographic data analysis and modeling. Version 2.2-12. http://CRAN.R-project.org/package=raster Hudiburg, T, B. Law, D.P. Turner, J. Campbell, D. Donato, and M. Duane. 2009. Carbon dynamics of Oregon and Northern California forests and potential land-based carbon storage. Ecological Applications 19(1):163-180. Hurteau, M. and M. North. 2010. Carbon recovery rates following different wildfire risk mitigation treatments. Forest Ecology and Management 260(5):930-937. Jenkins, J.C., D.C. Chojnacky, L.S. Heath, and R.A. Birdsey. 2003. National-scale biomass estimators for United States tree species. Forest Science 1:12-35. Kashian, D.M., W.H. Romme, D.B. Tinker, M.G. Turner, and M.G. Ryan. 2006. Carbon storage on landscapes with stand-replacing fires. BioScience 56(7):598-606. Keenan, T.F, D.Y. Hollinger, G. Bohrer, D. Dragoni, J.W. Munger, H.P. Schmid, and A.D. Richardson. 2013. Increase in forest water-use efficiency as atmospheric carbon dioxide concentrations rise. Nature 499:324-328. Keith, H., B.G. Macke, and D.B. Lindenmayer. 2009. Re-evaluation of forest biomass carbon stocks and lessons from the world’s most carbon-dense forests. Proceedings of the National Academy of Sciences of the United States of America 106(28):11635-11640. Kerhoulas, L.P., T.E. Kolb, M.D. Hurteau, and G.W. Koch. 2013. Managing climate change adaptation in forests: a case study from the U.S. Southwest. Journal of Applied Ecology 50:1311-1320. Kirkman, T.W. 1996. Statistics to use. Available online http://www.physics.csbsju.edu/stats/ Accessed 1March 2015. LANDFIRE. 2013. LANDFIRE Mean Fire Return Interval layer. U.S. Department of Interior, Geological Survey. Available online http://landfire.cr.usgs.gov/viewer/ . Accessed 1 October 2014. Latham, P., and J. Tappeiner. 2002. Response of old-growth conifers to reduction in stand density in western Oregon forests. Tree Physiology 22:137:146. Lee, J.J., and D.E. Weber. 1983. Effects of sulfuric acid rain on decomposition rate and chemical element content of hardwood leaf litter. Canadian Journal of Botany 61(3):872- 879. LEMMA. Landscape Ecology, Modeling, Mapping & Analysis. Available online at http://lemma.forestry.oregonstate.edu/data/structure-maps. Accessed 4 April 2013. Leopold, E.B., and R. Boyd. 1999. An ecological history of old prairie areas in southwestern Washington in R. boyd, editor, Indians, fires and the Land in the Pacific Northwest. Oregon State University Press, Corvallis pp 139-151. Littell, J.S., D.L. Peterson, and M. Tjoelker. 2008. Douglas-fir growth in mountain ecosystems: water limits tree growth from stand to region. Ecological Monographs 78(3):349-368. Littell, J.S., E.E. Oneil, D. McKenzie, J.A. Hicke, J.A. Lutz, R.A. Norheim, and M.M. Esner. 2010. Forest ecosystems, disturbance, and climatic change in Washington State, USA. Climatic Change 102:129-158. Loudermilk, E.L., R.M. Scheller, P.J. Weisberg, J. Yang, T.E. Dilts, S.L. Karam, and C. Skinner. 2013. Carbon dynamics in the future forest: the importance of long-term successional legacy and climate-fire interaction. Global Change Biology 19:3502-3515. Martin, K.M., M.D. Hurteau, B.A. Hungate, G.W. Koch, and M.P. North. 2015. Carbon tradeoffs of restoration and provision of endangered species habitat in a fire-maintained

47 forest. Ecosystems 18:76-88.Messier, C., and K.J. Puettmann. 2011. Forests as complex adaptive systems: implications for forest management and modeling. Italian Journal of Forest and Mountain Environments 66:249-258. Metherell A.K., L.A. Harding, C.V. Cole, and W.J. Parton. 1993. CENTURY soil organic matter model environment technical documentation. Agroecosystem Version 4.0. Fort Collins, CO:USDA-ARS. Millar, C.I., N.L. Stephenson, S.L. Stephens. 2007. Climate change and the forests of the future: managing in the face of uncertainty. Ecological Applications 17(8):2145-2151. Moore, T.R., J.A. Trofymow, C.E. Prescott, J. Fyles, and B.D. Titus, CIDET Working Group. 2006. Patterns of carbon, nitrogen, and phosphorus dynamics in decomposing foliar litter in Canadian forests. Ecosystems 9:46-62. Moss, R., M. Babiker, S. Brinkman, E. Calvo, T. Carter, J. Edmonds, I. Elgizouli, S. Emori, L. Erda, K. Hibbard, R. Jones, M. Kainuma, J. Kelleher, J. F. Lamarque, M. Manning, B. Matthews, J. Meehl, L. Meyer, J. Mitchell, N. Nakicenovic, B. O’Neill, R. Pichs, K. Riahi, S. Rose, P. Runci, R. Stouffer, D. van Vuuren, J. Weyant, T. Wilbanks, J. P. van Ypersele, and M. Zurek. 2008. Towards new scenarios for analysis of emissions, climate change, impacts, and response strategies. Technical Summary. Intergovernmental Panel on Climate Change, Geneva 25 pp. Mote, P.W., J.T. Abatzoglou, and K.E. Kunkel. 2013. Climate. Chapter 2 in Dalton, M.M, P.W. Mote, and A.K. Snover, editors. Climate change in the Northwest: implications for our landscapes, waters, and communities. Island Press, Washington, DC. NADP. 2012. National Atmospheric Deposition Program/National Trends Network 2012 Annual & seasonal data summary for site WA21 Part 1: summary of sample validity and completeness criteria. NADP Program Office, Illinois State Water Survey, 2204 Griffith Dr., Champaign, IL. Available online at nadp.sws.uiuc.deu/NTN/ntnData.aspx . Accessed 1 April2014. NCDC. 2013. National Oceanic and Atmospheric Administration, National Climatic Data Center. Available online at www.ncdc.noaa.gov. Accessed 1 May 2013. Norby, R.J., E.H. DeLucia, B. Glelen, C. Calfapietra, C.P. Giardina, J.S. King, J. Ledford, H.R. McCarthy, D.J.P. Moore, R. Ceulemans, P. DeAngelis, A.C. Finzi, W.H. Schlesinger, and R. Oren. 2005. Forest response to elevated CO2 is conserved across a broad range of productivity. Proceedings of the National Academy of Sciences of the United States of America 102(50):18052-18056.

Norby, R.J., J.M. Warren, C.M. Iversen, B.E. Medlyn, and R.E. McMurtrie. 2010. CO2 enhancement of forest productivity constrained by limited nitrogen availability. Proceedings of the National Academy of Sciences of the United States of America 107(45):19368-19373. NRCS. 2013a. The PLANTS Database. U.S. Department of Agriculture, Natural Resources Conservation Service. Greensboro, NC. Available online http://plants.usda.gov. Accessed 1 January 2014 NRCS. 2013b. Web Soil Survey. U.S. Department of Agriculture, Natural Resources Conservation Service. Available online at http://websoilsurvey.nrcs.usda.gov. Accessed 1 October 2013. Ohmann, J.L., and M.J. Gregory. 2002. Predictive mapping of forest composition and structure with direct gradient analysis and nearest-neighbor imputation in coastal Oregon, U.S.A. Canadian Journal of Forest Research 32:725-741

48 Parton, W.J., J.M.O. Scurlock, D.S. Ojima, T.G. Gilmanov, R.J. Scholes, D.S. Schimel, T. Kirchener, J-C. Menaut, T. Seastedt, E. G. Moya, A. Kamnalrut, and J.I. Kinyamario. 1993. Observations and modeling of biomass and soil organic matter dynamics for the grassland biome worldwide. Global Biogeochemical Cycles 7(4):785-803. Peter, D.H., and T.B. Harrington. 2014. Historical colonization of South Puget Sound prairies by Douglas-fir at Joint Base Lewis-McChord, Washington. Northwest Science 88(3):186- 205. Peterson, D.J., and R. Prasad. 1998. The biology of Canadian weeds. 109. Cytisus scoparius (L.) Link. Canadian Journal of Plant Science 78:497-504. R Core Team. 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org/ . Rehfeldt, G.E., N.L. Crookston, M.V. Warwell, and J.S. Evans. 2006. Empirical analyses of plant-climate relationships for the western United States. International Journal of Plant Science 167(6):1123-1150. Roberts, S.D., and C.A. Harrington. 2008. Individual tree growth response to variable-density thinning in coastal Pacific Northwest forests. Forest Ecology and Management 255:2771-2781. Rupp, D.E., J.T. Abatzoglou, K.C. Hegewisch, and P.W. Mote. 2013. Evaluation of CMIP5 20th century climate simulations for the Pacific Northwest USA. Journal of Geophysical Research: Atmospheres 118:10884-10906. Scheller, R.M., J.B. Domingo, B.R. Sturtevant, J.S. Williams, A. Rudy, E.J. Gustafson, and D.J. Mladenoff. 2007. Design, development, and application of LANDIS-II, a spatial landscape simulation model with flexible temporal and spatial resolution. Ecological Modelling 201:409-419. Scheller, R.M., D. Hua, P.V. Bolstad, R.A. Birdsey, and D.J. Mladenoff. 2011. The effects of forest harvest intensity in combination with wind disturbance on carbon dynamics in Lake States Mesic Forests. Ecological Modelling 222: 144-153. Scheller, R.M., and D.J. Mladenoff. 2004. A forest growth and biomass module for a landscape simulation model, LANDIS: design, validation, and application. Ecological Modelling 180: 211-229. Schmitz, R.F., and K.E. Gibson. 1996. Douglas-fir beetle. Forest Insect & Disease Leaflet 5 U. S. Department of Agriculture. Washington, D.C. 8 p. Schowalter, T.D, Y.L. Zhang, T.E. Sabin. 1998. Decomposition and nutrient dynamics of oak Quercus spp. Logs after five years of decomposition. Ecography 21:3-10. Smithwick, E.A.H., M.E. Harmon, S.M. Remillard, S.A. Acker, and J.F. Franklin. 2002. Potential upper bounds of carbon stores in forests of the Pacific Northwest. Ecological Applications 12(5):1303-1317. Spies, T.A., T.W. Giesen, F.J. Swanson, J.F. Franklin, D. Lach, and K.N. Johnson. 2010. Climate change adaptation strategies for federal forests of the Pacific Northwest, USA: ecological, policy, and socio-economic perspectives. Landscape Ecology 25:1185-1199. Sturrock, R.N., S.J. Frankel, A.V. Brown, P.E. Hennon, J.T. Kliejunas, K.J. Lewis, J.J. Worrall, A.J. Woods. 2011. Climate change and forest diseases. Plant Pathology 60:133-149. Sturtevant, B.R., R.M. Scheller, B.R. Miranda, D. Shinneman, and A. Syphard. 2009. Simulating dynamic and mixed-severity fire regimes: a process-based fire extension for LANDIS-II. Ecological Modelling 220:3380-3393.

49 Syphard, A.D., R.M. Scheller, B.C. Ward, W.D. Spencer, and J.R. Strittholt. 2011. Simulating landscape-scale effects of fuels treatments in the Sierra Nevada, CA, USA. International Journal of Wildland Fire 20:364-383. Thompson, R.S., K.H. Anderson, and P.J. Bartlein. 1999a. Atlas of relations between climatic parameters and distributions of important trees and shrubs in North America – Introductions and conifers: U.S. Geological Survey Professional Paper 1650-A. Available online at http://pubs.usgs.gov/pp/p1650-a/. Thompson, R.S., K.H. Anderson, and P.J. Bartlein. 1999b. Atlas of relations between climatic parameters and distributions of important trees and shrubs in North America – Hardwoods: U.S.Geological Survey professional Paper 1650-B. Available online at http://pubs.usgs.gov/pp/p1650-b/. Thrasher, B., J. Xiong, W. Wang, F. Melton, A. Michaelis, and R. Nemani. 2013. New downscaled climate projections suitable for resource management in the U.S. Eos, Transactions American Geophysical Union 94:321-323. Available online at https://portal.nccs.nasa.gov/portal_home/published/NEX.html. Accessed 1 December 2014. Tilman, D., and J.A. Downing. 1994. Biodiversity and stability in grasslands. Nature 367:363- 365. Tveten, R.K., and R.W. Fonda. 1999. Fire effects on prairies and oak woodlands on Fort Lewis, Washington. Northwest Science 73(3):145-158. Urban, D.L., M.E. Harmon, and C.B. Halpern. 1993. Potential response of Pacific Northwestern forests to climatic change, effects of stand age and initial composition. Climatic Change23:247-266. Valachovic, Y.S., B.A. Caldwell, K. Cromack Jr., and R.P. Griffiths. 2004. Leaf litter chemistry controls on decomposition of Pacific Northwest trees and woody shrubs. Canadian Journal of Forest Research 23:2131-2147. Van Tuyl, S., B.E. Law, D.P. Turner, and A.I. Gitelman. 2005. Variability in net primary production and carbon storage in biomass across Oregon forests—an assessment integrating data from forest inventories, intensive sites, and remote sensing. Forest Ecology and Management 209:273-291. Van Wagner, C.E., B.J. Stocks, B.D. Lawson, M.E. Alexander, T.J. Lynham, and R.S. McAlpine. 1992. Development and structure of the Canadian forest fire behavior prediction system. Fire Danger Group, Forestry Canada, Ottawa, Ontario. Walther, G., E. Post, P. Convey, A. Menzel, C. Parmesan, T.J.C. Beebee, J. Fromentin, O. Hoegh-Guldberg, and F. Bairlein. 2002. Ecological responses to recent climate change. Nature 416:398-395. Washington Department of Fish and Wildlife. 2013. Threatened and endangered wildlife in Washington: 2012 annual report. Listing and Recovery Section, Wildlife Program, Washington Department of Fish and Wildlife, Olympia, Washington 251 p. White, M.A., P.E. Thornton, S.W. Running, and R.R. Nemani. 2000. Parameterization and sensitivity analysis of the BIOME-BGC terrestrial ecosystem model: net primary production controls. Earth Interactions 4:1:85. Wickham, H. 2009. Ggplot2: elegant graphics for data analysis. Springer, New York, New York. Wickham, H. 2011. The split-apply-combine strategy for data analysis. Journal of Statistical Software, 40(1):1-29. http://www.jstatsoft.org/v40/i01/.

50 Williams, A.P., C.D. Allen, C.I. Millar, T.W. Swetnam, J. Michaelsen, C.J. Still, and S.W. Leavitt. 2010. Forest responses to increasing aridity and warmth in the southwestern United States. Proceedings of the National Academy of Sciences of the United States of America 107(50):21289-21294. Zulauf, A.S. 1979. Soil survey of Pierce County area, Washington. U.S. Department of Agriculture, Natural Resources Conservation Service.

Appendix A

Annual projected temperature data.

Projected mean annual temperature in degrees Celsius under two general circulation models (CCSM and CNRM), driven by two emission scenarios (moderate (RCP 4.5) and high (RCP 8.5) emissions).

Appendix B

Annual projected precipitation data.

Projected mean annual precipitation (cm) under two general circulation models (CCSM and CNRM), driven by two emission scenarios (moderate (RCP 4.5) and high (RCP 8.5) emissions).

Appendix C

LANDIS-II decadal summer (June, July, August) precipitation projections.

Projected decadal mean and standard deviation summer precipitation (June, July, August) under two general circulation models (CCSM and CNRM), driven by two emission scenarios (moderate (RCP 4.5) and high (RCP 8.5) emissions). .

Appendix D

LANDIS-II decadal temperature projections.

Projected decadal mean and standard deviation temperature under two general circulation models (CCSM and CNRM), driven by two emission scenarios (moderate (RCP 4.5) and high (RCP 8.5) emissions).