A Multi-Scale Analysis of the Potential Impacts of Rapid Climate Change on Forest Lands Managed by the Department of Defense in the United States

A Multi-scale Analysis of the Potential Impacts of Rapid Climate Change on Forest Lands Managed by the Department of Defense in the United States

Richard H. Odom, Jr.

Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of

Doctor of Philosophy In Geospatial and Environmental Analysis

Advisory Committee

W. Mark Ford, Co-Chair Stephen P. Prisley, Co-Chair Lynn M. Resler Nathan R. Beane

October 17, 2018

Blacksburg, VA

Keywords: climate change, landscape modeling, geospatial analysis, military installations, ecological site classification, Landis-II, Holdridge Life Zones

A Multi-scale Analysis of the Potential Impacts of Rapid Climate Change on Forest Lands Managed by the Department of Defense in the United States

Richard H. Odom, Jr.

ABSTRACT

Based on current projections from global climate models (GCM’s), regional climates in the coterminous U.S. are expected to become warmer and either wetter or drier over the next century depending on the GCM used to make projections. Forest communities and the species that comprise them are likely to respond to a changing climate in a number of different ways based on environmental tolerances that have evolved over the past several thousand years. If, as many scientists believe, global warming is occurring at a rate that is unique in the recent history of the Earth, many species and plant communities are likely to be stressed by temperature and moisture conditions unlike those in which they have evolved. Concurrently, some species and communities in boreal and cold temperate biomes may benefit from warmer temperatures and greater CO2 availability resulting in more successful reproduction, higher growth rates and increased competitiveness. Plant species and communities are likely to respond differently to climate change on different landscapes and at different scales, and therefore a multi-scale, ecoregional approach will be required to understand potential impacts of climate change on species, communities and entire ecosystems. This study is part of a broader effort by the U.S. Department of Defense to assess the vulnerability of military lands to rapid climate change and develop mitigation strategies to cope with projected impacts to natural systems, resource management activities and military missions.

The Holdridge Life Zone system was used to model the geographic extent of present and future climatic envelopes that influence the distribution of forest biomes and tree species in the coterminous U.S. The Holdridge system integrates mean annual temperature, mean annual precipitation and mean annual potential evapotranspiration to define bioclimatic life zones that are strongly correlated with the spatial distribution of major forest cover types and tree species distributions. Climate projections were based on an ensemble of 16 GCM’s and three future greenhouse gas emissions scenarios (low-B1, moderate-A1B and high-A2). Changes in the extent and location of Holdridge life zones over approximately 80 years were analyzed and

results interpreted in terms of potential impacts to forest tree species and major forest cover types. The magnitude of change from historic conditions also was evaluated for 663 U.S. military installations to aid in the development of vulnerability metrics for Department of Defense facilities and to better understand potential climate trajectories for different regions of the country. Cluster analysis was used to group installations on a regional basis and regional variation in projected climate conditions and assessed relative to important resource management issues at representative installations.

Forest cover was modeled at Ft. Drum, New York to illustrate potential changes in species composition and cover type distribution at a landscape scale under future climate change scenarios. Stand ages were estimated using data on site index trees available in the Forest Inventory and Analysis (FIA) database for New York. Ecological types were developed from large scale soil survey data (Natural Resource Conservation Service, Soil Survey Geographic Database, SSURGO) and stand-level forest inventory data available from the natural resources program at Ft. Drum. Stand age, ecological type, species life histories and soil properties were used to parameterize a stochastic forest landscape simulation model using the LANDIS-II application and project changes over 80 years under three future CO2 emissions scenarios. Results showed that there is potential for significant changes in the distribution of some tree species and forest cover types at Ft. Drum under the warmer climate conditions projected for the northeastern U.S. Cover types characterized by species at the northern end of their ranges (e.g., species associated with oak (Quercus rubra, Q. alba)-hickory (Carya cordiformis) forest) increased in abundance, especially on more xeric sites such as sand plains and convex landforms covered in coarse-textured glacial till. However, boreal and cool temperate species, such as sugar maple (Acer saccharum), yellow birch (Betula alleghaniensis), aspens (Populus tremuloides, P. grandidentata) and eastern hemlock (Tsuga canadensis) that are major current components of the northern hardwood-hemlock cover type therein, were projected to remain significant components of the Ft. Drum landscape late into the century on all but the most xeric sites. Overall, changes in species composition were less dramatic than expected at a landscape scale and highly sensitive to establishment probabilities related to specific site characteristics (e.g., soil texture and drainage). The lack of a strong climate response at Ft. Drum may be due to the

presence of a number of widely distributed tree species with presumed large climatic tolerances and the relatively homogeneous biophysical conditions that exist within this landscape.

A Multi-scale Analysis of the Potential Impacts of Rapid Climate Change on Forest Lands Managed by the Department of Defense in the United States

Richard H. Odom, Jr.

ABSTRACT (General)

The Holdridge Life Zone system was used to model the geographic extent of present and future climates that influence the distribution of forest biomes and tree species in the coterminous U.S. The Holdridge system integrates mean annual temperature, mean annual precipitation and mean annual potential evapotranspiration to define bioclimatic life zones that are strongly correlated with the spatial distribution of major forest cover types and tree species distributions. Climate projections were based on an ensemble of 16 GCM’s and three future greenhouse gas emissions scenarios (low-B1, moderate-A1B and high-A2). Changes in the extent and location of Holdridge life zones over approximately 80 years were analyzed and results interpreted in terms of potential impacts to forest tree species and major forest cover types. The magnitude of change from historic conditions also was evaluated for 663 U.S. military installations to aid in the development of vulnerability metrics for Department of Defense facilities and to better understand potential climate trajectories for different regions of the country.

Forest cover was modeled at Ft. Drum, New York to illustrate potential changes in species composition and cover type distribution at a landscape scale under future climate change scenarios. Results suggest that there is potential for significant changes in the distribution of some tree species and forest cover types at Ft. Drum over the next 50 to 100 years under warmer climate conditions projected for the northeastern U.S. Warm temperate tree species at the northern end of their ranges (e.g., oaks, hickories) increased in abundance, especially on more xeric sites such as sand plains and convex landforms covered in coarse-textured glacial till. However, boreal and cool temperate species, such as sugar maple, yellow birch, aspens and eastern hemlock were projected to remain significant components of the Ft. Drum landscape late into the century on all but the most xeric sites. Overall, changes in species composition were less dramatic than expected at a landscape scale and highly sensitive to establishment probabilities related to specific site characteristics (e.g., soil texture and drainage).

Acknowledgements

I would like to acknowledge the generous financial support of the U.S. Army Corps of Engineers, Engineering Research and Development Center, Environmental Lab, Vicksburg, Mississippi for this project.

Invaluable support was also provided by staff with the U.S. Geological Survey, Virginia Cooperative Fish and Wildlife Research Unit and the Virginia Tech Department of Fish and Wildlife Conservation.

The Natural Resources staff at Fort Drum, New York provided extensive forest inventory and geospatial data in support of the project.

Finally, I greatly appreciate the advice and patience of my doctoral committee.

Dedication

This dissertation is dedicated to my wife Pam. Without her support, encouragement and patience over the past seven years this achievement would not have been possible.

Table of Contents

List of tables and figures viii Project overview xvi

Chapter 1 – Assessing the Vulnerability of Military Installations in the Coterminous United States to Potential Biome Shifts resulting from Rapid Climate Change Abstract 1 Introduction 2 Methods 4 Results 11 Discussion and Conclusions 15 References 33 Appendices 51 Tables and Figures 55 Chapter 2 - Developing Species-Age Cohorts from Forest Inventory and Analysis Data to Parameterize a Forest Landscape Model Abstract 69 Introduction 70 Methods 75 Results 82 Discussion and Conclusions 84 References 89 Appendices 98 Tables and Figures 101 Chapter 3 – Simulated Effects of Climate Change on Soil Moisture Deficits, Species Distributions and Biomass in a Northern Hardwood Forest Abstract 110 Introduction 111 Methods 115 Results 123 Discussion and Conclusions 127 References 134 Appendices 147 Tables and Figures 154 vii

List of tables and figures. Page

Chapter 1 Table 1. Area of six largest Holdridge life zones in the coterminous U.S. based on mean 55 climate conditions from 1951-2006.

Table 2. Change in location and area of Holdridge life zones in eastern North America 56 from 2006 to 2085 under the A2 emissions scenario.

Table 3. Percent change in mean annual temperature and mean annual precipitation 57 projected from 2006 to 2085 for five groups of military installations identified through k-means cluster analysis. Cluster numbers correspond to Figure 7a, b and d.

Figure 1. Diagram illustrating Holdridge life zones and bioclimatic components along 58 latitudinal and altitudinal gradients (Halasz 2007; original image published at: http://en. wikipedia.org /wiki/ Image:Lifezones_Pengo.svg. This file is licensed under the Creative Commons Attribution ShareAlike license versions 2.5, 2.0, and 1.0).

Figure 2. Cartographic model illustrating the process of developing Holdridge life zone 59 maps from projected climate data based on average values from an ensemble of 16 global climate models (Girvetz et al. 2009). Example shown is for the B1 emissions scenario (IPCC 2007) for years 2055 (T1submodel) and 2085 (T2 submodel). Model created in ArcGIS 10.2.1 ModelBuilder (ESRI 2014).

Figure 3. Percent change in mean annual temperature, mean annual precipitation and 60 mean annual potential evapotranspiration ratio from 2006 to 2085 in the coterminous U.S. based on projections from an ensemble of 16 global climate models under a relatively high CO2 emissions scenario (A2).

Figure 4. Maps of Holdridge life zones for the coterminous U.S. based on historical and 61 projected climatic means (all zones are not labeled). Source data from the ClimateWizard tool, http://www.climatewizard.org/ (Girvetz et al. 2009).

Figure 5. Holdridge life zones and representative vegetation types for the eastern U.S. 62 based on historical climate conditions (1951-2006) and two future climate projections

viii

under different emissions scenarios (low-B1 and high-A2, IPCC 2007). Projections are based on median values derived from an ensemble of 16 GCMs. Black dots are locations of U.S. military installations (location of Ft. Drum, New York highlighted in cyan color).

Figure 6. Comparison of Holdridge life zones (color-shaded areas) developed from 63 historical climate data with major forest formations (a) and ecological provinces (b) for the eastern U.S. Modified Braun classification (Braun 1950; Dyer 2006) shown as white outlines in (a) and province-level data from the Hierarchical Framework of Ecological Units in the United States (Cleland et al. 1997) shown in (b).

Figure 7. Non-spatial cluster analysis (k = 5) of percent change in mean annual 64 temperature (CHGMAT) and mean annual precipitation (CHGMAP) projected for 596 military installations in the coterminous U.S. by 2085 under the A2 emissions scenario. Group medians in boxplots (a) and (b) are all significantly different (p < 0.01, ɑ = 0.05, Wilcoxon/Kruskal-Wallis non-parametric multiple comparisons). Map (c) illustrates geographic distribution of installation groups and principal components plot (d) shows distributions in bivariate data space.

Figure 8. Statistical summary (a), map of clusters (b) and parallel box plot (c) 65 comparing results of bivariate spatial cluster analysis from the ArcGIS 10.2 Grouping Tool (parameters: number of groups = 8; spatial constraint = K_NEAREST NEIGHBORS; distance method = EUCLIDEAN; number of neighbors k = 8). Colored circles in (a) and (c) are groups means; colored vertical lines in boxplots (a) are group minimum and maximum values. Box plots indicate range, interquartile range and medians for each variable. R2 in (a) = the total sum of squares – explained sum of squares) / total sum of squares.

Figure 9. Holdridge temperature classes for the western USA based on mean annual 66 temperature from 1951 – 2006 (a) and projected mean annual temperature in 2085 under the A2 high emissions scenario (b). Data source: ClimateWizard, www.climatewizard.org (Girvetz et al. 2009).

Figure 10. Holdridge life zones derived from historical (1951-2006) climate conditions 67 compared with select boreal species distributions (crosshatched polygons, Little 1971) in the eastern U.S.

Figure 11. Military installations in the coterminous U.S. distributed along mean annual 68 temperature and mean annual precipitation gradients based on historical climate data. Installations are represented by white triangles. Projected climate trajectories A, B and C (white arrows) are shown for select installations to illustrate potential biome shifts as discussed in the text. Climate data are derived from the 1981-2010 U.S. Climate Normals dataset (Arguez et al. 2012), available at: http://www.ncdc.noaa. gov/data- access/land-based-station-data/land-based-datasets/climate-normals/1981-2010- normals-data. A complete list of installations and acronyms is shown in Appendix A, Table A-1.

Chapter 2 Table A-1. Life history traits for species and cover types included in LANDIS-II Age- 99 Only Succession model.

Table A-2. Establishment probabilities for species included in LANDIS-II Age-Only 100 Succession model listed by ecological site type.

Table 1. Relative frequency, density and abundance of the 22 most common tree 101 species at Ft. Drum, NY, USA.

Table 2. Age-diameter relationships developed from site trees (n = 395) extracted from 102 the New York FIA database.

Figure 1. Study site location (inset) and location of military infrastructure (training 103 areas shown in green outline) at Fort Drum, New York, USA. Training areas are comprised of over 1,500 forest stands managed to support military training requirements, timber and fiber production, game and non-game wildlife management programs and ecosystem sustainability. The “Impact Zone” and developed areas were excluded from the study.

Figure 2. Relative basal area (a) for the twenty most frequent trees at Fort Drum and for 104 the same species on site tree plots extracted from the New York FIA database. Overall, species composition and relative density are similar, but significant differences are highlighted for several species (bold type and gray shading). Diameter distributions (Dbh) are shown for all trees for FIA site tree plots (b) and the Fort Drum forest inventory plots (c).

Figure 3. Age distributions for the 12 most common tree species at Ft. Drum, New 105 York, USA based on age-diameter equations developed from FIA site trees.

Figure 4. Age-diameter relationships for sugar maple derived from previous studies and 106 analysis of FIA site trees.

Figure 5. Age-diameter relationships for northern red oak derived from previous studies 107 and analysis of FIA site trees.

Figure 6. Landscape level change in the 12 most common overstory tree species at Fort 108 Drum, New York, USA over a 100-year simulation of forest succession under a no disturbance scenario.

Figure 7. Current (a) and simulated (b) community type-age cohorts at Ft. Drum, New 109 York, USA. Maple-dominated stands in maple-elm (bright green) and northern hardwood (dark brown) types increased substantially over the 100-year simulation while open grass-forb communities, oak woodlands (orange) and mixed pine (dark olive) stands declined. Forest stand data were not available for “INACTIVE” (gray) areas. In the map legend, “YNG” denotes stands less than 40 years of age and “HWD” denotes northern hardwood species.

Chapter 3

Table A-1. Reclassification of soil series into ecological site types (Jefferson and Lewis 147 counties, New York, USA).

Table 1. Biophysical characteristics used to develop and evaluate ecological site types 154 at Fort Drum, New York, USA. Four continuous variables (*) were used to define 5

preliminary types using k-means clustering and were included with soil drainage class and temperature regime in recursive partitioning analysis.

Table 2. Effect of projected soil moisture deficits on species establishment probabilities 155 by ecological site type at Fort Drum, New York, USA. Species adaptation to drought is based on literature (Pastor and Post 1986; Burns and Honkala 1990; Gustafson and Sturtevant 2013) and species-site analyses.

Table 3. Characteristics of 23 woody species and 2 non-forest cover types used in 156 LANDIS-II simulations.

Table 4. Multivariate correlation of environmental variables used to guide classification 157 of ecological site types at Fort Drum, New York, USA. All correlation statistics were significant except where noted by “ns” (Spearman’s U, D = 0.05, p < 0.0001).

Table 5. Indicator values based on relative frequency of 23 tree species at Fort Drum, 158 New York, USA. Maximum indicator values (IV) are statistically different than expected by chance except where noted * (p < 0.05, Monte Carlo randomization, n = 1000). Indicator values shown in bold font are more strongly associated with respective site types. See text for descriptions of ecological site types.

Table 6. Soil water deficit (mm) for seven ecological site types at Fort Drum, New 159 York, USA for mid- and late-century climate conditions. Deficits were derived using the WebWIMP water balance modeling tool (Wilmott et al. 1985, http://climate.geog.udel.edu/~wimp/, accessed 4/8/2018). Changes in mean annual temperature (o C) and mean annual precipitation (mm) for Representative Concentration Pathway (RCP) 4.5 and 8.5 are based on ensemble projections from the Community Climate System Model 3.0 (CCSM-3, NCAR GIS Program 2012). See text for descriptions of ecological site types.

Table 7. Above ground biomass projections for 23 tree species at Fort Drum, New 160 York, USA under a high emissions (RCP 8.5) climate change scenario.

Figure 1. Study site location (inset) and location of military infrastructure (training 161

xii

areas shown in green outline) at Fort Drum, New York, USA.

Figure 2. Historical mean annual temperature (oC), precipitation (mm) and Palmer 162 Drought Severity Index (PDSI, Palmer 1965) for Watertown, New York, USA (Sources: NOAA National Center for Environmental Information, Global Summary of the Year data, https://www.ncdc.noaa.gov/cdo-web/; U.S. Drought Risk Atlas, National Drought Mitigation Center, http://droughtatlas.unl.edu/). Fitted dashed lines illustrate trends over time.

Figure 3. K-means cluster analysis (n = 6,741) of four environmental variables at Ft. 163 Drum, New York, USA. Circle sizes in the principal components plot (a) are proportional to the number of sample units in each cluster and color-shaded regions contain 90% of the values in respective clusters. Cluster means were all significantly different (p = 0.05, nonparametric multiple comparisons, Dunn 1964) except where noted by superscripts (b). The map at right (c) illustrates the spatial distribution of each cluster on the installation.

Figure 4. Ecological site types at Fort Drum, New York, USA derived from 1:12,000 164 county soil series (Soil Survey Geographic (SSURGO) Database for Jefferson and Lewis counties, New York; available at https://sdmdataaccess.sc.egov.usda.gov. Accessed 04/28/2013). Areal proportions for each type shown in parentheses. See text for complete descriptions of ecological site types.

Figure 5. Recursive partitioning of seven ecological site types based on four 165 environmental variables at Fort Drum, New York, USA. The classification tree model explained r2 = 0.86 (n = 5416; RMSE = 0.22) of the variation with a misclassification rate of 5.6 percent. Twenty percent of the observations were withheld for validation (r2 = 0.859; n = 1324; RMSE = 0.23).

Figure 6. Change in species establishment probability (SEP) for Acer saccharum at Fort 166 Drum, New York, USA over time by ecological site type and emissions scenario. The Representative Concentration Pathway (RCP) 4.5 scenario assumes an increase in mean annual temperature in the early part of the century (+ 2.1o C) that is maintained through

xiii

the end of the century by conservation measures. The RCP 8.5 scenario assumes continued production of greenhouse gases at current rates and increasing mean annual temperatures to the end of the century (+4.7o C). Climate projections are based on the Community Climate System Model 3.0 (National Center for Atmospheric Research (NCAR) GIS Program. 2012). See text for descriptions of ecological site types.

Figure 7. Predicted above ground biomass (AGB) for all species and cover types by 167 ecological site type at Fort Drum, New York, USA under current climate conditions (base model, top) and the high emissions RCP 8.5 scenario (bottom). Results for the RCP 4.5 scenario were intermediate and not shown for clarity of presentation. See text for descriptions of ecological site types.

Figure 8. Total biomass (g m2 -1) of five tree species with different shade and drought 168 tolerances shown for two ecological site types at Fort Drum, New York, USA. The SND site type had an estimated soil water deficit approximately twice that of the TER site type at year 2095 under the high emissions RCP 8.5 scenario. See text for descriptions of ecological site types.

Figure 9. Total projected above ground biomass (AGB) for 16 tree species at Fort 169 Drum, New York, USA under a high emissions (RCP 8.5) climate change scenario based on an ensemble Community Climate System Model 3.0 (NCAR GIS Program 2012). Solid lines represent relatively shade tolerant species and dashed lines represent relatively shade intolerant species. Biomass projections based on the lower emissions RCP 4.5 scenario were only slightly higher than those under the RCP 8.5 scenario and are not shown for clarity of presentation.

Figure 10. Projected change in community type over time under a high emissions (RCP 170 8.5) climate change scenario at Fort Drum, New York, USA. Area shown in dark gray on map was not classified into a forest type due to the lack of forest inventory data.

Figure 11. Monthly water balance (mm) for three soils at Ft. Drum, New York, USA for 171 years 2015 and 2095. Available water holding capacity (AWC) was derived from SSURGO data for the top 100 cm of soil (Soil Survey Staff 2013). Water balances are

xiv based on projected mean annual temperature and mean annual precipitation (CCSM-3 ensemble RCP 8.5 scenario, NCAR GIS Program 2012) and calculated using the WebWIMP water balance model (Wilmott et al. 1985).

Project Overview

Despite uncertainties about the effects of the current global warming trend, managers responsible for natural resource management on Department of Defense (DoD) lands in the U.S. must continue to meet range management and training needs at major U.S. military installations. To support resource management planning under uncertain global warming scenarios, the United States (U.S.) Army, Engineer Research and Development Center (ERDC) is conducting research on a wide range of environmental issues that influence the dual-missions of large, land-based installations in the coterminous U.S., which is to provide the necessary mission training to meet national defense objectives and to manage natural resources within the context of federal environmental regulations to achieve sustainable ranges and lands (U.S. Army ERDC, http://www.erdc.usace.army.mil/About/ERDCCapabilities/Ongoing Research.aspx). In addition to conducting and sponsoring research designed to understand how ecosystems might be affected by global warming, ERDC is also involved in developing decision support tools to assist resource managers with long-term resource planning under uncertain future climate scenarios.

In 2011, the Department of Fish and Wildlife Conservation at Virginia Tech, in collaboration with the U.S. Geological Survey Virginia Cooperative Fish and Wildlife Research Unit, received a grant from ERDC to develop a vulnerability framework to better understand potential threats from global warming to military installations nationwide. The goal was to integrate information from Global Climate Models (GCMs) and greenhouse gas emission scenarios developed by the International Panel for Climate Change (IPCC 2007; 2014; http://www.ipcc.ch/) with extant environmental data to create a modeling framework that would indicate the relative risk of significant environmental change for all military lands in the coterminous U.S. An ecoregional analysis approach was suggested by the ERDC research team as the best way to approach the problem since conducting site-specific research at every installation was not feasible for logistic and financial reasons. It was assumed that installations within similar bioclimatic regions and with similar ecological conditions and processes would respond to global warming in a similar manner, at least for this “coarse filter” risk analysis. However, since the coarse scale analysis could not address site-specific issues important to individual resource management programs at all installations, the project also sought to conduct fine scale research at representative xvi installations that would help identify the environmental data and analytical approaches required to better understand the potential effects of global warming on species, communities, land management practices and military missions at a management scale.

This dissertation will address the coarse scale vulnerability analysis conducted for military installations throughout the coterminous U.S. and fine scale modeling at one installation, Ft. Drum, New York. Ft. Drum was selected for several reasons: 1) it was assumed that this installation located in upstate New York would be representative of the potential impacts of global warming in cool temperate forest types in the northeastern U.S. (e.g., northern hardwoods- hemlock (Fagus grandifolia, Betula alleghaniensis, Acer saccharum, Tsuga canadensis), elm- ash-cottonwood (Ulmus americana, Fraxinus americana, Populus deltoides) and Lake States pine (Pinus strobus, Pinus resinosa), Eyre 1980); 2) cooperative relationships were already established with resource managers at Ft. Drum with on-going natural resources research projects; and 3) a relatively large amount of resource data were available for the installation, including a recently completed forest inventory. In addition, previous research has indicated that forest communities and tree species in the northeastern and north-central U.S. might be particularly vulnerable to rapid warming of regional climates (McKenney et al. 2007; Hayhoe et al. 2008; Mohan, Cox and Iverson 2009; Yaussy, Iverson and Matthews 2013; Clark et al. 2014).

Other members of the ERDC climate change project team addressed the potential impacts of rapid climate change in grassland (Hovick et al. 2014) and other forest ecosystems (Matthews et al. 2014) to provide additional knowledge that will be integrated into a comprehensive decision- support framework for use by installation resource managers.

xvii

Chapter 1 Assessing the Vulnerability of Military Installations in the Coterminous United States to Potential Biome Shifts resulting from Rapid Climate Change

Abstract A biome classification based on current climate conditions was developed for the coterminous United States using the Holdridge Life Zone system. The classification was validated by comparison to existing ecoregional classifications, the distribution of major forest formations and tree species ranges in eastern North America. Life zones were projected for mid- and late century time periods under three greenhouse gas emission scenarios (low - B1, moderate - A1B and high - A2) using an ensemble of global climate models. The potential vulnerability of installations managed by the U.S. Department of Defense (n = 596) was analyzed relative to projected biome shifts using spatial cluster analysis to characterize interregional variation and identify representative installations for subsequent landscape-level analyses. Results indicated that while mean annual temperatures are expected to increase in all parts of the country, installations located in the northeastern U.S., northern Lake States and western Great Plains are likely to experience the largest proportional increases in mean annual temperature in comparison to historical conditions. Forest and grassland communities at these installations that are managed to support a wide range of military training and environmental objectives may be adversely affected by altered disturbance regimes, excessive heat and moisture stress. However, precipitation is projected to increase in the northeast and Lake States and may mitigate some effects of increased atmospheric temperatures on biological communities. Given the uncertainty of how species and communities may respond to climate change in the coming decades in different ecoregions, additional environmental and cultural attributes need to be developed for installation locations and integrated within a decision support framework to assist resource managers in understanding potential vulnerabilities and planning appropriate responses.

Introduction Scientific evidence continues to mount that the current period of climatic warming is exceptional when compared to temperature fluctuations over the past 150 years, especially at higher latitudes (IPCC 2014). Arctic sea ice appears to be shrinking at an unprecedented rate and mean global sea levels continue to rise, trends linked to the continued increase of carbon dioxide (CO2) and other greenhouse gases in the Earth’s atmosphere (Hinzman et al. 2005). Mean global air and sea temperatures also continue to rise and most current General Circulation or Global Climate Models (GCMs) project these trends to continue over the next century under most emissions scenarios (IPCC 2014). A wide range of potential impacts on terrestrial, aquatic and marine ecosystems throughout the world have been documented and many more postulated if the current rate of warming continues as projected (Walther et al. 2002; Parmesan and Yohe 2003; Menzel et al. 2006; Botkin et al. 2007; Post et al. 2009; Balmaseda et al. 2013; Moritz and Agudo 2013; IPCC 2014). Bergengren et al. (2011) estimate that as much as 37 percent of the Earth’s terrestrial ecosystems could experience biome-level changes under the most severe warming scenarios. In the United States (U.S.), numerous studies over the past decade have also outlined the potential impacts of global warming on forest ecosystems including general declines in growth and productivity (Potter, Li and Hiatt 2012), increased likelihood of damage from insect pests and pathogens (Lovett et al. 2006; Dukes et al. 2009), significant shifts in tree species abundance and ranges (Iverson et al. 2008; Dale et al. 2010; Pucko et al. 2011), negative impacts to life cycles and distributions of fauna (Rodenhouse et al. 2008; 2009) and changes to biogeochemical and hydrologic cycles (Hayhoe et al. 2007; Campbell et al. 2009). Many of these potential impacts are of concern to resource managers at military installations that are responsible for managing large areas to support sustainable range management and training activities while also supporting natural resource stewardship. Similar to other federal land managers, resource professionals on military installations must comply with requirements of the Endangered Species Act, Clean Water Act, National Environmental Protection Act, the Sikes Act and other federal environmental legislation, as well as state and local environmental regulations. Accordingly, military land managers are seeking guidance on exactly how rapid climate change could affect resources under their charge in both the short- and long-term.

Unfortunately, while there may be agreement among many climate scientists on the general trend in atmospheric warming during the 20th century and projections for CO2-induced temperature increases for the next few decades, there remains considerable uncertainty concerning exactly how warmer climates will impact forest ecosystems at regional (sub-continental) and larger scales (Clark et al. 2001; Millar et al. 2007; Herr et al. 2016; Luce et al. 2016). A major source

of uncertainty is disagreement among global climate models in their predictions of future precipitation patterns. As mean annual temperature increases, evapotranspiration is expected to increase in many areas of the world (Seneviratne et al. 2006; Bonan 2008; Jung et al. 2010), but if precipitation also increases, the effects of higher temperatures could be mitigated to some degree. For eastern North America, most models predict a 0-10 percent increase in precipitation by the latter part of the century (IPCC 2014). However, some models indicate a 0-10 percent reduction in precipitation for southern tier states, and overall, precipitation predictions from current GCMs are within the range of natural climatic variability for much of the eastern half of the country (IPCC 2014, Summary for Policymakers, Fifth Assessment Synthesis Report, Figure SPM.7b). Further, Deser et al. (2012) showed that significant regional variations in both temperature and precipitation projections can occur even within a single GCM when natural climatic variation is taken into account. Using 40 simulations based on the National Center for Atmospheric Research, Community Climate System Model (CCSM3) with identical climate forcings for North America, Deser et al. (2012) concluded that consistent prediction of similar climate trends (similar sign and magnitude) was only achievable for a few years beyond the limits of observed conditions. Over the next 55 years, their results indicated that all of the 40 ensemble projections were “plausible outcomes” and that natural variation in temperature and precipitation patterns contributed significantly to model uncertainty. Fangxing et al. (2014) highlighted that both GCM’s and downscaled climate models exhibit significant biases in their predictions of future temperature and precipitation patterns when applied at a regional scale and found no statistically significant trends in precipitation for the northeastern U.S. for end of century projections under relatively high CO2 emissions scenarios. Hall (2014) suggested that downscaling of GCM outputs might be appropriate in areas with relatively homogeneous regional climates (e.g., mid-continent regions with little topographic relief), but might produce unreliable estimates of future climatic conditions in regions with complex topography or contrasting land cover (e.g., strong regional effects on temperature and precipitation from the Great Lakes).

Given the level of uncertainty in projections of future temperature and precipitation patterns associated with current global and regional climate models and the expected differential response

3 of ecosystems in different biogeographic settings, we chose to focus on the relative vulnerability of military installations to potential biome shifts as opposed to attempting to quantify specific resource impacts. Responses to rapid global warming are likely to be highly regionalized and it is unclear if predicting specific local impacts (e.g., increase or decrease in wetland area, changes in endangered species populations or likelihood of a tree species being extirpated) at various military installations in widely differing bioclimatic regions is appropriate using current global and regional climate models (Araujo et al. 2005; Guisan and Thiuller 2005). Pielke et al. (2007) suggested that a vulnerability or risk assessment approach might provide a more comprehensive framework for examining the many potential impacts of rapid global warming on ecological systems. Factors such as patterns and trends in land use, human population density, availability of required habitat components and even socio-political processes can have far more pervasive effects on species distribution and alteration of ecological processes than climate – especially at local geographic scales and over relatively short time periods (Hansen et al. 2001; Bonan 2008; Hof et al. 2011). Therefore, understanding how potential changes in broad temperature and moisture patterns might affect major biomes and their constituent ecosystem services is a critical starting point for developing a comprehensive vulnerability index at an ecoregion scale (ERDC 2010).

Methods

Data acquisition and processing Study sites included 596 U.S. military installations located in the coterminous U.S. (see Appendix A, Table A-1 for list of installations). In addition to major active installations where large numbers of troops are housed and trained, the installation dataset contained sites such as training ranges, munitions storage sites (arsenals), historical sites, National Guard training areas and other locations managed by the Department of Defense (DoD). To model potential biome shifts, the Holdridge life zone system (Holdridge 1947; 1965) was applied in the coterminous U.S. and projected over time using climatic outputs from global climate models. The Holdridge system incorporates mean annual temperature (MAT), mean annual precipitation (MAP) and potential evapotranspiration (PEVT) to define major life zones or biomes. It was developed primarily through research in Central America and the Caribbean, but has been applied in many 4 areas of the world including North America (Lugo et al. 1999), Australia (Jia et al. 2012) and globally (Emanuel, Shugart and Stevenson 1985; Olson et al. 2001; Sisneros et al. 2011). Although not designed to be as detailed as more recently developed ecological classifications used in the U.S. (e.g., Cleland et al. 1997; Comer et al. 2003; Omernik and Griffith 2014), the Holdridge system relies directly on simple climatic variables modeled by all GCMs and in many ways avoids the complex geologic, edaphic and floristic components used by other classifications that are difficult to relate objectively to output from global climate models (Lugo et al. 1999). Moreover, the coarse resolution of the Holdridge system is consistent with the spatial, temporal and predictive limitations of current climate models and provides a more suitable framework for analyzing small-scale (large area) phenomena such as biome shifts or changes in the geographic range of a species. For the purposes of this project, the Holdridge system provided a means to analyze present and projected climate regimes for military installations in the U.S. that was very climate-centric and relatable in a natural resource management context.

Visually, the Holdridge system can be represented as a triangular diagram with each axis corresponding to one of the three climatic variables (Figure 1). Mean annual temperature (oC), mean annual precipitation (mm) and mean annual evapotranspiration ratio (PEVT divided by MAP) interact to create life zones represented by hexagonal shapes within the triangle. Hexagons approximate the bioclimatic boundaries of major life zones or ecoregions along both latitudinal and altitudinal gradients; they do not attempt to describe specific vegetations or forest formations. Mean annual temperature is expressed as “biotemperature”, meaning all temperatures above the freezing point averaged over the year. Temperatures below 0o C are not biologically meaningful for most terrestrial communities because all water would be frozen and unavailable for biological processes. Potential evapotranspiration ratio is calculated using the following formula:

PEVT = (MAT (oC) * 58.93mm/ oC) / MAP (mm); where MAT = Holdridge biotemperature (Holdridge 1959).

Mean annual temperature and precipitation values were graphically analyzed within the context of the Holdridge life zone system to better understand how installations were distributed across life zone classes and to begin developing bioclimatic groupings for subsequent analyses.

Climate data (MAT, MAP) in ASCII raster format covering the coterminous U.S. were downloaded from the ClimateWizard web site (Girvetz et al. 2009; http://www.climatewizard.org) for three time periods (current, mid-21st century and late-21st century) and three future emissions scenarios. Current conditions were based on monthly temperature and precipitation data collected at approximately 8,000 observation points from 1895-1997 (Gibson et al. 2002) and modeled as a 4 km-resolution grid using climatologically- aided interpolation (Parameter-elevation Relationships on Independent Slopes Model, PRISM, Daly et al. 1994; 2008). The PRISM model incorporates geographic and physiographic factors to correct for orographic effects on temperature and precipitation in mountainous and coastal regions (Daly et al. 2008). The ClimateWizard tool utilizes PRISM data from 1951-2006 as the basis for the “Historical 4 km Lower 48” dataset because observational data from the latter part of the 20th century were more abundant (larger number of weather stations) and considered more reliable (C. Zganjar 2012, personal communication). Mean annual temperature and mean annual precipitation were modeled for 2055 and 2085 using average values from an ensemble of 16 global climate models (Girvetz et al. 2009, see Appendix A, Table A-2 for a complete list of climate models available on the ClimateWizard.org site). Many climate change studies have shown that using averaged values from multiple models produces more reliable estimates of climatic means than using any single GCM (Gleckler et al. 2008; Pierce et al. 2009). Ensemble averages tend to minimize variability in climate projections among different individual models, but may not always result in the most accurate estimates for some climate variables. Data available from the ClimateWizard site are based on downscaled global climate projections from the World Climate Research Programme's, Coupled Model Intercomparison Project phase 3 (CMIP3) multi-model dataset (Maurer et al. 2007; Meehl et al. 2007). Three emissions scenarios (Nakicenovic et al. 2000) were available from the ClimateWizard site:

● B1 – aggressive conservation and alternative energy utilization strategies are adopted early in the century and CO2 levels decrease by 25 percent by 2100. ● A1B – carbon dioxide levels continue to rise until mid-century when conservation and alternative energy utilization strategies begin to reduce emissions to approximately 25 percent above current levels by 2100; ● A2 – human populations continue to utilize high levels of fossil fuels over the next century and CO2 concentrations continue to rise to triple the current concentration by 2100.

These scenarios are projected to result in mean annual global surface temperature increases of 1.7, 2.7 and 3.5o C respectively above the current mean global surface temperature by the end of the century (despite eventual decreases in atmospheric CO2 under the B1 scenario, there is a time lag effect due to the amount of CO2 already present in the Earth’s atmosphere by 2050). Due to the imprecise nature of GCMs and speculation about how much fossil fuel will be used in the future, each scenario has a large variance associated with projected temperature increases. For example, for the A2 scenario, projected increases in mean annual global surface temperature by 2100 range from 2.0 to 5.4o C. By comparison, mean annual global temperature has increased by approximately 4o C since the last glacial maxima 18,000 years ago (Delcourt and Delcourt 1998).

To model trajectories of boreal life zones north of the Canadian border, additional climate data based on a single GCM (the National Center for Atmospheric Research, Parallel Climate Model 1.4 (PCM) available at http://gdo-dcp.ucllnl.org/downscaled_cmip_projections/; Washington et al. 2000) were acquired directly from the CMIP3 web site. These data were used to estimate shifts in boreal life zones from the northeastern U.S. into Canada, but only data projected for 2085 under the A2 scenario were utilized. Data were imported to a geographic information system (GIS; ArcGIS 10.2, ESRI 2014) and processed to derive mean annual temperature (MAT), mean annual precipitation (MAP) and mean annual potential evapotranspiration (PEVT) surfaces for 2085 covering the coterminous U.S. and southern Canada. Life zones in Canada derived from the single PCM dataset could be quite different from those derived from ensemble projections acquired from the ClimateWizard site, but pre-processed ensemble data with similar

7 resolution and coverage were not available from the several climate data repositories reviewed. Acquiring and processing data, including downscaling from global to CONUS scales, from multiple GCM’s was beyond the scope of this project. The PCM model has been shown to approximate median results in comparison to other widely used GCM’s in terms of projecting mean annual global surface air temperatures at multi-decadal scales (AchutaRao et al. 2004).

Historical climate data were resampled from 4 km to 12 km spatial resolution to match the resolution of data for future GCM projections. All data were projected and georeferenced to the North American Lambert Conformal Conic projection to insure that subsequent calculations of area and distance were more accurate relative to using non-projected data. After resampling and projecting data, 67 installation locations were dropped from the analysis because their locations were outside the geographic boundaries of raster datasets. All of these installations were located on coastlines, but were in close proximity to neighboring installations that were included in analyses. A preliminary cartographic model was developed to guide processing of climate data into maps of Holdridge life zone classes covering the coterminous U.S. Continuous raster data downloaded from the ClimateWizard site or derived from CMIP3 data were reclassified into categories using class thresholds from the Holdridge system. The COMBINE function in the ArcGIS Spatial Analyst Toolbox was used to delineate all unique combinations of categorical data and create raster maps of life zone classes for each date and emissions scenario. A majority filter was applied to life zone maps to remove individual outlier pixels. Once all processing steps were tested and finalized, a final cartographic model was developed using ArcGIS ModelBuilder to automate the classification process (Figure 2).

Final life zone maps were visually compared to distributions of major forest formations and ecological units in the eastern U.S. to assess how well life zones based on historical climate data fit with widely used descriptions of vegetation and ecological systems. Spatial data depicting an updated version of forest formations originally described by Braun (1950) were overlaid on Holdridge life zones in ArcGIS (Figure 6a). Dyer (2006) used Forest Inventory and Analysis (FIA) data to modify the original Braun classification, which was largely based on observations in mature and relatively undisturbed forests, to reflect changes in species composition resulting

8 from widespread disturbance during the late 19th and early 20th centuries (i.e., secondary forest). Province level data from the National Hierarchical System of Ecological Units (ECOMAP, Cleland et al. 1997) were also acquired for comparison with life zone maps. The ECOMAP national ecological classification was developed by the USDA Forest Service to support ecosystem management objectives and was first released in 1993 (Bailey 2004). It incorporates a hierarchical framework for classifying and mapping ecological units from a continental scale (1:7,500,000) to landscape units applicable to forest management (1:24,000), and has been widely used as the basis for analyzing species distributions, climate change and other biogeographic studies. Major climate zones in North America are captured at the “Division” level in ECOMAP, but vegetation formations (e.g., boreal forest, prairie, steppe-shrub) are described at the Province level and provided a higher level of geographic detail for relating Holdridge Life Zones to broad biological communities.

Assessing change For the eastern half of the U.S., the latitude and longitude of centroids derived from ellipses fitted to each Holdridge life zone were extracted from spatial data layers for all emission scenarios at years 2006, 2055 and 2085 using the Zonal Geometry Tool in ArcGIS 10.2. The distance between current and projected life zone centroids and general direction of shifts was also determined. Life zones with relatively similar climatic conditions and geographic distributions were combined to simplify analyses and interpretation of results. In addition, life zones in the southern Appalachian Mountains with very small areas (< 1000 km2) were not included in tabular summaries. Many of these zones were comprised of only 1-2 map pixels and although interesting ecologically (e.g., cool temperate superhumid zones in the higher southern Appalachians were projected to become warm temperate humid zones in 2085 under the A2 scenario), confidence in their classification and positional accuracy is low given the scale of the analysis. Similarly, using ellipse centroids to quantify biome shifts was not deemed useful for the western U.S. where the spatial pattern of life zones was highly irregular and fragmented due to topographic variation. Climate zones in the mountainous West are largely determined by changes in altitude rather than latitude and higher resolution data may be necessary to accurately depict the distribution of life zones in much of the western U.S. As a result, some analytical

9 results will be discussed primarily in the context of the eastern U.S. Percent change in mean climatic variables was determined for 596 military installations by comparing climate values from projected data to historical baselines (projected values – historic values) / historic values * 100).

Non-hierarchical, k-means cluster analysis (JMP®, Version 11. SAS Institute Inc., Cary, NC, 1989-2007) was used to analyze percent change in climate variables at installation locations to explore potential regional variation and assist in identifying representative installations at a regional level for additional research. K-means clustering is a widely used tool for visually exploring multivariate variation in large datasets where hierarchical relationships are not of interest. However, its effectiveness in identifying meaningful clusters is highly dependent on cluster size, cluster shape and the density of observations in multivariate space (Jain 2010). K- means clustering seeks to minimize the variance in groups of similar observations while maximizing the between-group distance. Determining the optimal number of groups (k) is not an exact science so multiple analyses where conducted using percent change in mean annual temperature (CHGMAT), mean annual precipitation (CHGMAP) and mean annual potential evapotranspiration ratio (CHGPEVT) as input variables and the number of groups ranging from 4 to 10. Variables were scaled individually and were not standardized because they were already expressed on a proportional scale. Differences in groups were evaluated using bivariate plots of principle component scores resulting from k-means analyses, standard box plots and a nonparametric test for differences in group medians (Kruskal andWallis 1952). The Kruskal-Wallis Test utilizes ranked sums rather than raw data and does not depend on assumptions of univariate normality (distributions of temperature and precipitation data were non-normal and could not be transformed to normal distributions using standard approaches). However, the Kruskal-Wallis Test does require homogeneity of variance and similar distributions for all groups. If these assumptions are met, then testing for differences in group central tendencies (means and medians) is valid (Zar 2010). Distributions for CHGMAT, CHGMAP and CHGPEVT were compared using histograms and evaluation of summary statistics. All statistical analyses were performed in JMP Professional (JMP®, Version 11. SAS Institute Inc., Cary, NC, 1989-2007).

Traditional cluster analysis techniques such as k-means do not generally take into account the relationships of sample locations in geographic space in forming groups. Therefore, in order to define more regionally meaningful groups of military installations, spatial cluster analysis was used to analyze projected rates of change in mean annual temperature and precipitation. Spatial cluster analysis takes into account the proximity of sample locations in geographic space in addition to variability in non-spatial multivariate space (Duque et al. 2007). A simple nearest neighbors approach was utilized with the number of neighbors and distance calculation method set to default values (k = 8 and Euclidean distance respectively) to explore how data were clustered when 4, 6 and 8 groups were arbitrarily specified. Specifying a minimum number of neighboring locations considered in forming groups prevents the formation of groups with very low membership, which can occur with outlier values. The relative importance of each explanatory variable in forming groups was assessed by comparing overall and group-wise r2 statistics and reviewing parallel box plots for all variables. A pseudo-F statistic (Caliński and Harabasz 1974) was calculated to estimate the optimal number of final groups based on analysis of between-group differences and within-group similarities.

Results Approximately two-thirds of the country was projected to experience mean annual temperature increases in the 12 to 19 percent range, but the Intermountain West, northern Great Plains and northern New England were projected to have increases ranging from 20 to 30 percent (Figure 3). In contrast, mean annual temperatures in southern tier states and the West Coast were projected to increase by 10 to 13 percent and the immediate Gulf Coast area by less than 10 percent from current conditions by 2085 under the A2 scenario. Projected changes in precipitation were more variable than for temperature with some areas of the country projected to decrease by more than 10 percent (southern California, southern Great Plains and some portions of the desert Southwest) and other areas to increase by a similar amount (northern Great Plains and northeastern U.S.). Patterns of projected changes in potential evapotranspiration are directly related to projected temperature and precipitation patterns, but the relatively large increases in mean annual temperature projected for the Rocky Mountain plateau and northern tier states appear to offset projected increases in precipitation in these areas resulting in very large

increases in potential evapotranspiration relative to other areas of the country (80 to greater than 1,000 percent; Figure 3c).

Classification of climate surfaces using thresholds from the Holdridge life zone system resulted in 65 zones being identified in the coterminous U.S. based on historical climate data (Figure 4a), and 71, 71 and 68 classes identified for the 2085 B1, A1B and A2 scenarios respectively (Figure 4b, only life zones for the 2085-A2 scenario are shown). The thirteen largest life zones, which represent 20 percent of the total number, comprise 78 percent of the total area. The three largest life zones make up over 37.6 percent of the total area (Table 1). Under the A2 scenario, boreal life zones are projected to be eliminated from the eastern U.S. by 2085 with the exception of very isolated areas in the highest elevations of the northern Appalachian Mountains. The centers of boreal zones in the northeastern U.S. shift 500 km to 600 km northward into Canada and are largely replaced by cool temperate humid and perhumid regimes. Concomitantly, subtropical zones in the southern U.S. increase substantially in area (250 to 400 percent) and expand from their current positions in southern Florida and the Gulf Coast as far north as western Kentucky and southeastern Virginia. By 2085, subhumid and semiarid tropical zones that do not currently exist in the U.S. cover most of the Florida Peninsula and south Texas (gold to yellow colors in Figures 4b and 5b, c). The warm temperate humid life zone that currently dominates the southeastern U.S. from Texas to Virginia does not change substantially in areal extent (12 percent decline from current conditions to 2085-A2), but shifts northward approximately 600 km and spans an area from Missouri to southern New England. Under all emissions scenarios, most life zones in the eastern U.S. shift northward as expected given the general zonal nature of mean annual temperature in the region. However, a relatively small warm temperate, subhumid steppe- woodland class (life zone 543) currently located in central Oklahoma is projected to expand significantly into the Midwest as the climate warms, increasing its areal coverage by over 1,400 percent. The average shift for all life zone centroids in the eastern U.S. by 2085 was 600 km in a north-northeast direction or approximately 8 km per year (Table 2).

Visual comparison of Holdridge life zones with widely used maps of eastern forest formations and ecological provinces showed excellent correlation in some regions, but relatively poor

12 correlation in transitional areas between warm temperate and boreal zones (Figure 6). Dyer’s (2006) hemlock (Tsuga canadensis)-northern hardwood and northern hardwood-red pine (Pinus resinosa) formations closely match boreal life zones with perhumid and humid moisture regimes in the northeast and western Great Lakes regions respectively. In the southeastern U.S., the southern mixed and subtropical evergreen forest formations also correspond well with warm temperate humid and subtropical humid life zones. The beech (Fagus grandifolia)-maple (Acer saccharum), oak (Quercus spp.)-hickory (Carya spp.) and mesophytic formations are distributed across at least two and as many as four life zones, which is not unexpected given the influences of regional topography (Appalachian and Quachita-Ozark Highlands) and the very broad geographic distribution of the mesophytic forest formation as defined by Dyer (2006). Distributions of oak-chestnut (Castanea dentata) and mixed mesophytic formations as originally defined by Braun (1950) match cool temperate perhumid and humid life zones along the Appalachian Mountains very well. However, The Quachita-Ozark Highlands are apparently not high enough in elevation for their effects on regional climate to be seen at this scale of analysis and the oak-hickory formation as mapped by Dyer (2006) or Braun (1950) does not correspond well with the distribution of any life zone in the Holdridge system.

The influence of the Appalachian Mountains is also evident when comparing life zones to ECOMAP units in the eastern U.S. Ecological units with the “M” designation for mountainous regions (M211 and M221) correspond very well with montane boreal superhumid and cool temperate perhumid life zones in the northern and southern Appalachians. Non-montane boreal life zones correspond fairly well with Laurentian (212) and Northeastern (211) Mixed Forest Provinces in the ECOMAP system and cool temperate life zones also are reasonably correlated with the Eastern Broadleaf Forest Province (221) and northern portions of the Central Interior Broadleaf Forest Province (223). However, the southern half of the Central Interior Broadleaf Forest Province was classified as warm temperate humid forest in the Holdridge system, suggesting similarities with climate conditions farther south. Two ECOMAP units cover most of the southeastern U.S., the subtropical Southern Mixed Forest (231) and Outer Coastal Plain (232) Mixed Forest provinces, both of which fall largely into the Holdridge warm temperate humid forest life zone. Only the immediate Gulf Coast and Florida Peninsula were classified as

13 subtropical under the Holdridge system, whereas ECOMAP suggests that subtropical conditions extend northward along the Atlantic Coast to the Delmarva Peninsula. Lugo et al. (1997) took issue with this broad definition of subtropical temperature conditions in the ECOMAP classification since Holdridge and others define subtropical regions as areas with little or no frost, which in the southeastern U.S. is restricted to the immediate Gulf Coast and Florida peninsula. The Holdridge system identifies almost all of Florida as having a distinctly warmer climate than most of the Atlantic Coastal Plain to the north and classifies central Florida as subtropical dry forest.

Five groups of military installations were identified from cluster analysis of percent change in climate variables from 2006 to 2085 (A2 emissions scenario) at 596 locations in the coterminous U.S. (Figure 7 and Table 3). Examination of biplot graphs and eigen values for several k-means cluster analyses with the number of predefined groups (k) ranging from 4 to 10 suggested that including percent change in potential evapotranspiration ratio (CHGEVT) added little explanatory value relative to percent change in mean annual temperature (CHGMAT) and mean annual precipitation (CHGMAP). Specifying larger numbers of groups (k > 5) tended to create outlier clusters with relatively few observations and resulted in reassignment of marginal values between groups that did little to elucidate regional patterns. Smaller numbers of groups (k < 4) simply grouped large numbers of installations into “clusters” with large geographic ranges and a high degree of within-cluster variability. Four groups appeared to represent overall regional differences with good discrimination, but specifying a fifth group resulted in a cluster of installations in the Interior West being identified (cluster number 2 in Figure 8) where mean annual temperature is projected to increase by almost 16 percent and precipitation is projected to decrease by an average of 6 percent from 2006 to 2085 under the A2 scenario. Group medians for all 5 groups were significantly different (ɑ = 0.05, p < 0.01) for both variables based on the Kruskal-Wallis multiple comparisons test. However, group distributions for CHGMAT and CHGMAP were not all the same and groups 1 and 5 contained significant outliers for the change in temperature variable (Figure 8b).

Results from spatial cluster analysis were similar to those from non-spatial k-means clustering when a similar number of groups was specified (k = 4 and 6). However, when 8 groups were specified, clusters of installations with very different mean values (outlier clusters) were more clearly identified using spatial clustering and all installations were grouped into more contiguous regional clusters (Figure 8b). Overall, spatial cluster analysis explained a large amount of the variation for all groups (CHGMAP R2 = 0.81; CHGMAT R2 = 0.78). Groups 2 (red color) and 6 (brown color) had very low “share” values for change in precipitation indicating substantially different ranges and variances relative to the combined range and variance for all groups (Figure 8a). These clusters of installations are located in the northeastern and north central U.S. and are projected to experience a 12 percent increase in mean annual precipitation by 2085. Group 3 (green), located in the high plains of eastern Colorado and Wyoming, and group 7 (pink), which was broadly distributed from the southern Great Plains to coastal California, had relatively high share values and their mean values for percent change in precipitation over the next 70 years (- 4.7 and -9.0 respectively) were clearly lower than other group means. The ranges and variances of other groups were more similar to overall values for change in precipitation. For percent change in temperature, groups 2 (red), 3 (green) and 6 (brown) had relatively low share values indicating relatively different ranges and variances compared to values for all groups based on CHGMAT. However, group 2, which is a tight cluster of installations in the coastal mid-Atlantic area, appears to have a low share score because its range and variance is small relative to the overall range and variation of the CHGMAT variable. Mean percent change in mean annual temperature for the group is not very different from the mean for all groups. Groups 1 (blue) and 7 (pink) had the lowest mean changes in temperature and groups 3 (green) and 6 (brown) had substantially higher means. A parallel box plot produced from the Grouping Analysis Tool (Figure 8c) visually summarized the relative positions of mean values for all groups by climatic variable and highlighted group means that were higher and lower than interquartile ranges for each variable.

Discussion The Holdridge life zone system appeared to capture regional variation in the magnitude and directionality of projected climate changes fairly well at the relatively coarse scale of this

15 analysis. As expected, major life zones conformed to general temperature and precipitation gradients in the coterminous U.S. that are determined by latitude, prevailing wind direction, major mountain ranges and proximity to large sources of atmospheric moisture. Lugo et al. (1999) applied the Holdridge life zone system in the coterminous U.S. using similar input data with a 4 km resolution, but only identified 38 life zones, 95 percent of which were in the western U.S. The eastern half of the country was comprised almost entirely of two life zones; warm temperate and cool temperate moist forests (see Figure 3, Lugo et al. 1999). The smaller number of life zones in the Lugo classification is clearly the result of identifying far fewer temperature classes and slightly fewer precipitation or humidity classes for the majority of the U.S. The only boreal life zone identified was a small area in the Cascade Mountains in Washington State and the subtropical zone only covered extreme southern Florida. By comparison, approximately 20 percent of the U.S. was classified as boreal and 7 percent as subtropical in this study, both orders of magnitude larger in areal coverage in comparison to the Lugo classification. The number of precipitation classes was more similar in the two studies, but the earlier study did not identify any areas in the eastern U.S. with a superhumid precipitation regime, which reduced the number of life zones identified in high rainfall areas in the northern Appalachians. The two classifications used very different approaches and the Lugo study relied on an older version of the PRISM dataset, so perhaps it is not surprising that the resulting classifications appear quite different. Much of the difference may also be attributable to the splitting of some temperate life zones in the current study into finer classes to more closely match the distribution of widely recognized forest types (e.g., life zone 453, which corresponds roughly to distributions of the central oak-hickory and mixed mesophytic forest types, is a transition zone between the warm temperate southeastern mixed forest and the cool temperate beech-maple forests that dominate much of the eastern U.S.). Despite the above differences, Lugo et al. (1999) also concluded that the Holdridge system provided a more objective and climate-centric means of classifying ecoregions in comparison to other schemes evaluated in their study.

Future distributions of Holdridge life zones in the eastern U.S. may shift as much as 600 km to the north-northeast by the latter part of the century resulting in a drastic reduction in the area of boreal and cool temperate life zones and a concomitant increase in the extent of warm temperate

and subtropical zones. If more northern life zones and military lands located within these zones in the eastern U.S. are at greater risk for significant biological changes as a result of projected warming trends, what aspects of forest ecosystems might be affected?

Northeastern U.S. The potential shifts in life zones found in this study, both in terms of magnitude and rate of change over time, are largely consistent with the findings of many others for eastern North America that have modeled changes in species distributions based on climate and other environmental data (Overpeck et al. 1991; Iverson et al. 1998, 2008; McKenney et al. 2007, 2011; Woodall et al. 2009, 2010). Iverson et al. (2008) estimated that habitat for 134 tree species in the eastern U.S. would shift from 400 km to 800 km northward by the end of the century under a range of climate change projections representing “best case” (low CO2-sensitivity model and B1 emissions scenario) and “worst case” (high CO2-sensitivity model and A1F1 scenario) scenarios. Their results indicated that common boreal tree species such as Balsam fir (Abies balsamea), quaking aspen (Populus tremuloides), bigtooth aspen (Populus grandidentatum), red spruce (Picea rubra), black spruce (Picea mariana), yellow birch and paper birch (Betula papyrifera) would decline in importance relative to more southern species such as sweetgum (Liquidambar styraciflua), loblolly pine (Pinus taeda), southern red oak (Quercus falcata), winged elm (Ulmus alata) and eastern cottonwood (Populus deltoides). Using a climatic envelope approach to predict change in the ranges of 130 North American tree species, McKenney et al. (2007) also found that species ranges moved northward an average of approximately 700 km (ranging from 330 km to 1100 km) when modeled using six different GCMs and two emissions scenarios (less severe B2 and more severe A2). Interestingly, since the McKenney study covered all of North America instead of just the coterminous U.S., they were able to illustrate the potential shifts of boreal species northward into Canada and therefore did not report these species as losing large expanses of habitat. Instead, they emphasized the potential loss of habitat for many southern tee species (see McKenney et al. 2007, Table 1) due to temperatures increasing beyond their assumed tolerances and their inability to expand ranges northward at a rate that would keep pace with projected environmental changes. Many of these southern species were predicted to increase in importance in the eastern U.S. by Prasad et al.

2007 and Iverson et al. 2008. The aforementioned two studies used species abundance or importance values as response variables to explain how tree species might respond to changing climate. Canham and Thomas (2010) reported that the average local frequency of a tree species (number of plots where a species was present within a particular climate zone) was more informative in defining species-climate relationships than relative abundance in their evaluation of the 24 most common trees in the northeastern U.S. The general trend of change in species frequency over large temperature gradients was consistent with the other studies mentioned above (i.e., boreal species giving way to temperate species with increasing mean annual temperature), but their results indicated that several cool temperate species (e.g., red maple (Acer rubrum), American beech (Fagus grandifolia), white ash (Fraxinus americana) and black cherry (Prunus serotina)), whose habitats are predicted to decline in other studies might still be well within their climatic ranges with mean annual temperatures 3o to 4o C higher than historical averages. In addition, they found that sugar maple, a species predicted to decline substantially under high CO2 emission scenarios in several studies (Iverson and Prasad 1998; Prasad et al. 2007; McKenney et al. 2007; Mohan, Cox and Iverson 2009), maintained relatively high frequencies across a broad range of mean annual temperature suggesting that this widely distributed species may be able to maintain a significance presence in some parts of the eastern U.S. even under more extreme global warming scenarios. For example, sugar maple appears to be able to maintain 20-25 percent relative frequency in forests with a mean annual temperature of 11o C (Canham and Thomas 2010, Figure 1A), which is approximately the mean annual temperature projected at F. Drum, New York for 2085 under the A2 emissions scenario in this study. Based on a recent forest inventory completed at this installation, sugar maple is currently present on 25 percent of the inventory plots (Ft. Drum Natural Resource Program, unpublished data). This is not to suggest that there will be no changes in sugar maple abundance, regeneration or growth rates under a warmer climate, but current research based on relatively coarse species distribution models do not appear to provide conclusive evidence on exactly how this and other tree species will be affected in the future by rapid climate change across the broad range of environmental conditions found in the eastern U.S.

Research discussed above utilized predictive species distribution models based on correlations between the current distribution of tree species as estimated by FIA plot data and environmental conditions modeled for the same geographic area, but do FIA or other data provide direct empirical evidence of species ranges shifting northward over time? Woodall et al. (2009; 2010) and Zhu et al. (2012) conducted extensive analyses of FIA tree regeneration data (i.e., distribution of seedlings and saplings) relative to the distribution of adult trees of the same species. The intent was to test the widely held view that species responding to a changing climate should show increased regeneration where the environment is becoming more favorable (northern margin of species ranges in eastern North America) and decreased regeneration where conditions are becoming less favorable (southern or western margins of retreating climate niches). Their results indicated that the ranges of the majority of tree species in the eastern U.S. (62 percent of the 92 species studied) have contracted at both northern and southern margins. Some species (21 percent) expanded their ranges to the north and some (16 percent) expanded in a southerly direction. These studies and others suggest that at a macro-scale there is little empirical evidence of tree species responding to current warming trends in the eastern U.S. by increasing occupancy of more northerly sites that should climatically be more favorable for establishment. There is some evidence in the northeastern U.S. from larger scale studies that species have expanded upslope in montane environments and that these shifts are correlated with local and regional increases in atmospheric temperature (Hamburg and Cogbill 1988; Beckage et al. 2008; Pucko et al. 2011). However, studies of past changes in species composition in New England have also indicated that disturbance and land use history are strongly associated with the decline of some species (e.g., red spruce) and increased abundance of pioneer species such as red maple and paper birch in the post-Colonial era (Foster et al. 1998; Vadeboncoeur et al. 2012). This raises the question of whether the current suite of disturbance-dependent species common across much of southern New England (e.g., birch, red maple, oak and pine) is in equilibrium with historic climatic conditions, and if not, does this introduce bias into species distribution models by not accurately representing the full range of climatic conditions under which more shade-tolerant and generally mesophytic species thrived prior to extensive human disturbance (Nowacki and Abrams 2015)?

Clearly, which tree species might be “losers” or “gainers” in eastern forests under various climate change scenarios depends greatly on the perspectives of different researchers, characteristics of the datasets employed, which region of the country is being analyzed, scale of analyses and how results are interpreted. Nevertheless, most research suggests that rapid warming of the climate over the next century will result in a reduction of optimal climatic conditions for boreal tree species and create more favorable growing conditions for temperate species. This could significantly affect DoD resource management programs throughout the Great Lakes region that currently manage pine (red, eastern white (Pinus strobus) and Jack pine (Pinus banksiana)), aspen-birch and mixed hemlock-northern hardwood forests to provide saw logs, pulpwood, fiber for energy production and wildlife habitat. A key question is the extent to which other species associated with boreal or cold temperate forests that appear to be less at risk (e.g., sugar and red maple, northern red oak, eastern hemlock, eastern white pine) or species at the northern edge of their climatic ranges (e.g., white oak (Quercus alba), black cherry, yellow poplar (Liriodendron tulipifera) will fill niches vacated by the presumed loss of more cold- tolerant species. Our current understanding of potential tree migration rates (approximately 1 km/year maximum) based on studies of post-Pleistocene forest recovery (Overpeck et al. 1991; Davis and Shaw 2001) and modern seed dispersal studies (Clark 1998) indicates that tree species in eastern North America will not be able to keep pace with the rate of projected climate change (average 8 km/year in this study) over the next century (Loarie et al. 2009; Zhu et al. 2012; Snell and Cowling 2015). This suggests that extant tree species in northern forests with greater tolerance for heat-related stress are more likely to capture sites made available by the decline of boreal species than species with northern range boundaries hundreds of kilometers to the south, at least in the near term. Nevertheless, this may result in significant local shifts in species abundance, requiring resource managers to maintain an adaptive strategy when planning long- term projects such as bioenergy generation that may rely on fiber from a particular species mix. For example, if regeneration of aspen-birch stands at Ft. Drum, New York becomes more problematic due to a shift in dominance to more temperate early successional species such red maple or black cherry, this could affect plans for bioenergy production at the installation that are currently based on fast growing aspen-birch stands managed as a coppice. It may be that intensive aspen-birch production could be maintained on certain soil types and moisture regimes 20 where it may have a competitive advantage over more temperate species, but there may also be biological and operational limitations that preclude this approach.

Potential changes in species composition driven by rapid climate change also represent challenges for endangered species and game management programs on installations in the northeastern U.S. forest type changes projected in northern Michigan by Duveneck et al. (2014) suggest a potential decline in the area of Jack pine forests under warmer and wetter future climates conditions, which could have negative impacts on endangered Kirtland warbler (Dendroica kirtlandii) populations managed at Camp Grayling Army National Guard facility. Broader impacts could be seen at a number of installations for a variety of neotropical migratory songbirds that rely on boreal forest habitats for summer breeding and foraging activities (Rodenhouse et al. 2008). Ralston and Kirchman (2013) projected significant declines in 15 northeastern songbird bird species and complete loss of 12 species from montane boreal habitats in New York, Vermont and New Hampshire by 2080. There are few military installations located wholly in the boreal forest biome south of Alaska, but several in the northeastern U.S. are on the southern margins of boreal life zones and are likely to experience significant changes in songbird communities under a rapidly warming climate as habitat for northern species declines and more southern species increase in importance (Rodenhouse et al. 2008). Similar impacts are expected for important game species such as ruffed grouse (Bonasa umbellus) that depend heavily on early successional brushy habitats dominated by birch, aspen and willow (Salix spp.) species. In contrast, more austral game species such as wild turkey (Meleagris gallopavo) could increase in abundance, especially on installations in the upper Midwest if mast producing oak-hickory and oak-savanna habitat types displace northern hardwood communities as implied by projected shifts in Holdridge life zones (Figure 6). Other wildlife species of concern for military installations in the northeast region include the endangered Indiana (Myotis sodalis) and northern long-eared (Myotis septentrionalis) bats. Populations of these species have declined drastically over the past decade due to a fungal pathogen (Pseudogymnoascus destructans) that disrupts bat physiology and behavior while hibernating in caves during the winter, but global warming would appear to benefit bat species at the northern margins of their ranges by reducing exposure-related mortality during the winter by shortening hibernation

21 duration and increasing the abundance of tree species associated with oak-hickory and mixed mesophytic forests (e.g., shagbark hickory (Carya ovata) and white oak) that provide good roosting sites during the summer (Menzel et al. 2001, Silvis et al. 2016). Depending on the amount and seasonality of future precipitation events, insect populations that are the primary food source for bats may also increase with earlier spring emergence and warmer temperatures expected for the region (Hayhoe et al. 2007). Loeb and Winters (2013) modeled potential changes in the summer distribution of Indiana bats using three GCMs and two emissions scenarios and found that increased temperature and decreased precipitation projected for the western portion of the range may cause these areas to become unsuitable for summer maternity colonies within 10 to 20 years. The bulk of the species’ summer roosting range would shift eastward to the Appalachian Mountains and northeastern U.S., potentially increasing management concerns for this species on military installations throughout the region.

The hydrologic and edaphic implications of rapid climate change in the northeastern U.S. are complex and difficult to predict due to uncertainties associated with snow and rainfall projections by different GCMs under different future emissions scenarios. Based on the ensemble of models used in this analysis, precipitation is expected to increase by an average of 10 percent across the region by 2085 under the A2 emissions scenario. This is in agreement with Hayhoe et al. (2007) who projected a 7 to 14 percent increase in mean annual precipitation based on an ensemble of nine GCMs and three emissions scenarios. Precipitation was projected to increase during the winter months and remain unchanged during the summer. Snow cover was projected to decrease in both extent and duration, but average annual soil moisture was expected to increase slightly under a low emissions scenario (B1) or remain unchanged relative to current conditions under a high emissions scenario (A1F1). Surface run-off could therefore increase during the winter months and low streamflow and soil moisture levels that typically occur in the late summer are likely to occur earlier in the season as temperature and evapotranspiration increase (Huntington et al. 2009). Reduced snow cover and wetter soil conditions could result in challenging range management situations and increased environmental impacts for installations that operate heavy vehicles during the winter months. Soils at installations such as Ft. Drum that would normally be covered by up to 1 meter of snow December through February could be

exposed to increased churning and erosion, increasing road maintenance and range rehabilitation costs. Drier soils and fine fuels resulting from higher summer temperatures are likely to increase the risk of wildfires from live fire exercises, but these conditions may also provide opportunities for the use of prescribed fire on installations that heretofore have had little opportunity to use fire as a management tool due to the prevalence of wet to moist soil conditions for most of the year. On well drained, coarse textured soils, prescribed fire could be used at northern installations to maintain red and Jack pine stands, pine-savanna and oak-savanna habitat conditions that benefit many game species, neotropical migrants that prefer open and edge habitats, and endangered species such as the Kirtland’s warbler. Open forest conditions are also preferred for many military training exercises.

Southeastern U.S. Relative to the northeastern U.S., installations in the southeast should see somewhat less dramatic changes in forest species composition and wildlife habitat due to more modest projected increases in mean annual temperature (10.5%) and precipitation (3.5%) across most of the region by 2085 under the A2 emissions scenario. However, temperature regimes in much of the region are projected to shift from warm temperate to subtropical, and while humid conditions are likely to persist in much of the region, significantly lower precipitation is projected for the southern half of the Florida Peninsula and much of Texas. This could result in severe seasonal moisture deficits and increased likelihood of wildfire that may promote the development of subtropical subhumid oak- and pine-shrub savannas in drier portions of the southeast (Clark et al. 2014). This may be largely beneficial to endangered species management on some DoD lands where humid conditions and wildfire suppression have promoted the development of dense woody shrub thickets that are not optimal habitat for some species (e.g., the Florida scrub jay, Aphelocoma coerulescens, at Avon Park Air Force Range in central Florida and the golden- cheeked warbler, Setophaga chrysoparia, at Fort Hood and Camp Bullis in central Texas). However, intensification of subtropical to tropical subhumid climates in central Texas could also shift the species composition of some woody shrub communities to favor species that are more typical of ecosystems in western and southern portions of the state (e.g., conversion of mixed oak woodland-grassland habitats comprised of live oak (Q. virginiana), Ashe’s juniper (Juniperus

23 ashei) and mesquite (Prosopis glandulosa) to semiarid communities dominated by creosote bush (Larrea tridendata) or subtropical mesquite-blackbrush (Vachellia rigidula) thorn scrub), which could be detrimental to species of concern such as the black-capped vireo (Vireo atricapilla) (Grzybowski et al. 1994; McFarland et al. 2013).

Pine-dominated communities from Virginia to Texas are in general expected to respond favorably to warming trends and many pine species have been projected to significantly expand their ranges northward (Iverson and Prasad 1998). However, Clark et al. (2014) and Klos et al. (2009) found that many pine species, including loblolly pine (Pinus taeda), are more sensitive to drought than many southern oak and mesophytic hardwood species and may exhibit higher mortality on upland sites with increasing frequency or intensity of droughts. In addition, Klos et al. (2009) suggested that forest stands with higher species richness and lower densities had lower mortality and better growth responses under various drought conditions in comparison to monospecific stands. This suggests that loblolly pine plantations throughout the South may be more at risk from regional warming and increased evapotranspiration than mixed species forests comprised of more drought tolerant species such as longleaf (Pinus palustris) and slash pine (P. elliotti), or southern oaks and hickories. Depending on how fire regimes are affected and local edaphic conditions, installations in the southeastern Piedmont may experience increasing encroachment by mesophytic hardwoods (e.g., red maple, American beech, flowering dogwood (Cornus florida), sweetgum, southern magnolia (Magnolia grandiflora), and holly (Ilex spp.) in pine stands due to projected increases in temperature and precipitation. A trend towards increasing dominance of mesophytic species has been documented in many eastern forests, although the reason for changes in species composition has been attributed more to the lack of disturbance (primarily anthropogenic fire) rather than increased precipitation related to climate change (Alexander and Arthur 2014; Hanberry et al. 2012; Nowacki and Abrams 2008; 2015). Many southeastern pine-dominated forests are maintained through natural and prescribed fires that kill fire-intolerant hardwoods and reduce competition for pine regeneration from woody shrubs, forbs and grasses. The higher temperatures and increased evapotranspiration expected under projected climate change scenarios would appear to support continued use of prescribed fire on installations throughout the southern U.S. However, expected increases in precipitation in

24 southeastern states, especially during winter months when many prescribed fires are implemented, may offset temperature increases making burning more difficult and less effective in suppressing hardwood growth. It may be that on certain landscape types (e.g., fine textured soils in pine flatwoods), higher soil and fine fuel moistures resulting from increased precipitation will reduce the effectiveness of prescribed fire to the point that shade-tolerant woody shrubs and hardwood tree species become more competitive.

Shifts from pine to hardwood-dominated systems on many southern installations would have important implications for timber production and management of endangered species such as the red-cockaded woodpecker (Leuconotopicus borealis) that depend on pine-dominated ecosystems with open understories. However, sites characterized by coarse textured soils, such as southeastern Sand Hill communities, may be less affected by increased precipitation and maintenance of longleaf pine communities using prescribed fire may remain a relatively easy task for land managers at installations such as Ft. Bragg, North Carolina, Ft. Jackson, South Carolina and Ft. Benning, Georgia. These installations contain significant areas of Sand Hill communities where local edaphic conditions may help maintain xerophytic communities despite the increasingly humid climate projected for the region. Clark et al. (2014) suggested that significant warming in the southeastern U.S. under more extreme CO2 emissions scenarios could push southern forests on more xeric sites towards subtropical savanna-like vegetation, which could potentially reverse the effects of mesophication and help restore pine and oak savannas on many upland landscapes throughout the region. In effect, increasing temperatures and more arid soil moisture conditions could help restore the physiognomy of many upland forest ecosystems across much of the South that existed prior to widespread conversion to cropland and loblolly pine plantations (i.e., a mosaic of southern hardwoods on mesic sites, mixed pine-hardwood forests and more open oak-hickory and longleaf pine forest-savanna; Frost 2006).

Central Prairie-Forest Ecotones Based on the ensemble GCM projections used in this analysis, the northern plains and western Great Lakes region may experience some of the largest increases in mean annual temperature (>20 percent) and mean annual precipitation (> 10 percent) of any area in the coterminous U.S.

25 outside of the Rocky Mountains by late-century (Figure 1). This region is characterized by strong north-south and east-west gradients in temperature and precipitation respectively and as a result encompasses major ecotones between humid boreal and cool temperate forests in the east and the drier grassland communities to the west. The prairie-woodland ecotone in the central U.S. has shifted spatially with variations in climate and fire frequency throughout the Holocene (Williams et al. 2009) and large increases in potential evapotranspiration expected with strongly warming temperatures would appear to favor expansion of prairie and warm temperate woody species over more cold tolerant boreal tree species (e.g., aspen, birches and spruce-fir). However, natural fire regimes that historically favored grassland systems by limiting encroachment from woody vegetation no longer exist in much of the region due to wildfire suppression and landscape fragmentation by agricultural ecosystems. Over the past several decades, woody plants have been encroaching on prairie landscapes at the margins of the prairie-woodland ecotone primarily due to the absence of frequent wildfires that once maintained grassland and open woodland habitats (Radeloff et al. 1999; Briggs et al. 2005; Rogers and Russell 2014). In addition, increased precipitation projected for the region over the next 50 to 70 years may further shift the competitive advantage towards woody warm temperate species resulting in expansion of oak (Quercus alba, Q. rubra)-hickory (Carya cordiformis, C. ovata) and oak (Q. rubra)-maple (A. saccharrum)-basswood (Tilia americana) communities westward into grassland habitats and northward into areas currently dominated by more boreal tree species.

There are relatively few military installations with large land bases in the northern plains, but Army National Guard installations located within prairie-forest ecotones, such as Camp Ripley in central Minnesota and Fort McCoy in Wisconsin, should expect increasing encroachment of woody plant species into grassland habitats with concomitant negative effects on plant and animal species adapted to open landscapes (e.g., the federally endangered Karner blue butterfly (Lycaeides samuelis) at Fort McCoy; Wood et al. 2011). However, both installations appear to be successfully maintaining grassland and open woodland conditions through a combination of prescribed fire, mechanical removal of woody plants and herbicide applications (Minnesota Department of Natural Resources and Minnesota Army National Guard 2015; Wood et al. 2011). Increasing precipitation from climate change and lack of frequent wildfires in most of the Upper

Midwest will in general promote the establishment and growth of temperate mesophytic woody species, but projected increases in regional temperature regimes will also increase evapotranspiration and likely produce significant soil moisture deficits during the growing season in some years. Maintenance of open oak woodland communities on more xeric sites (e.g., sand plains) should be possible if the relationships between fire frequency, site conditions and competition can be fully understood and appropriate management prescriptions implemented. Species that depend on open, grass-dominated communities may be more at risk under expected future climate conditions, but increased forest cover and increased woody species diversity resulting from warmer and more humid conditions should be beneficial to many mammal (e.g., forest bats, white-tail deer (Odocoileus virginianus), migratory songbird and herptofauna species at these installations.

The southern and western portions of the Great Plains present a distinctly different scenario with mean annual temperatures projected to increase by as much as 15 percent and mean annual precipitation projected to decline by 5 to 10 percent in areas from eastern Colorado to central Texas. Ecosystems in the region are likely to experience abnormally long and intense periods of drought by mid-century that exceed any that have occurred in the past 1,000 years (Cook et al. 2015; Williams et al. 2012), which could shift subhumid, cool temperate shortgrass prairie communities (blue grama (Bouteloua gracilis)-buffalograss (Bouteloua dactyloides)-galleta (Hilaria jamesii) in the western high plains towards warm temperate, semi-arid grassland (black grama, Bouteloua eriopoda; tobosagrass, Pleuraphis mutica) and shrub-steppe communities (e.g., mesquite-creosote bush, pinyon (Pinus edulis)-juniper (Juniperus monosperma)) more typical of regions to the south and southwest. Fort Carson, Colorado and Ft. Sill, Oklahoma could see declines in temperate grassland communities and expansion of xeromorphic shrubs and other plants characteristic of more arid biomes, which may in the short term increase cover type and species diversity and potentially improve habitat structure for some species of concern (e.g., golden-cheeked warbler, black-capped vireo and lesser prairie-chicken (Tympanuchus pallidicinctus) at Ft. Sill, all of which prefer or require some woodland-savanna habitat. However, shortgrass prairie communities in the southern Great Plains have declined significantly over the past 100 years due to overgrazing by cattle, conversion to cropland and urban

27 development. Further reductions in the areal extent or ecological function of extant native grassland and woodland-savanna habitats on these installations are likely to have negative effects on populations of black-tailed prairie dogs (Cynomys ludovicianus) and other small mammals that are important food resources for a number of at-risk predator species (e.g., black-footed ferret (Mustela nigripes), swift fox (Vulpes velox velox) and ferruginous hawk (Buteo regalis)).

Along the eastern Front Range and foothills of the Rocky Mountains, expected decreases in soil moisture under warmer and drier climate conditions are likely to reverse the expansion of ponderosa pine (Pinus ponderosa) forests into the western Great Plains that has been occurring over the past several decades (Kaye et al. 2010). By the end of the century, pine forests will likely be restricted to sheltered canyons and higher elevations where more mesic conditions may persist. This could have significant consequences for Mexican spotted owl (Strix occidentalis) populations at Ft. Carson if development and expansion of woody shrub habitat (e.g., pinyon Pinus-juniper Juniperus spp.) does not occur fast enough or is an inadequate substitute for pine forests that provide better cover and nesting sites. The owl is at the northern limit of its range in southeastern Colorado and the climatic conditions expected to develop over the next 50 to 100 years may remain within the environmental tolerances of the species in terms of thermo- regulated behaviors such as fecundity and roost selection (Peery et al. 2012). However, it seems likely that spotted owls will be forced to migrate to higher elevations as forested habitat becomes less dense and more fragmented at lower elevations.

Western Montane Forests, Southwest Deserts and Great Basin Woodlands Given the topographic complexity of the western USA, the most significant biome shifts in the interior portions of the region are likely to be driven by changes in temperature and moisture patterns related to altitude. A number of studies have identified upslope shifts in species and forest communities in the Rockies (Elliott 2012) and Sierra Nevada Mountains (Van Mantgem and Stephenson 2007; Kelly and Goulden 2008) that were attributed to recent warming trends, but downslope movement of tree species has also been reported (Crimmins et al. 2011). Projected changes in montane forest communities due to climate change will likely be more complex than simple upward movement of tree species along elevational gradients.

Physiographic and edaphic conditions are highly variable in mountainous areas and variation in the amount and distribution of snowfall will influence how high elevation species respond to changing climate conditions (Malanson et al. 2007). In this study, projected changes in mean annual temperatures for the western USA by 2085 under the high emissions A2 scenario resulted in almost complete loss of high elevation life zones (Figure 9), which is consistent with estimates from other studies (e.g., greater than 97 percent loss of subalpine and alpine habitats; Finch 2012).

There are relatively few military installations at high elevations in the western U.S., but semiarid woodland and steppe-shrub communities at lower elevations are also at risk from projected increases in mean annual temperature and drought over the next century. Cool temperate, semiarid biomes characterized by big sagebrush (Artemisia tridentata) communities are projected to be displaced northward by 200 to 300 km and contract in area by 40 to 69 percent (Rehfeldt et al. 2006, Finch 2012). Military ranges located near the southern limits of sagebrush distribution in Nevada and Utah could experience encroachment from more drought tolerant communities such as desert scrub and pinyon-juniper woodlands that would be detrimental to several species of concern including the greater sage‐grouse (Centrocercus urophasianus) and pygmy rabbit (Brachylagus idahoensis) (Blomberg et al. 2012; Finch 2012). The Yakima Training Center in Washington and Mountain Home Air Force Base in Idaho are projected to remain at least marginally in a cool temperate, steppe-shrub life zone and could become focal points for conservation of sagebrush communities and their obligate wildlife species. Extensive drought- induced mortality has been documented in pinyon-juniper systems in the Southwest and these communities may be further impacted by increased intensity and duration of future droughts (Breshears et al. 2005; Williams et al. 2012). This vegetation type may expand in the Great Basin and Front Range of Colorado with warmer temperatures, but will likely be replaced by more arid desert scrub and grassland communities in northern Arizona and New Mexico (Finch 2012). This could impact at risk bird species such as the pinyon jay (Gymnorhinus cyanocephalus) and gray vireo (Vireo vicinior) that occur at several DoD installations in the region (e.g., Kirkland Air Force Base and Army National Guard Camel Tracks Training Area, New Mexico; Johnson et al. 2011).

Warm temperate and subtropical desert life zones will also be affected by rising temperatures, although predictions of changes in species distributions based on bioclimatic models remain approximate due to uncertainties associated with future precipitation patterns (e.g., summer monsoon) and winter temperature regimes (Abatzoglou and Kolden 2011). However, in general, climate in the southwestern U.S. is expected to become substantially warmer and drier (super- arid in the Holdridge Life Zone classification) with periodic drought conditions exceeding those experienced in the past several hundred years (Seager et al. 2007). Species whose ranges are currently limited by freezing temperatures are expected to migrate northward and to higher elevations in southwest mountains, eventually displacing more cold-tolerant species and plant communities. At lower elevations, desert scrub communities are projected to continue as the dominant vegetation type and may expand northward and eastward (Weiss and Overpeck 2005) while isolated montane forests (e.g., Madrean forest and woodlands) are projected to completely disappear over time (Rehfeldt et al. 2006). There is concern that milder winters in combination with more intense summer drought conditions will favor the establishment of frost-intolerant, exotic annual grasses (e.g., Bromus rubens, Pennisetum ciliare, Eragrostis lehmanniana) in the Mojave and Sonoran deserts that would introduce surface fuels and potentially establish fire regimes that most native desert species are not adapted to. Warmer temperatures could also alter synchronicity between pollinators and flowering plants potentially affecting a number of insect, bird and bat species (Memmott et al. 2007). Bagne and Finch (2012; 2013) identified a number of animal and plant species at the Barry M. Goldwater Range and Fort Huachuca, Arizona that illustrate the potentially wide-ranging impacts of climate change on military installations located in southwest desert ecosystems, including the endangered Mexican long-nosed bat (Leptonycteris nivalis), lesser long-nosed bat (Leptonycteris yerbabuenae) and Sonoran pronghorn antelope (Antilocapra americana sonoriensis).

Conclusions Based on projections from an ensemble of 16 global climate models, military installations within humid boreal to cool temperate forest biomes in the northeastern U.S. and subhumid cool temperate steppe biomes in the northern Great Plains are most vulnerable to significant shifts in mean annual temperature and/or mean annual precipitation over the next 70 years. Installations

30 such as Minot Air Force Base in North Dakota, Camp Riley in Minnesota, Fort McCoy in Wisconsin, Fort Drum in upstate New York and the Fort Carson complex in eastern Colorado were projected to experience larger proportional changes in mean annual temperature in comparison to installations in other regions of the country. Installations in the northern Great Plains and Lake States could see increases in mean annual precipitation on the order of 10 to 13 percent, which may mitigate some aspects of increasing temperature on biological systems. Fort Carson and other installations in the western Great Plains and southern California are projected to receive 5 to 10 percent less precipitation by 2085 under the A2 emissions scenario. All biomes in North America are projected to be impacted by increasing mean annual temperatures to some degree, but biological systems in relatively colder climates at higher latitudes and at higher elevations may experience the largest proportional increases in mean annual temperature and mean annual evapotranspiration in comparison to current conditions.

Significant impacts on biological systems are likely even under less severe future emissions scenarios (A1B and B1 relative to A2) and over shorter time frames. Significant northward shifts of major life zones in the eastern U.S. may occur as early as 2055 under all emission scenarios. This is presumably due to how concentrations of emissions are modeled in IPCC scenarios that indicate CO2 continuing to increase at a rapid rate over the next few decades before moderating (A1B scenario) or declining (B1 scenario) late in the century (Price et al. 2013). Even under the less severe B1 emissions model, over half of the boreal life zone is projected to be lost from the eastern U.S. and subtropical conditions may expand from the immediate Gulf Coast to eastern North Carolina and southeastern Oklahoma by 2055.

Most military installations do not encompass large elevational gradients, especially those located in the eastern U.S., and are therefore not likely to experience wholesale changes in tree species assemblages in the next 50 years. However, it is likely that the relative frequency and abundance of some tree species will change and will adversely affect boreal species more than temperate species. As a result, installations in boreal and cold temperate life zones should implement periodic forest inventories and biological monitoring programs to determine the extent to which projected changes in species composition and other ecosystem properties are occurring. Despite the projected rapid changes in atmospheric temperature, current research indicates that most 31 changes will occur slowly and may be difficult to observe. However, there is also evidence that some species may have temperature-related “tipping points” where various biological processes (e.g., root growth, seed production) are disrupted and could significantly affect growth rates and regeneration over a short time span as ecological thresholds are approached. It may be appropriate to closely monitor populations of cryophytic genera (e.g., Abies, Picea, Betula, Pinus (red and Jack pine), Populus (except P. deltoides), and Thuja) that are at the southern extent of their geographic ranges on many northern installations to detect potential problems with mortality, growth or regeneration (Figure 10). Changes in species composition should have relatively little impact on military training objectives as long as basic structural components of land cover remain manageable (e.g., relative amounts of open land, shrub-dominated and closed forest conditions required for different types of training operations). Training does not depend on a particular species mix as much as a particular vegetation physiognomy.

Plant and animal communities at DoD installations located within a specific Holdridge Life Zone may have approximately similar responses to changing climate conditions, but uncertainty in projections of precipitation by global climate models and lack of knowledge about how species (and individual trees) will respond suggests that resource managers must plan for multiple possible outcomes (Figure 11, B). If predicted increases in precipitation do not materialize in ecoregions where future mean annual temperatures are expected to be relatively high (e.g., the southeastern U.S.) or if species respond differently than predicted by coarse-scale bioclimatic models, forest communities may development in very different directions and create novel communities with few analogues in the continental United States. For example, despite the current humid climate and potential for increased precipitation projected by some climate models in the southeastern U.S., Clark et al. (2014) and McNulty et al. (2013) suggest that higher evapotranspiration rates, more intense drought events and increased frequency of wildfires in the future could drive southern pine forests towards subtropical pine-savanna vegetation, which would have profound impacts on biological communities, economies and land use patterns in the region.

References Abatzoglou, J. T., and C. A. Kolden. 2011. Climate change in western US deserts: potential for increased wildfire and invasive annual grasses. Rangeland Ecology & Management 64:471- 478.

AchutaRao, K., C. Covey, C. Doutriaux, M. Fiorino, P. Gleckler, T. Phillips, K. Sperber, and K. Taylor. 2004. An appraisal of coupled climate model simulations. D. Bader, ed. Lawrence Livermore National Laboratory, UCRL-TR-202550, August 16, 2004, 197pp. Available at http://www-pcmdi.llnl.gov/appraisal.php).

Alexander, H. D., and M. A. Arthur. 2014. Increasing red maple leaf litter alters decomposition rates and nitrogen cycling in historically oak-dominated forests of the eastern U.S. Ecosystems 17:1371-1383.

Araujo, M. B., R. G. Pearson, W. Thuiller, and M. Erhard. 2005. Validation of species–climate impact models under climate change. Global Change Biology 11:1504-1513.

Arguez, A., I. Durre, S. Applequist, R.S. Vose, M.F. Squires, X. Yin, R.R. Heim, Jr., and T.W. Owen. 2012. NOAA's 1981-2010 U.S. Climate Normals: an overview. Bulletin of the American Meteorological Society 93:1687-1697.

Bagne, K.E. and D.M. Finch. 2012. Vulnerability of species to climate change in the Southwest: threatened, endangered, and at-risk species at the Barry M. Goldwater Range, Arizona. Gen. Tech. Rep. RMRS-GTR-284. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 139 p.

Bagne, K.E. and D.M. Finch. 2013. Vulnerability of species to climate change in the Southwest: threatened, endangered, and at-risk species at Fort Huachuca, Arizona. Gen. Tech. Rep. RMRS-GTR-302. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 183 p.

Bailey, R.G. 2004. Identifying ecoregion boundaries. Environmental Management, 34(S1), S14- S26.

Balmaseda, M. A., K. E. Trenberth, and E. Källén. 2013. Distinctive climate signals in reanalysis of global ocean heat content. Geophysical Research Letters 40:1754-1759.

Beckage, B., B. Osborne, D.G. Gavin, C. Pucko, T. Siccama, and T. Perkins. 2008. A rapid upward shift of a forest ecotone during 40 years of warming in the Green Mountains of Vermont. Proceedings of the National Academy of Sciences of the United States of America 105(11):4197-4202

Bergengren, J.C., D.E.Waliser, and Y.L. Yung. 2011. Ecological sensitivity: a biospheric view of climate change. Climatic Change 107:433–457.

Blomberg, E. J., J. S. Sedinger, M. T. Atamian, and D. V. Nonne. 2012. Characteristics of climate and landscape disturbance influence the dynamics of greater sage-grouse populations. Ecosphere 3(6):55.

Bonan, G. B. 2008. Forests and climate change: forcings, feedbacks, and the climate benefits of forests. Science 320:1444-1449.

Botkin, D. B., H. Saxe, M. B. Araujo, et al. 2007. Forecasting the effects of global warming on biodiversity. BioScience 57:227-236.

Braun EL. 1950. Deciduous forests of eastern North America: The Blakiston Co. 596 p.

Breshears, D.D., N.S. Cobb, P.M. Rich, et al. 2005. Regional vegetation die-off in response to global-change type drought. Proceedings of the National Academy of Sciences, USA. 102: 15144–15148.

Briggs, J. M., A. K. Knapp, J. M. Blair, J. L. Heisler, G. A. Hoch, M. S. Lett, and J. K. McCarron. 2005. An ecosystem in transition: causes and consequences of the conversion of mesic grassland to shrubland. BioScience 55:243-254. 34

Block, W.M., J.L. Ganey, P.E. Scott, and R. King. 2005. Prey ecology of Mexican spotted owls in pine–oak forests of northern Arizona. Journal of Wildlife Management 69(2): 618-629.

Caliński, T. and J. Harabasz. 1974. A dendrite method for cluster analysis. Communications in Statistics 3(1):1-27.

Campbell, J. L., L. E. Rustad, E. W. Boyer, et al. 2009. Consequences of climate change for biogeochemical cycling in forests of northeastern North America. This article is one of a selection of papers from NE Forests 2100: A synthesis of climate change impacts on forests of the northeastern U.S. and eastern Canada. Canadian Journal of Forest Research 39:264- 284.

Canham, C.D. and R.Q. Thomas. 2010. Frequency, not relative abundance, of temperate tree species varies along climate gradients in eastern North America. Ecology 91(12):3433-3440.

Clark, J.S. 1998. Why trees migrate so fast: confronting theory with dispersal biology and the paleorecord. The American Naturalist, 152, 204–224.

Clark, J.S. et al. 2001. Ecological forecasts: an emerging imperative. Science 293 (5530) (Jul. 27, 2001): 657-660.

Clark, J.S., A.E. Gelfand, C.W. Woodall and K. Zhu. 2014. More than the sum of the parts: forest climate response from joint species distribution models. Ecological Applications 24(5): 990–999.

Cleland, D.T.; P.E. Avers, W.H. McNab, M.E. Jensen, R.G. Bailey, T. King, and W.E. Russell, 1997. National Hierarchical Framework of Ecological Units. Pp. 181-200 in: Boyce, M. S.; Haney, A. (eds.), Ecosystem Management Applications for Sustainable Forest and Wildlife Resources. Yale University Press, New Haven. Data accessed August 2015 from USDA Forest Service Geodata site at http://data.fs.usda.gov/geodata/edw/datasets.php.

Colorado Natural Heritage Program. 2005. Ecological system descriptions and viability guidelines for Colorado. Colorado Natural Heritage Program, Colorado State University, Fort Collins, Colorado. Ecological descriptions and maps accessed at http://www.cnhp.colostate.edu/download/projects/eco_systems/eco_systems.asp.

Comer, P.J., D. Faber-Langendoen, R. Evans, S.C. Gawler, C. Josse, G. Kittel, S. Menard, M. Pyne, M. Reid, K. Schulz, K. Snow, and J. Teague. 2003. Ecological systems of the United States: a working classification of U.S. terrestrial systems. NatureServe, Arlington, VA.

Cook, B.I., T.R. Ault, and J.E. Smerdon. 2015. Unprecedented 21st century drought risk in the American Southwest and Central Plains. Science Advances 12 Feb 2015; 1:e1400082.

Crimmins, S. M., S. Z. Dobrowski, J. A. Greenberg, J. T. Abatzoglou, and A. R. Mynsberge. 2011. Changes in climatic water balance drive downhill shifts in plant species’ optimum elevations. Science 331:324-327.

Dale, V.H., M.L. Tharp, K.O. Lannom, and D.G. Hodges. 2010. Modeling transient response of forests to climate change. Science of the Total Environment 408:1888-1901.

Daly, C., R.P. Neilson, and D.L. Phillips. 1994. A statistical-topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140-158.

Daly, C., M. Halbleib, J.I. Smith, W.P. Gibson, M.K. Doggett, G.H. Taylor, J. Curtis, and P.A. Pasteris. 2008. Physiographically-sensitive mapping of temperature and precipitation across the conterminous United States. International Journal of Climatology, 28: 2031-2064.

Delcourt, P. A., and H. R. Delcourt. 1998. Paleoecological insights on conservation of biodiversity: a focus on species, ecosystems, and landscapes. Ecological Applications 8:921- 934.

Deser, C., R. Knutti, S. Solomon and A.S. Phillips. 2012. Communication of the role of natural variability in future North American climate. Nature Climate Change 2, 775–779.

Doe, W. W., R. B. Shaw, R. G. Bailey, D. S. Jones, and T. E. Macia. 1999. Locations and environments of U.S. army training and testing lands: An ecoregional framework for assessment. Federal Facilities Environmental Journal 10:9-26.

Dukes, J.S., J. Pontius, D. Orwig, et al. 2009. Responses of insect pests, pathogens, and invasive plant species to climate change in the forests of northeastern North America: What can we predict? in NE Forests 2100: A Synthesis of Climate Change Impacts on Forests of the Northeastern U.S. and Eastern Canada. Canadian Journal of Forest Research 39:231-248.

Duque, J. C., R. Ramos, and J. Suriñach. 2007. Supervised regionalization methods: a survey. International Regional Science Review 30:195-220.

Duveneck, M.J., R.M. Scheller, M.A. White, S.D. Handler, and C. Ravenscroft. 2014.Climate change effects on northern Great Lake (USA) forests: A case for preserving diversity. Ecosphere 5(2): art23.

Dyer, J. M. 2006. Revisiting the deciduous forests of eastern North America. BioScience 56:341- 352. Spatial data accessed 7/6/2015 at http://www.ohio.edu/ people/ dyer/forest_regions.html.

Elliott, G.P. 2012. Extrinsic regime shifts drive abrupt changes in regeneration dynamics at upper treeline in the Rocky Mountains, USA. Ecology, 93(7), 2012, pp. 1614-1625

Emanuel, W.R., H.H. Shugart, and M.P. Stevenson. 1985. Climate change and the broad-scale distribution of terrestrial ecosystem complexes. Climate Change 7: 29–43.

ERDC 2010. Climate change-induced biome shifts and contaminant management for DoD lands, Project Management Plan 6.2/6.3. Unpublished internal document available from U.S. Army Corps of Engineers, Engineer Research and Development Center, Environmental Lab, Vicksburg, MS. 22p.

Fangxing, F., R.S. Bradley, and M.A. Rawlins. 2013. Climate change in the northeastern U.S.: regional climate model validation and climate change projections. Climate Dynamics 43:1- 2, 145-161.

Finch, D. M., ed. 2012. Climate change in grasslands, shrublands, and deserts of the interior American West: a review and needs assessment. Gen. Tech. Rep. RMRS-GTR-285. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 139 p.

Foster, D. R., G. Motzkin, and B. Slater. 1998. Land-use history as long-term broad-scale disturbance: regional forest dynamics in central New England. Ecosystems 1:96-119.

Frost, C. 2006. History and Future of the Longleaf Pine Ecosystem. In: Jose S., E.J. Jokela, D.L. Miller. The longleaf pine ecosystem: ecology, silviculture, and restoration. New York: Springer; 2006;2007 web.

Gibson, J., G. Moisen, T. Frescino, and T. Edwards, Jr. 2014. Using publicly available forest inventory data in climate-based models of tree species distribution: examining effects of true versus altered location coordinates. Ecosystems 17:43-53.

Gibson, W.P., C. Daly, T. Kittel, D. Nychka, C. Johns, N. Rosenbloom, A. McNab, and G. Taylor. 2002. Development of a 103-year high-resolution climate data set for the conterminous United States. In: Proc., 13th AMS Conf. on Applied Climatology, Amer. Meteorological Soc., Portland, OR, May 13-16, 181-183.

Girvetz, E.H., C. Zganjar, G.T. Raber, E.P. Maurer, P. Kareiva, and J.L. Lawler. 2009. Applied climate-change analysis: The Climate Wizard Tool. PLoS ONE 4(12): e8320. doi:10.1371/journal.pone.0008320. Climate data accessed from www.climatewizard.org 2/20/2014.

Gleckler, P. J., K. E. Taylor, and C. Doutriaux. 2008. Performance metrics for climate models, J. Geophys. Res., 113, D06104, doi:10.1029/2007JD008972.

Grzybowski, J. A., D. J. Tazik, and G. D. Schnell. 1994. Regional analysis of black-capped vireo breeding habitats. The Condor 96:512-544.

Guisan, A., and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993-1009.

Halasz, P. 2007. Original image published at: http://en. wikipedia.org /wiki/Image: Lifezones_Pengo.svg. This file is licensed under the Creative Commons Attribution ShareAlike license versions 2.5, 2.0, and 1.0.

Hamburg, S.P., and C.V. Cogbill. 1988. Historical decline of red spruce populations and climatic warming. Nature, 331(6155): 428–431.

Hanberry, B., B. Palik, and H. He. 2012. Comparison of historical and current forest surveys for detection of homogenization and mesophication of Minnesota forests. Landscape ecology 27:1495-1512.

Hansen, A. J., R. P. Neilson, V. H. Dale, C. H. Flather, L. R. Iverson, D. J. Currie, S. Shafer, R. Cook, and P. J. Bartlein. 2001. Global change in forests: responses of species, communities, and biomes: interactions between climate change and land use are projected to cause large shifts in biodiversity. BioScience 51:765-779.

Hayhoe, K., C.P. Wake, T.G. Huntington, L. Luo, M.D. Schwartz, J. Sheffield, E. Wood, B. Anderson, J. Bradbury, A. DeGaetano, T.J. Troy, and D. Wolfe. 2007. Past and future changes in climate and hydrological indicators in the U.S. Northeast. Clim. Dyn. 28: 381– 407.

Hayhoe, K., C.P. Wake, B.T. Anderson, X.-Z. Liang, E.P Maurer, J. Zhu, J.A. Bradbury, A.T DeGaetano, A.M. Stoner, and D.J. Wuebbles. 2008. Regional climate change projections for the Northeast. USA. Mitig. Adapt. Strategies Glob. Change, 13: 425–436.

Herr, A., J. M. Dambacher, E. Pinkard, M. Glen, C. Mohammed, and T. Wardlaw. 2016. The uncertain impact of climate change on forest ecosystems – How qualitative modelling can guide future research for quantitative model development. Environmental Modelling & Software 76:95-107.

Hinzman, L., N. Bettez, W. R. Bolton, et al. 2005. Evidence and implications of recent climate change in northern Alaska and other Arctic Regions. Climatic Change 72(3):251-298.

Hof, C., I. Levinsky, M.B. Araujo, and C. Rahbek. 2011. Rethinking species’ ability to cope with rapid climate change. Global Change Biology, 17, 2987–2990.

Holdridge, L.R. 1947. Determination of world plant formations from simple climatic data. Science 108: 367-368.

Holdridge, L.R. 1959. Simple method for determining potential evapotranspiration from temperature data. Science 130 (3375): 572.

Holdridge, L.R. 1965. Life Zone Ecology, Revised Edition. Tropical Science Center, San Jose, Costa Rica. 149p.

Hollander, M. and D. A. Wolfe. 1973. Nonparametric statistical methods. New York: Wiley & Sons, 503p.

Hovick, T. J., R. D. Elmore, B. W. Allred, S. D. Fuhlendorf, and D. K. Dahlgren. 2014. Landscapes as a moderator of thermal extremes: a case study from an imperiled grouse. Ecosphere 5:art35.

Intergovernmental Panel on Climate Change (IPCC). 2007. Climate Change 2007: Synthesis Report. Contribution of Working Groups I, II and III to the Fourth Assessment Report of the 40

Intergovernmental Panel on Climate Change [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC, Geneva, Switzerland, 104p.

IPCC, 2013: Summary for Policymakers. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

IPCC, 2014: Summary for Policymakers. In: Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Field, C.B., V.R. Barros, D.J. Dokken, K.J. Mach, M.D. Mastrandrea, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken, P.R. Mastrandrea, and L.L. White (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1-32.

Iverson L.R., and A.M. Prasad 1998 Predicting abundance of 80 tree species following climate change in the eastern United States. Ecol Monogr 68:465–485.

Iverson, L.R., A.M. Prasad, S.N. Matthews, and M. Peters. 2008. Estimating potential habitat for 134 eastern U.S. tree species under six climate scenarios. Forest Ecology and Management 254:390–406.

Iverson, L.R and D. McKenzie. 2013. Tree-species range shifts in a changing climate: detecting, modeling, assisting. Landscape Ecology (2013) 28:879–889.

Jia M, D. Liu, K. Song, Z. Wang, and C. Ren. 2012. Mapping biomes of Australia based on the Holdridge Life Zone Model. International Conference on Computer Vision in Remote Sensing. IEEE; 2012:362-365.

Jain, A. K. 2010. Data clustering: 50 years beyond K-means. Pattern Recognition Letters 31 (2010) 651–666.

Johnson, K., L. Wickersham, T. Neville, et al. 2011. Habitat use at multiple scales by pinyon- juniper birds on Department of Defense lands: landscape scale. Natural Heritage New Mexico Publication 10-GTR-360. 49p. Available at https://nhnm.unm.edu/pubs/.

Jung, M., M. Reichstein, P. Ciais, et al. 2010. Recent decline in the global land evapotranspiration trend due to limited moisture supply. Nature 467:951-954.

Kaye, M.W., C.A. Woodhouse, and S.T. Jackson. 2010. Persistence and expansion of ponderosa pine woodlands in the west-central Great Plains during the past two centuries. Journal Of Biogeography, 37(9):1668-1683.

Kelly, A. E., and M. L. Goulden. 2008. Rapid shifts in plant distribution with recent climate change. Proceedings of the National Academy of Sciences of the United States of America 105:11823-11826.

Keys, Jr.,J., C. Carpenter, S. Hooks, F. Koenig, W.H. McNab, W. Russell, and M.L. Smith. 1995. Ecological units of the eastern United States – First Approximation (CD-ROM). Atlanta, GA: U.S. Department of Agriculture, Forest Service. GIS coverage in ARCINFO format, selected imagery, and map unit tables.

Kottek, M., J. Grieser, C. Beck, B. Rudolf, and F. Rubel, 2006: World Map of the Köppen- Geiger climate classification updated. Meteorol. Z. 15:259-263.

Kruskal, W.H. and W.A. Wallis. 1952. Use of ranks in one-criterion variance analysis. Journal of the American Statistical Association 47 (260):583-621.

Landscape Change Research Group. 2014. Climate change atlas. Northern Research Station, U.S. Forest Service, Delaware, OH. http://www.nrs.fs.fed.us/atlas.

Little, E.L., Jr. 1971. Atlas of United States trees, volume 1, conifers and important hardwoods: U.S. Department of Agriculture Miscellaneous Publication 1146, 9 p., 200 maps. Shape files accessed 6/29/2015 at http://esp.cr.usgs.gov/data/little/.

Loarie S.R., P.B. Duffy, H. Hamilton, G.P. Asner, C.B. Field, and D.D. Ackerly. 2009. The velocity of climate change. Nature 462 1052–1055.

Loeb, S. C., and E. A. Winters. 2013. Indiana bat summer maternity distribution: effects of current and future climates. Ecology and Evolution 3:103-114.

Lovett, G. M., C. D. Canham, M. A. Arthur, K. C. Weathers, and R. D. Fitzhugh. 2006. Forest Ecosystem Responses to Exotic Pests and Pathogens in Eastern North America. BioScience 56(5):395-405.

Luce, C. H., J. M. Vose, N. Pederson, J. Campbell, C. Millar, P. Kormos, and R. Woods. 2016. Contributing factors for drought in United States forest ecosystems under projected future climates and their uncertainty. Forest Ecology and Management 380:299-308.

Lugo, A.E., S.L. Brown, R. Dodson, T.S. Smith, and H.H. Shugart. 1999. Special Paper: The Holdridge life zones of the conterminous United States in relation to ecosystem mapping. Journal of Biogeography 26:1025-1038.

Malanson, G. P., et al. 2007. Alpine treeline of western North America: linking organism-to- landscape dynamics. Physical Geography 28:378-396.

Mast, J.N., T.T.Veblen, and Y.B. Linhart. 1998. Disturbance and climatic influences on age structure of ponderosa pine at the pine/grassland ecotone, Colorado Front Range. Journal of Biogeography 25(4):743–755.

Matthews, S., L. Iverson, M. Peters, A. Prasad, and S. Subburayalu. 2014. Assessing and comparing risk to climate changes among forested locations: implications for ecosystem services. Landscape Ecology 29:213-228.

Maurer, E.P., L. Brekke, T. Pruitt, and P.B. Duffy. 2007. Fine-resolution climate change projections enhance regional climate change impact studies. Eos Transactions American Geophysical Union 88(47):504.

McFarland, T. M., H. A. Mathewson, J. E. Groce, M. L. Morrison, and R. N. Wilkins. 2013. A Range-Wide Survey of the Endangered Black-Capped Vireo in Texas. Southeastern Naturalist 12:41-60.

McNulty, S., P. Caldwell, T. Doyle, K. Johnsen, et al. 2013. Forests and Climate Change in the Southeast USA. In: Ingram, K.; Dow, K.; Carter, L.; Anderson, J., eds. 2013. Climate of the Southeast United States: Variability, change, impacts, and vulnerability. Washington, DC: Island Press. 165-189.

McKenney D.W., J.H. Pedlar, M.F. Hutchinson, K. Lawrence, and K. Campbell. 2007. Potential impacts of climate change on the distribution of North American trees. BioScience 57:939– 948.

McKenney D.W., J.H. Pedlar, R.B. Rood, and D. Price. 2011. Revisiting projected shifts in the climate envelopes of North American trees using updated general circulation models. Glob Change Biol 17:2720–2730.

Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stouffer, and K. E. Taylor. 2007. The WCRP CMIP3 multi-model dataset: A new era in climate change research. Bulletin of the American Meteorological Society 88:1383-1394.

Memmott, J., P. G. Craze, N. M. Waser and M. V. Price. 2007. Global warming and the disruption of plant pollinator interactions. Ecology Letters 10: 710-717.

Menzel, A. et al. 2006. European phenological response to climate change matches the warming pattern. Global Change Biology 12:1969-1976.

Menzel, M. A., J. M. Menzel, T. C. Carter, W. M. Ford, and J. W. Edwards. 2001. Review of the forest habitat relationships of the Indiana bat (Myotis sodalis). U.S. Department of Agriculture, Forest Service, Northeastern Research Station. Report NE-284.

Millar, C., N. Stephenson, and S. Stephens. 2007. Climate change and forests of the future: managing in the face of uncertainty. Ecological Applications, 17(8):2145-2151.

Minnesota Department of Natural Resources and Minnesota Army National Guard. 2015. Minnesota Army National Guard, Camp Ripley Training Center and Arden Hills Army Training Site, 2014 Conservation Program Report, January 1-December 31, 2014. Compiled by Nancy J. Dietz and Brian J. Dirks, Camp Ripley Series Report No. 24, Little Falls, MN, USA. 208 pp.

Mohan, J.E., R.M. Cox and L.R. Iverson. 2009. Composition and carbon dynamics of forests in northeastern North America in a future, warmer world. Can. J. For. Res. 39: 213–230.

Moritz, C., and R. Agudo. 2013. The future of species under climate change: resilience or decline? Science 341:504-508.

Nakicenovic, N., J. Alcamo, G. Davis, et al. 2000. IPCC Special Report on Emissions Scenarios. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 599 p.

Nowacki, G. J., and M. D. Abrams. 2008. The demise of fire and "mesophication" of forests in the eastern United States. BioScience. 58(2):123-138.

Nowacki, G. J., and M. D. Abrams. 2015. Is climate an important driver of post-European vegetation change in the Eastern United States? Global Change Biology 21:314-334.

Olson, D.M., E. Dinerstein, E.D. Wikramanayake, et al. 2001. Terrestrial Ecoregions of the World: A New Map of Life on Earth: A new global map of terrestrial ecoregions provides an innovative tool for conserving biodiversity. BioScience 51:933-938.

Omernik, J.M. and G.E. Griffith. 2014. Ecoregions of the conterminous United States: evolution of a hierarchical spatial framework. Environmental Management 54(6):1249-1266.

Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421:37-42.

Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics 37:637-669.

Peery, M. Z., R. J. Gutiérrez, R. Kirby, O. E. LeDee, and W. LaHaye. 2012. Climate change and spotted owls: potentially contrasting responses in the Southwestern United States. Global Change Biology 18:865-880.

Pierce, D. W., T. P. Barnett, B. D. Santer, and P. J. Gleckler. 2009. Selecting global climate models for regional climate change studies. Proceedings of the National Academy of Sciences of the United States of America 106:8441-8446.

Pielke Sr., R.A., et al. 2007. A new paradigm for assessing the role of agriculture in the climate system and in climate change. Agricultural and Forest Meteorology 142(2–4): 234-254.

Post, E., M. C. Forchhammer, M. S. Bret-Harte, et al. 2009. Ecological dynamics across the Arctic associated with recent climate change. Science 325:1355-1358.

Prasad, A.M., L.R. Iverson, and A. Liaw. 2006. Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems 9:181–199.

Prasad, A.M., L.R. Iverson, S. Matthews, and M. Peters. 2007. A climate change atlas for 134 forest tree species of the eastern United States [database]. Northern Research Station, USDA Forest Service, Delaware, OH. www.nrs.fs.fed.us/atlas/tree.

Price, D.T., R.I. Alfaro, K.J. Brown, et al. 2013. Anticipating the consequences of climate change for Canada’s boreal forest ecosystems. Environ. Rev. 21:322–365.

Pucko, C., B. Beckage, T. Perkins, and W. S. Keeton. 2011. Species shifts in response to climate change: Individual or shared responses? The Journal of the Torrey Botanical Society 138:156-176.

Radeloff, V. C., D. J. Mladenoff, H. S. He, and M. S. Boyce. 1999. Forest landscape change in the northwestern Wisconsin Pine Barrens from pre-European settlement to the present. Canadian Journal of Forest Research 29:1649-1659.

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 Sciences 167:1123-1150.

Rodenhouse, N.L., S.N. Matthews, K.P. McFarland, J.D. Lambert, L.R. Iverson, A. Prasad, T.S. Sillett, and R.T. Holmes. 2008. Potential effects of climate change on birds of the Northeast. Mitigation and Adaptation Strategies for Global Change 13(5-6):517-540.

Rodenhouse, N.L., L.M. Christenson, D. Parry, and L.E. Green. 2009. Climate change effects on native fauna of northeastern forests. This article is one of a selection of papers from NE forests 2100: A synthesis of climate change impacts on forests of the northeastern U.S. and eastern Canada. Can. J. For. Res. 39:249–263.

Rogers, T. R., and F. L. Russell. 2014. Historical patterns of oak population expansion in the Chautauqua Hills, Kansas. Journal of Biogeography 41:2105-2114.

Rondeau, R. J., K. T. Pearson, and S. Kelso. 2013. Vegetation response in a Colorado grassland- shrub community to extreme drought: 1999-2010. The American Midland Naturalist, 170(1):14-25.

Rustad, L., J. Campbell, J.S. Dukes, T. Huntington, K. Fallon-Lambert, J. Mohan, and N. Rodenhouse. 2012. Changing climate, changing forests: The impacts of climate change on forests of the northeastern United States and eastern Canada. Gen. Tech. Rep. NRS-99.

Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 48 p.

Seager, R., M. Ting, I. Held, Y. Kushnir, et al. 2007. Model projections of an imminent transition to a more arid climate in southwestern North America. Science 316:1181.

Seneviratne, S. I., D. Luthi, M. Litschi, and C. Schar. 2006. Land-atmosphere coupling and climate change in Europe. Nature 443:205-209.

Silvis, A., R.W. Perry and W.M. Ford. 2016. Relationships of three species of white-nose syndrome-impacted bats to forest condition and management. U.S. Forest Service Southern Research Station General Technical Report. SRS–214, Ashville, NC. 57p.

Sisneros, R., J. Huang, G. Ostrouchov, and F. Hoffman. 2011. Visualizing Life Zone boundary sensitivities across climate models and temporal spans. Procedia Computer Science 4 (2011):1582–1591.

Smith, W. B. 2002. Forest inventory and analysis: a national inventory and monitoring program. Environmental Pollution 116, Supplement 1:S233-S242.

Snell, R. S. and S. A. Cowling. 2015. Consideration of dispersal processes and northern refugia can improve our understanding of past plant migration rates in North America. J. Biogeogr. (2015) 42: 1677–1688.

Thuiller W. 2004. Patterns and uncertainties of species' range shifts under climate change. Global Change Biol. 2004; 10:2020–2027.

Vadeboncoeur M.A., S.P. Hamburg, C.V. Cogbill, and W.Y. Sugimura. 2012. A comparison of presettlement and modern forest composition in central New Hampshire. Canadian Journal of Forest Research 42:190–202.

Van Mantgem, P. J., and N. L. Stephenson. 2007. Apparent climatically induced increase of tree mortality rates in a temperate forest. Ecology Letters 10:909-916. 48

Walther, G.-R., E. Post, P. Convey, A. Menzel, C. Parmesan, T. J. C. Beebee, J.-M. Fromentin, O. Hoegh-Guldberg, and F. Bairlein. 2002. Ecological responses to recent climate change. Nature 416:389-395.

Washington, W. M., J. W. Weatherly, G. A. Meehl, A.J. Semtner, T. W. Bettge, A. P. Craig, W. G. Strand, J. Arblaster, V. B. Wayland, R. James, and Y. Zhang. 2000. Parallel Climate Model (PCM) control and transient simulations. Climate Dynamics 16(10/11): 755-774.

Weiss, J. L. and J. T. Overpeck. 2005. Is the Sonoran Desert losing its cool? Global Change Biology 11:2065-2077.

Wiens, J. A., D. Stralberg, D. Jongsomjit, C. A. Howell, and M. A. Snyder. 2009. Niches, models, and climate change: Assessing the assumptions and uncertainties. Proceedings of the National Academy of Sciences of the United States of America 106:19729-19736.

Williams, A.P., C. D. Allen, A. K. Macalady, et al. 2012. Temperature as a potent driver of regional forest drought stress and tree mortality. Nature Clim. Change 3:292-297.

Williams, J. W., B. Shuman, and P. J. Bartlein. 2009. Rapid responses of the prairie-forest ecotone to early Holocene aridity in mid-continental North America. Global and Planetary Change 66:195-207.

Woodall, C. W., C. M. Oswalt, J. A. Westfall, C. H. Perry, M. D. Nelson, and A. O. Finley. 2009. An indicator of tree migration in forests of the eastern United States. Forest Ecology and Management 257:1434-1444.

Woodall, C.W., C.M. Oswalt, J.A. Westfall, C.H. Perry, M.D. Nelson, and A.O. Finley. 2010. Selecting tree species for testing climate change migration hypotheses using forest inventory data. Forest Ecology and Management 259 (2010) 778–785.

Woolfenden, G.E., and J.W. Fitzpatrick. 1984. The Florida scrub-jay:demography of a cooperative-breeding bird. Princeton University Press, Princeton, New Jersey, USA.

Xu, T. and M.F. Hutchinson. 2013. New developments and applications in the ANUCLIM spatial climatic and bioclimatic modelling package. Environmental Modelling & Software 40:267-279.

Yaussy, D. A., L. R. Iverson, and S. N. Matthews. 2013. Competition and climate affects U.S. hardwood-forest tree mortality. Forest Science 59:416-430.

Zar, J. H. 2010. Biostatistical Analysis, 5th edition. Prentice Hall, Upper Saddle River, NJ.

Zhu K., C. Woodall, and J. Clark. 2012. Failure to migrate: lack of tree range expansion in response to climate change. Glob Change Biol 18:1042–1052.

Appendix A.

Table A-1. Department of Defense installations in the continental United States (Source: Military Installations, Ranges, and Training Areas (point locations and boundaries), Version 1.0, June 21, 2010, accessed 11/16/2012 from http://www.acq.osd.mil/ie/bei/opengov /installations_ranges.zip).

Mean annual Mean annual Code Installation name Latitude Longitude temperature precipitation (oC) (mm)

29PA Twentynine Palms Marine Corps Base 34.4400 -116.1117 20.1 114.3 ABER Aberdeen Proving Ground 39.3415 -76.2908 13.3 1122.7 ADAI Camp Adair Military Reservation 44.7155 -123.2747 11.5 1110.0 ANNI Anniston Army Depot 33.7120 -85.9575 16.9 1318.3 ARLG Arlington National Cemetery 38.8773 -77.0723 13.2 1036.3 AROT Army Reserve Outdoor Training Area 41.1441 -96.4138 10.4 767.1 ATAN Army Training Area (NE) 41.3082 -98.3110 9.9 657.9 ATTE Camp Atterbury Military Reservation 39.3091 -86.0409 11.7 1066.8 AVON Avon Park Air Force Bombing Range 27.6470 -81.2919 22.2 1247.1 BADG Badger Army Ammunition Plant 43.3594 -89.7360 6.3 858.5 BARS Barstow Marine Corps Logistics Base 34.8655 -116.9427 18.7 109.2 BEAR Bearmouth National Guard TA 46.7021 -113.3125 5.9 320.0 BEAU Beaufort Marine Corps Air Station 32.4734 -80.7157 18.7 1264.9 BELL Belle Mead General Depot 40.4882 -74.6682 11.4 1237.0 BELV Ft Belvoir Military Reservation 38.7153 -77.1665 13.2 1036.3 BENN Ft Benning Military Reservation 32.3967 -84.8273 18.4 1234.4 BLIS Ft Bliss 31.8977 -106.2594 18.2 239.5 BOGU Bogue Field 34.6940 -77.0288 17.6 1478.3 BRAG Ft Bragg Military Reservation 35.1521 -79.1380 16.2 1186.2 BUCK Buckley Air National Guard AF Base 39.7073 -104.7639 10.0 401.3 BULL Camp Bullis 29.6778 -98.5810 20.4 835.7 CAMP Ft Campbell 36.6296 -87.6210 14.4 1318.3 CARS Ft Carson Military Reservation 37.4910 -103.8746 9.4 421.6 CHPT Cherry Point Marine Corps Air Sta 34.9239 -76.8953 16.8 1389.4 CNWS Crane Naval Weapons Support Ctr 38.8348 -86.7953 11.9 1127.8 CORN Cornhusker Army Ammunition Plant 40.9254 -98.3866 9.9 657.9 CUSI Cusick Survival Traning Site 48.4917 -117.4233 8.2 556.3 CUST Custer Reserve Forces Training Area 42.3041 -85.3253 8.9 894.1 DARE Dare County Range 35.7167 -75.8846 16.6 1310.6 DETR U.S. Garrison, Ft Detrick 39.4369 -77.4524 13.4 1031.2 DODG Camp Dodge Military Reservation 41.7159 -93.7253 10.0 881.4 DRUM Ft Drum 44.1331 -75.6083 7.4 1082.0 DUGW Dugway Proving Grounds 40.2261 -113.0440 11.2 419.1 EGLN Eglin AFB 30.5822 -86.5494 20.1 1632.7 ELLS Ellsworth AFB 44.1552 -103.0916 7.4 467.4 ETAL Ft Ethan Allen Military Reservation 44.4679 -72.8999 7.3 915.7 EUST Ft Eustis Military Reservation 37.1261 -76.5891 14.8 1247.1 FDIX Ft Dix Military Reservation 39.9942 -74.5360 12.2 1196.8 FLEE Ft Lee Military Reservation 37.2518 -77.3321 13.9 1150.6 FLOR Florence Military Reservation 33.0920 -111.3488 20.9 254.0

Mean annual Mean annual Code Installation name Latitude Longitude temperature precipitation (oC) (mm)

FORD Ft Ord Military Reservation 36.6319 -121.7656 14.1 393.7 GILL Ft Gillem Heliport 33.6219 -84.3292 16.3 1196.3 GOOD Goodfellow AFB 31.4361 -100.3651 18.0 530.9 GORD Ft Gordon 33.3572 -82.2384 17.3 1130.3 GRAY Camp Grayling Military Reservation 44.6178 -84.8820 5.4 848.4 GRFO Grand Forks AFB 47.9608 -97.2473 4.6 497.8 HAAD Hawthorne Army Ammunition Depot 39.4320 -76.1755 13.3 114.3 HIGH Highlands, NC 35.0545 -83.2021 10.3 2225.0 HILL Ft A. P. Hill Military Reservation 38.1210 -77.2668 13.6 1130.3 HNLG Hunter-Liggett Military Reservation 35.9633 -121.1741 15.6 312.4 HOME Homestead AFB 25.4980 -80.3893 23.8 1486.7 HOOD Ft Hood 31.1762 -97.6910 20.2 854.7 HUAC Ft Huachuca 31.5609 -110.3157 17.4 360.7 IOWA Iowa Army Ammunition Plant 40.7893 -91.2503 10.9 962.7 IRWN Ft Irwin 35.3779 -116.6227 20.1 114.0 JACK Ft Jackson 34.0394 -80.8341 18.9 1196.3 JOLI Joliet Army Ammunition Plant 41.3859 -88.1238 9.7 937.3 KAAP Kansas Army Ammunition Plant 37.2917 -95.2067 13.4 1069.3 KING Kingsville NAS 27.4983 -97.8258 22.0 736.6 KNOX Ft Knox 37.8980 -85.9023 13.2 1270.0 LAOP Louisiana Ordnance Plant 32.5651 -93.4000 18.7 1303.0 LAUG Laughlin AFB 29.3594 -100.7826 20.9 475.0 LCAP Lake City Army Ammunition Plant 39.0968 -94.2489 11.9 1097.3 LEAV Ft Leavenworth Military Reservation 39.3613 -94.9095 12.2 1038.9 LEJE Camp Lejeune Marine Corps Base 34.6504 -77.3179 16.8 1389.4 LEOW Ft Leonard Wood Military Reserv 37.7014 -92.1549 12.9 1130.3 LETT Letterkenny Army Depot 40.0199 -77.6972 11.3 1011.7 LEWI Ft Lewis Military Reservation 47.0297 -122.5964 11.5 990.6 LONG Longhorn Ordnance Ammo Plant 32.6695 -94.1368 17.6 1300.5 MACD MacDill AFB 27.8433 -82.5012 22.8 1137.9 MACK Camp MacKall Military Reservation 35.0284 -79.4921 16.2 1186.2 MACR Ft MacArthur 33.7133 -118.2979 17.4 334.0 MALA Malabar Transmitter Annex 28.0228 -80.6797 22.3 1226.8 MALM Malmstrom AFB 47.5082 -111.1868 6.5 378.5 MANG Mead Army National Guard Facility 41.1951 -96.4382 10.4 767.1 MCLS Marine Corps Logistics Support Base 31.5528 -84.0509 18.6 1356.4 MCOY Ft McCoy 44.0303 -90.6853 7.0 838.2 MCPH Ft McPherson 33.7072 -84.4353 16.3 1196.3 MEAD Ft George G. Meade 39.1022 -76.7448 13.3 1178.6 MELR Melrose Air Force Range 34.3096 -103.7883 14.3 469.9 MHAB Mountain Home Air Base 43.0568 -115.8701 10.3 269.2 MILA Milan Arsenal 35.8845 -88.7016 13.7 1379.2 MINO Minot AFB 48.4207 -101.3461 5.3 469.9 MONM Ft Monmouth Military Reservation 40.3129 -74.0471 11.6 1237.0 MONR Ft Monroe Military Reservation 37.0171 -76.2985 15.3 1163.3 MTBA Mount Baker Helicopter Training Area 48.9318 -121.9750 10.1 2057.4 NAAP Newport Army Ammunition Plant 39.7955 -87.4000 12.1 1137.9 NAPE Army Helicopter Training Area 46.7356 -122.6127 11.8 1206.5 NATI Natick Laboratories Military Reserv 42.4012 -71.4719 10.1 1245.9 52

Mean annual Mean annual Code Installation name Latitude Longitude temperature precipitation (oC) (mm)

NAVA Navajo Army Depot (Closed) 35.1886 -111.7848 5.7 589.3 NOLA New Orleans Naval Air Station 29.8452 -89.9938 20.9 1612.9 NSTF Naval Survival Training Facility 44.9865 -70.4465 3.4 1031.2 OAAP U.S. Army Ammunition Depot (OK) 34.8176 -95.9302 16.5 1150.6 OTSP Military Ocean Terminal Sunny Point 34.0082 -77.9715 17.7 1450.3 PAIS Parris Island U.S. Marine Corps 32.3357 -80.7096 18.7 1264.9 PARK Camp Parks Military Reservation 37.7215 -121.8974 15.4 375.9 PEND Camp Pendleton Marine Corps Base 33.3546 -117.4211 16.1 281.9 PIBL Pine Bluff Arsenal 34.3230 -92.0789 16.1 1318.3 PICA Picatinny Arsenal 40.9555 -74.5419 9.1 1219.2 PICK Ft Pickett Military Reservation (VNG) 37.0455 -77.9151 13.8 1165.9 POLK Ft Polk Military Reservation 31.0883 -93.0586 19.2 1602.2 POWE Powell Air Force Sta 44.7693 -108.3048 6.9 170.2 PUEB Pueblo Army Depot 38.3139 -104.3301 11.9 330.2 QUAN Quantico Marine Corps Base 38.5616 -77.4265 13.3 1087.1 RAAP Radford Army Ammunition Plant 37.1847 -80.5454 10.6 1082.0 RAVE Ravenna Arsenal 41.1974 -81.0760 8.8 960.1 REDR Red River Army Depot 33.4258 -94.2983 17.1 1300.5 REDS Redstone Arsenal 34.6285 -86.6026 15.9 1460.5 RILE Camp Riley Military Reservation 46.1308 -123.9451 10.6 1704.3 RILY Ft Riley Military Reservation 39.1669 -96.8112 12.7 883.9 RIPL Camp Ripley MNG 46.0889 -94.3590 4.8 674.0 ROBE Camp Roberts Military Reservation 35.7888 -120.7869 15.6 309.9 ROBI Camp Joseph T. Robinson 34.8885 -92.3095 16.2 1237.0 ROCK Rock Island Arsenal 41.5168 -90.5421 10.6 894.1 RUCK Ft Rucker Military Reservation 31.4119 -85.7408 18.7 1445.3 SAAP Sunflower Army Ammunition Plant 38.9226 -95.0096 13.6 1010.9 SENE Seneca Army Depot 42.7434 -76.8594 8.9 848.4 SHAR Sharpe General Depot (Field Annex) 37.8360 -121.2641 16.1 421.6 SIER Sierra Army Depot 40.2371 -120.2576 9.5 340.4 SILL Ft Sill Military Reservation 34.7024 -98.5176 16.3 802.6 SMOK Smoky Hill ANG Range 38.6986 -97.8128 13.2 817.9 SNOQ Snoqualmie National Forest 47.0468 -121.6122 9.8 1468.1 STEW Ft Stewart 31.9868 -81.5970 19.8 1229.4 STOR Ft Story Military Reservation 36.9212 -76.0192 15.3 1163.3 SWIF Camp Swift N. G. Facility 30.2202 -97.2465 20.2 854.7 TINK Tinker AFB 35.4150 -97.3973 15.6 911.9 TOAD Tooele Army Depot 40.5282 -112.4047 11.0 500.4 UMAT Umatilla Chemical Depot 45.8396 -119.4405 11.7 264.2 VAND Vandenberg AFB 34.7219 -120.5568 15.4 401.3 WIHA Ft William H. Harrison Military Reserv 46.6311 -112.1046 6.7 287.0 WILU Camp Williams, UT 40.4498 -112.0123 11.4 325.1 WILW Camp Williams, WI 43.9384 -90.2521 6.9 845.8 WOLT Ft Wolters 32.8732 -97.9940 19.1 807.7 WSMR White Sands Missile Range 32.9537 -106.4183 16.2 335.3 WSPT West Point U.S. Military Academy 41.3639 -74.0300 10.6 1287.8 YAKI Yakima Firing Center 46.7098 -120.1944 9.4 210.8 YUMA Yuma Proving Ground 33.1587 -114.4203 23.9 78.7

Table A-2. General Circulation Models (GCM) available at Climatewizard.org that were used collectively to generate temperature and precipitation projections to formulate Holdridge Life Zones (Girvetz et al. 2009).

Model Country Organization

Bjerknes Centre for Climate Research BCCR-BCM2.0 Norway Canadian Centre for Climate Modelling &

CGCM3.1(T47) Canada Analysis Météo-France / Centre National de

CNRM-CM3 France Recherches Météorologiques

CSIRO-Mk3.0 Australia CSIRO Atmospheric Research U.S. Dept. of Commerce / NOAA /

GFDL-CM2.0 USA Geophysical Fluid Dynamics Laboratory U.S. Dept. of Commerce / NOAA /

GFDL-CM2.1 USA Geophysical Fluid Dynamics Laboratory

GISS-ER USA NASA / Goddard Institute for Space Studies

INM-CM3.0 Russia Institute for Numerical Mathematics

IPSL-CM4 France Institute Pierre Simon Laplace Center for Climate System Research (The University of Tokyo), National Institute for

MIROC3.2(medres) Japan Environmental Studies, and Frontier Research Center for Global Change (JAMSTEC) Meteorological Institute of the University of

ECHO-G Germany / Korea Bonn, Meteorological Research Institute of KMA, and Model and Data group.

ECHAM5/MPI-OM Germany Max Planck Institute for Meteorology

MRI-CGCM2.3.2 Japan Meteorological Research Institute

CCSM3 USA National Center for Atmospheric Research

PCM USA National Center for Atmospheric Research Hadley Centre for Climate Prediction and

UKMO-HadCM3 UK Research / Met Office

Table 1. Area of six largest Holdridge life zones in the coterminous U.S. based on mean climate conditions from 1951-2006. Cumulative Area Area Life zone Associated vegetation 2 area (km ) (Percent) (Percent) Warm temperate humid forest Southeastern mixed forest; oak-hickory forest 1,116,576 14.7 14.7 (5531) Cool temperate subhumid steppe High plains; short-grass and sagebrush 873,936 11.5 26.2 (434) Boreal and subalpine perhumid- Boreal/subalpine spruce-fir forest 861,264 11.4 37.6 superhumid forest (351, 352) Cool temperate humid forest (443) Beech-maple forest; oak-hickory savanna; long- 829,584 10.9 48.5 grass prairie Boreal humid-subhumid forest- Cold short-grass prairie; sagebrush steppe 491,472 6.5 55.0 steppe (333) Cool temperate humid forest (453) Oak-hickory and mixed mesophytic forests 453,888 6.0 61.0

1 Numbers represent MAT, MAP and PEVT classes in the Holdridge system. Larger values indicate higher temperature, precipitation and evapotranspiration regimes.

Table 2. Change in locationa and area of Holdridge life zones in eastern North America from 2006 to 2085 under the A2 emissions scenario.

------Present ------2085-A2 ------Change metrics Representative Centroid Rate of Holdridge Life Zone Approximate ECOMAP Unit installation Area Longitude Latitude Area Longitude Latitude Longitude Latitude Area shift shift (currently) (km2) (deg) (deg) (km2) (deg) (deg) (deg) (deg) (%) (km) (km/yr)

Boreal and montane M212 New England-Adirondack Camp Ethan Allen, 151,056 -78.8 46.1 3600c -74.5 50.2 4.3 4.1 -98 570 7 superhumid forest Mixed Forest NY 241/251/351 Boreal perhumid forest 212 Laurentian Mixed Forest Camp Grayling, MI 393,264 -87.2 45.3 288 -80.7 49.3 6.5 4.0 -100 632 8 242/342/352 Fort Drum, NY

Cool temperate and 221 Eastern Broadleaf Forest West Point, NY 311,904 -79.6 41.5 251,568 -73.9 45.7 5.7 4.2 -19 656 8 montane perhumid forest (Oceanic) and M221 Central 442/452 Appalachian Broadleaf- Coniferous Forest Cool temperate humid 222 Eastern Broadleaf Forest Ravenna AAP, OH 544,608 -89.9 41.3 479,520 -91 44.7 -1.1 3.4 -12 526 7 forest 443 (Continental) and 251 Temperate Prairie Parkland Cool temperate humid 221/222 Humid Temperate Camp Atterbury, 433,728 -85.7 39.6 110,448 -73.3 43.9 12.4 4.3 -75 1127 14 forest 453 Eastern Broadleaf Forest IN

Warm temperate humid 231 Subtropical Southeastern Fort Bragg, NC 1,104,624 -87.4 34.7 967,104 -84.9 40 2.5 5.3 -12 632 8 forest 553 Mixed Forest, 232 Coastal Plain Fort Benning, GA Mixed Forest Fort Campbell, TN

Warm temperate subhumid 251 Temperate Prairie Parkland Tinker AFB, OK 21,312 -97.1 36.8 318,240 -89.8 42.2 7.3 5.4 1393 871 11 steppe-woodland 543

Subtropical humid forest 232 Subtropical Coastal Plain Eglin AFB, FL 233,280 -86.8 29.5 827,856 -86.5 33.7 0.3 4.2 255 463 6 653 Mixed Forest and 234 Lower Mississippi Riverine Forest Subtropical subhumid forest 255 Subtropical Prairie Parkland Avon AF Range, FL 45,216 -88.6 28.1 223,632 -90.7 32.4 -2.1 4.3 395 511 6 654 and 232 Coastal Plain Mixed Forest Tropical subhumid dry 400 Humid Tropical Savanna NAS Key West, FL 3,367 -81.5 24.6 72,144 -81.6 27.6 -0.1 3.0 2043 317 4 forest 754 Means 4 4 631 8 a Coordinates for centroids of ellipses that approximate the shape and distribution of predicted life zone classes under current and future (2085) climate conditions. b Keys et al. 1995. c Values in shaded cells derived from CMIP3 PCM1.4 model (Maurer et al. 2007) instead of 16 GCM ensemble from ClimateWizard.org (Girvetz et al. 2009). d Area of subpolar life zones combined with boreal zones in northern Minnesota and Maine.

Table 3. Percent change in mean annual temperature and mean annual precipitation projected from 2006 to 2085 for five groups of military installations identified through k-means cluster analysis. Cluster numbers correspond to Figure 7a, b and d.

Mean change from 2006 to 2085 ± SD General U.S. Region Cluster N MAT (%) MAP (%)

5 122 17.4 ± 1.3 10.0 ± 3.1 Northeast, Northern Plains 2 54 15.6 ± 1.6 -5.9 ± 3.5 Central and Western High Plains 3 145 13.8 ± 1.0 11.1 ± 2.0 Mid-Atlantic, Pacific NW 1 121 10.8 ± 1.0 -10.6 ± 3.4 Coastal California, Southwest, Southern Plains 4 154 10.6 ± 1.1 3.5 ± 3.1 Southeast

Figure 1. Diagram illustrating Holdridge life zones and bioclimatic components along latitudinal and altitudinal gradients (Halasz 2007; original image published at: http://en. wikipedia.org /wiki/ Image:Lifezones_Pengo.svg. This file is licensed under the Creative Commons Attribution ShareAlike license versions 2.5, 2.0, and 1.0).

Figure 2. Cartographic model illustrating the process of developing Holdridge life zone maps from projected climate data based on average values from an ensemble of 16 global climate models (Girvetz et al. 2009). Example shown is for the B1 emissions scenario (IPCC 2007) for years 2055 (T1submodel) and 2085 (T2 submodel). Model created in ArcGIS 10.2.1 ModelBuilder (ESRI 2014). Figure 3. Percent change in mean annual temperature, mean annual precipitation and mean annual potential evapotranspiration ratio from 2006 to 2085 in the coterminous U.S. based on projections from an ensemble of 16 global climate models under a relatively high CO2 emissions scenario (A2).

(a)

(b)

Figure 4. Maps of Holdridge life zones for the coterminous U.S. based on historical and projected climatic means (all zones are not labeled). Source data from the ClimateWizard tool, http://www.climatewizard.org/ (Girvetz et al. 2009). (a) (b) (c)

Figure 5. Holdridge life zones and representative vegetation types for the eastern U.S. based on historical climate conditions (1951-2006) and two future climate projections under different emissions scenarios (low-B1 and high-A2, IPCC 2007). Projections are based on median values derived from an ensemble of 16 GCMs. Black dots are locations of U.S. military installations (location of Ft. Drum, New York highlighted in cyan color). Figure 6. Comparison of Holdridge life zones (color-shaded areas) developed from historical climate data with major forest formations (a) and ecological provinces (b) for the eastern U.S. Modified Braun classification (Braun 1950; Dyer 2006) shown as white outlines in (a) and province-level data from the Hierarchical Framework of Ecological Units in the United States (Cleland et al. 1997) shown in (b).

Figure 7. Non-spatial cluster analysis (k = 5) of percent change in mean annual temperature (CHGMAT) and mean annual precipitation (CHGMAP) projected for 596 military installations in the coterminous U.S. by 2085 under the A2 emissions scenario. Group medians in boxplots (a) and (b) are all significantly different (p < 0.01, ɑ = 0.05, Wilcoxon/Kruskal-Wallis non-parametric multiple comparisons). Map (c) illustrates geographic distribution of installation groups and principal components plot (d) shows distributions in bivariate data space. 64

Figure 8. Statistical summary (a), map of clusters (b) and parallel box plot (c) comparing results of bivariate spatial cluster analysis from the ArcGIS 10.2 Grouping Tool (parameters: number of groups = 8; spatial constraint = K_NEAREST NEIGHBORS; distance method = EUCLIDEAN; number of neighbors k = 8). Colored circles in (a) and (c) are groups means; colored vertical lines in boxplots (a) are group minimum and maximum values. Box plots indicate range, interquartile range and medians for each variable. R2 in (a) = the total sum of squares – explained sum of squares) / total sum of squares.

a b

Figure 9. Holdridge temperature classes for the western USA based on mean annual temperature from 1951 – 2006 (a) and projected mean annual temperature in 2085 under the A2 high emissions scenario (b). Data source: ClimateWizard, www.climatewizard.org (Girvetz et al. 2009). 66

Figure 10. Holdridge life zones derived from historical (1951-2006) climate conditions compared with select boreal species distributions (crosshatched polygons, Little 1971) in the eastern U.S.

Figure 11. Military installations in the coterminous U.S. distributed along mean annual temperature and mean annual precipitation gradients based on historical climate data. Installations are represented by white trianglesU. Projected climate trajectories A, B and C (white arrows) are shown for select installations to illustrate potential biome shifts as discussed in the text. Climate data are derived from the 1981-2010 U.S. Climate Normals dataset (Arguez et al. 2012), available at: http://www.ncdc.noaa.68 gov/data-access/land-based-station-data/land-based- datasets/climate-normals/1981-2010-normals-data. A complete list of installations and acronyms is shown in Appendix A. Chapter 2 Developing Species-Age Cohorts from Forest Inventory and Analysis Data to Parameterize a Forest Landscape Model

Abstract Simulating long term, landscape-level changes in forest composition and spatial structure requires estimates of community or stand age to initialize models of succession and other ecological processes. However, detailed tree or stand age data are rarely available for most forested areas in the eastern United States and even general information on stand history is often lacking. Data from the USDA-Forest Service Forest Inventory and Analysis (FIA) database were used to estimate broad age classes for a forested landscape in northern New York, USA in order to simulate changes in landscape composition and structure that may occur under future climate change scenarios. Relationships between tree diameter and age were developed for FIA site trees using simple linear regression and applied to forest stands at Fort Drum, a 43,000 ha U.S. Army installation located near Watertown, New York. Overall, approximately half of the variation in age was explained by diameter breast height (DBH) across all species studied (r2 ranged from 0.42 for Acer saccharum to 0.63 for Fraxinus americana). Data on site trees that are used to estimate site quality and stand age were extracted from ecoregions similar to the study site, but specific differences in site quality between stands at Fort Drum and FIA plots were not analyzed. Age-diameter relationships from published research on northern hardwood species were evaluated to calibrate results from the FIA- based analysis, but age estimates showed a high degree of variation both within species and across study sites. Nevertheless, this approach provided some objective basis for assigning broad age classes to stands with unknown histories without extensive collection and analysis of tree stem cores. In addition to stand age, species life histories and environmental conditions represented by ecological site types were used to parameterize a stochastic forest landscape model (LANDIS-II) to spatially and temporally model potential changes in forest communities under three climate change scenarios. Forest stands modeled over 100 years without significant disturbance appeared to reflect expected patterns of increasing dominance (50 to 100 percent increase in areal extent) by shade tolerant mesophytic tree species such as sugar maple (Acer saccharum), red maple (Acer rubrum) and Eastern hemlock (Tsuga canadensis) on sites where abundant soil moisture appears to provide a

competitive advantage. However, on drier sandy soils, pine (Pinus strobus, P. resinosa) and oak (Quercus rubra, Q. alba) species continued to be important components of forest stands throughout the modeling period (0 to 30 percent increase respectively, but no net loss at the landscape scale). This suggests that despite abundant precipitation and relatively low evapotranspiration rates for the region in general, relatively low soil water holding capacity and fertility may be limiting factors for the establishment and spread of mesophytic species on excessively drained soils at Fort Drum and elsewhere in the region. Increasing atmospheric temperatures projected for the Northeastern U.S. over the next century could alter temperature and moisture regimes for many coarse textured soils in the region providing a possible mechanism for expansion of xerophytic tree species.

Introduction Over the past twenty years, a number of studies have documented the potential impacts of climate change on forest biomes, broad forest cover types and the ranges of individual species in eastern North America (Canham and Thomas, 2010; Iverson et al. 2008; McKenney et al. 2007; Woodall et al. 2009). Most of these studies were based on correlations between current bioclimatic conditions and current distributions of forest types or tree species, which were then extrapolated into the future using outputs from global or regional climate models to characterize future distributions. These projections either implicitly or explicitly assume that the current distributions of tree species approximate the range of environmental tolerances that they have adapted to over thousands of years and that the absence of a species or community suggests the presence of bioclimatic conditions unsuitable for sustained regeneration and growth. A number of authors have pointed out limitations of the “climatic envelope” approach as the basis for species distribution models (SDM), including reliance on temporally and spatially limited climate data, incomplete information on the distribution of tree species and lack of detailed information on adaptations to wide ranging biophysical conditions (Loehle and LeBlanc 1996; Pearson and Dawson 2003; Thuiller 2004; Wiens et al. 2009). Many species distribution models are based on a few observed or modeled climate variables from the past 100 to 150 years that do not encompass the complete range of climatic conditions under which species have evolved during the Holocene. Often only data considered the most consistent, accurate and available from the past few decades, a period widely recognized as warmer and wetter in comparison to long-term climatic variation in North America, are used to develop correlations with extant tree distributions. Some tree species, especially those 70

with relatively large north-south geographic distributions in eastern North America (e.g., Acer rubrum, A. saccharum, Quercus rubra, Q. alba, Ulmus americana), appear to have the capacity to grow under a wide range of temperature regimes as long as adequate moisture is available (Prasad and Iverson 2003; Tardif et al. 2006; Johnson et al. 2009; Martin-Benito and Pederson 2015). Current climate models are not consistent in forecasting how precipitation may change under a warming climate, but most indicate that average annual precipitation in the northeastern U.S. will continue to increase over the next century (Hayhoe et al. 2008; Huntington et al. 2009; Melillo et al. 2014; Janowiak et al. 2017). In addition to uncertainties about whether future temperature and moisture conditions will exceed the environmental tolerances of tree species, global and regional climate models typically produce grid-based estimates of climatic variables at resolutions of 12 to 20 km, and while this scale may be adequate to support modeling efforts at continental or even regional scales, it may not be effective in describing species response to climate change at scales more appropriate for studying ecosystem function and the influence of disturbances such as timber harvesting, fire or disease. Nevertheless, small-scale studies that correlate extant species distributions with climatic attributes are important in highlighting the range and severity of potential impacts on forest species, communities and ecosystems from anthropogenic climate change, in assessing the variation in ecosystem response that is likely to occur on a regional basis and for pointing out the large uncertainties that exist in current modeling approaches and results. Relatively few studies have attempted to connect results from small-scale climate and vegetation response models to specific landscapes to explore exactly what aspects of the biophysical environment and community dynamics are likely to change under future climate scenarios and how these changes might influence or be influenced by current forest management plans.

A number of different approaches and tools are available to model changes in forest structure and function over time and space including tree and stand-level models (e.g., Climate-Forest Vegetation Simulator, Crookston et al. 2010), species-specific niche (Matthews et al. 2011) and process models (SORTIE, Loehle 2000;) dynamic global vegetation models (MC1, Daley et al. 2000; Morin and Thuiller 2009) and many others. The various trade-offs inherent in these different approaches have been well summarized by several papers citing characteristics such as application scale, spatial versus non-spatial approaches, complexity in constructing parameters, availability of adequate input data, applicability across different ecoregions, statistical approach and interpretation of model 71

outputs (Guisan and Zimmerman 2000; Pearson and Dawson 2003; Thuiller 2004; Gustafson 2013). LANDIS-II (LANdscape DIsturbance and Succession) is a stochastic forest landscape simulator that can incorporate site adaptations, biological interactions (e.g., competition for light, reproductive strategies) and disturbance ecology (effects of fire, wind and timber harvesting) within a spatially- explicit context to model change in forest communities over time (Mladenoff and He 1999; Scheller et al. 2007; de Bruijn et al. 2014). Species-age cohorts and site types that define limitations on establishment and growth are specified for each cell in a raster data structure and life history attributes for each species of interest are used to model change in species composition and biomass over time. Cohorts age and senesce, compete for light and reproduce based on life history attributes such as species longevity, shade tolerance, age when seed is produced and seed dispersal strategies. Unless killed by a disturbance such as timber harvesting or fire, an age-dependent mortality function us used to remove cohorts from the model. Raster maps and associated attribute files are produced that allow visualization and analysis of model results. LANDIS-II has been applied at landscape scales in several regions of North America and has been shown to be a flexible and accurate tool for modeling changes in species composition and distribution over time as a function of climate change while taking into account a variety of natural and anthropogenic disturbances (He et al. 1999; Mladenoff 2004). To provide information in support of long-range forest and wildlife management plans, we proposed to use LANDIS-II to model forest succession at Fort Drum, a U.S. Army installation located in northern New York, USA. Our initial effort focused on developing a preliminary forest succession model to better understand species-site relationships on the installation and evaluate if available information resources were adequate to parameterize simulations and produce credible models of successional patterns over the next 100 years. Disturbances such as fire, timber harvesting or insect and disease events were not included in initial modeling efforts, but will be evaluated in subsequent research.

An important input to initialize and model succession in LANDIS-II is the spatial distribution and structure of species-age cohorts across the landscape. However, the age of individual trees or forest stands is often lacking for many forest landscapes in the eastern U.S. and determining age from increment cores requires a substantial amount of fieldwork and laboratory analysis of tree growth rings. Detailed stand histories were not available for Fort Drum, but a systematic, variable-plot timber inventory completed in 2009-2011 provided basic information on species composition, 72

abundance, diameter and basal area distributions for 1,450 stands covering approximately 25 percent of the installation. Equations to predict tree age from diameter are not abundant in the literature and some forest biologists question whether age can be accurately predicted from tree diameters in mixed species stands with different developmental histories and site characteristics (Gibbs 1963; Cogbill 2003). Differences in shade tolerances and growth rates between tree species in mixed stands as well as the effects of past disturbances can make it very difficult to establish reliable age-diameter relationships. Nevertheless, several studies have shown that a significant portion of the variation in tree age is correlated with diameter for northern hardwood stands in the northeastern United States. Tubbs (1977) analyzed a mature northern hardwood stand dominated by sugar maple in the Upper Peninsula of Michigan after 50 years of selection cutting and found that diameter measured from stumps of recently felled trees explained 88 percent of the variation in tree age. A similar study based on stump diameters for 60 sugar maples trees harvested from a managed forest in Wisconsin indicated that 64 percent of variation in age was explained by diameter (Dey et al. 2017). Leak (1985) developed regression equations based on basal diameters for several northern hardwood and conifer species at two old-growth sites in New Hampshire that explained r2 = 0.47, r2 = 0.79 and r2 = 0.86 of the variation in age based on diameter measured at breast height (DBH) for sugar maple, yellow birch and American beech respectively. Kenefic and Nyland (1999) reported an r2 of 0.81 for age-DBH relationships in a managed, uneven-aged stand comprised of 96 sugar maple trees in central New York State. Collectively, age-diameter relationships reported for sugar maple- dominated stands distributed from New Hampshire to Wisconsin provide reasonable evidence that age can be approximated from stem diameters, albeit with unexplained variation in predicted ages ranging from 12 to as much as 53 percent. Does this hold true for other important tree species with different shade tolerances, rates of growth and site adaptations?

Northern red oak (Quercus rubra) is less shade tolerant, generally faster growing than sugar maple and the two species have largely overlapping geographic ranges (Little 1971; Shifley 1987; Burns and Honkala 1990; Teck and Hilt 1991). As with sugar maple, several studies have developed age- diameter relationships for northern red oak in the eastern U.S. and southern Canada. Rentch (2001) studied five old-growth stands on the Allegheny Plateau, but tree ages derived from increment cores for x trees were not strongly correlated with DBH (r2 = 0.34). Conversely, data derived from an in-depth study of a managed, x year old northern red oak stand in Connecticut showed a very 73

high degree of correlation between age and DBH (r2 = 0.93; Oliver 1975). Analysis of data from northern red oak stands in southeastern New York State (Black Rock Forest 2017) and southern Quebec (Tardif et al. 2006) resulted in age-DBH correlations of r2 = 0.42 and r2 = 0.49 respectively. Correlations could not be derived from data for an old-growth stand in North Carolina (van de Gevel et al. 2012) and managed stands in West Virginia (Miller et al. 1997), but best-fit lines illustrated age-diameter relationships within the range of other studies (Figure x). As with sugar maple, age predictions from published age-diameter relationships for northern red oak are highly variable, but given the substantial differences in site conditions and management histories that no doubt exist for stands in the above locations, it appears that approximately half of the variation in age can be attributable to stem diameter across a large portion of the range of northern red oak in the Northeast. In addition, variability in age predictions might be reduced if predictions are limited to dominant and co-dominant trees and not applied to shade tolerant species in the understory that may be of similar age, but have significantly smaller diameters.

Given the lack of information on tree or stand ages in the Fort Drum forest inventory and the goal of using extant information resources as much as possible, I chose to develop age-diameter relationships from site tree records contained in the New York Forest Inventory and Analysis database (O’Connell, B.M. et al. 2014). The Forest Inventory and Analysis (FIA) program is administered by the USDA Forest Service in cooperation with state agencies. It incorporates a three-phase approach to assessing the status of forest resources throughout the U.S.: phase 1 relies on remote sensing to stratify land cover and determine the areal extent of each stratum; phase 2 consists of field data collection using a systematic sampling framework of fixed-area plots to assess major forest conditions of interest; and phase 3 collects additional forest health attributes on a subset of phase 2 plots. Each standard FIA plot is comprised of four circular subplots covering 0.4 hectares on which attributes are collected or computed for all trees greater than 5 inches in diameter. A standard FIA plot represents approximately 2,429 hectares (6,000 acres) and 15 to 20 percent of each state is assessed annually. Aggregate statewide reports are produced every 5 years that summarize key findings and compare trends over time (O’Connell, B.M. et al. 2014). Statistical details pertaining to the sampling framework, attributes collected at each plot, data processing procedures and accuracy of FIA data can be found in Bechtold and Patteron (2005). Site trees are dominant or co-dominant trees located on FIA subplots that are used to estimate site index and 74

stand age. The age of site trees is determined by counting growth rings on increment cores extracted at 1.37 m above the ground and each tree is assigned a weighting factor that approximates the proportion of overstory trees represented by each site tree (Woodall et al. 2010). If FIA site tees are representative of the diameters and ages of dominant and co-dominant species that comprise the majority of the overstory in their respective stands, then age-diameter relationships derived from these trees should approximate the mean age of stands at Fort Drum when applied to the dominant species in each stand as defined by relative basal area. It is important to note that the goal in developing age-diameter relationships was not to determine the precise ages of individual trees with a high degree of accuracy, but to develop generalized age-diameter relationships that would support assignment of broad age classes to stands at Fort Drum in order to parameterize the LANDIS-II base succession model.

Methods Study site Fort Drum is a U.S. Army installation covering over 43,000 ha near Watertown, NY, USA, approximately 25 km east of Lake Ontario (latitude 44.10o N, longitude 75.65o W; Figure 1). The installation lies primarily in the Saint Lawrence Glacial Lake Plain physiographic region (Bailey et al. 1994) with elevations ranging from 126 m along West Creek to 280 m in the northeast portion of the base near Lake Bonaparte. The region has a humid, cool temperate climate with an average annual temperature of 7.9o C and mean annual precipitation of 1100 mm. Average monthly precipitation is highest in the late fall (114 mm in November) and lowest in mid-winter (70 mm in February), but is distributed evenly throughout the growing season. A substantial amount of precipitation falls as snow during the winter months averaging 2847 mm per year over the past 35 years. The average frost-free growing season runs from May 15 through September 25 (133 days), but frost can occur as late as June 2 and as early as September 9 (Arguez et al. 2012, 1981-2010 U.S. Climate Normals dataset, Watertown, NY weather station, accessed 10/10/2016 at https://www.ncdc.noaa.gov/cdo-web/). Due to the greater distance from Lake Ontario and slightly higher elevations, the northeastern portion of the base may experience cooler conditions, earlier frosts and slightly shorter growing seasons (Northeast Regional Climate Center, interpolated regional climate maps accessed 10/10/2016 at http://www.nrcc.cornell.edu/regional/ climatenorms/climatenorms.html). 75

Three physiographic units characterize the majority of the installation: relatively flat, low elevation plains derived from fine textured glacial lacustrine deposits (28 percent of total area), slightly more elevated and coarse textured sand terraces and plains derived from glacial outwash and wind-blown deposits (25 percent of total area), and bedrock controlled uplands covered by coarse glacial till in the northeastern third of the installation (27 percent of total area) (McDowell 1989). Topography is generally level to gently rolling and both alluvial and depressional wetlands are common throughout the installation. Soils in the western and southern portions of the base formed from post-glacial, fine textured lacustrine deposits and sandy outwash plains underlain by sandstone and limestone bedrock. Mesic to wet, circumneutral Hapludualfs and Endoaqualfs (e.g., Collamer, Hudson, Niagara and Rhinebeck series) are common on lowland plains while excessively drained Udipsamments (Plainfield and Windsor series) predominate on higher sand plains and terraces. Intermediate to the above are relatively narrow terraces of mesic, loamy fine sands (e.g., Wareham and Deerfield series). There are significant areas of calcareous glacial till covering lower elevation landscapes that have given rise to soils with relatively high base saturation and pH (Eutrudepts in Amenia, Benson and Nellis series). The northeastern upland portion of the installation is dominated by acidic, coarse textured soils derived from glacial till underlain by bedrock comprised of gneiss, schist and granite. Most upland soils are mesic, relatively shallow to bedrock and have a frigid temperature regime (mean annual temperature < 8o C and difference between summer and winter temperature means > 6o C; Soil Survey Staff 1999). Predominant soil groups in the uplands are Lithic Haplorthods (e.g., Lyman-Abram complex) and Typic Dystrudepts (e.g., Insula-Millsite- Quetico complex). Organic soils (Haplosaprists such as Carlisle, Bucksport and Pondicherry series) associated with bogs and alluvial wetlands cover approximately 6 percent of the installation and are especially common in the northeastern uplands as a result of glacial erosion of underlying bedrock during the latter stages of the Pleistocene (McDowell 1989).

Prior to European settlement in the 18th century, most of northern New York State outside of the Adirondack Mountains was covered by mixed pine (Pinus strobus, P. resinosa), hemlock (Tsuga canadensis) and northern hardwood (Fagus grandifolia, Betula alleghaniensis, Acer saccharum) forests typical of cool temperate regions in eastern North America (Braun 1950; Cogbill, Burke and Motzkin 2002). After the Revolutionary War, large numbers of settlers migrated from New England to western New York State and began clearing land to establish farms and produce wood products

for fuel, construction materials and commercial trade (Cronon 1983). Over the next 100 to 150 years, much of the land in the St. Lawrence River Valley, including the Watertown - Fort Drum area, was converted to agriculture with forests remaining in relatively isolated patches on soils either too wet or too rocky to farm. The U.S. Army began acquiring land in the early 1900’s and expanded the installation to over 35,000 ha during World War II. Fort Drum reach its current extent in the 1980’s as new facilities were developed to house and train the U.S. Army’s 10th Mountain Division (U.S. Army, Fort Drum web site, accessed 10/27/16 at http://www.drum.army.mil/AboutFortDrum/Pages/History_lv2.aspx). Concurrent with the growth of Fort Drum, the number of farms and agricultural acreage in the region in general declined substantially during the latter half of 20th century (50 to 90 percent in Jefferson and Lewis counties respectively; Stanton and Bills 1996). These lands have largely reverted to old field, shrub and early successional forest communities that characterize much of the landscape at Fort Drum and in surrounding areas.

Approximately 30 percent of the undeveloped lowland landscapes at Fort Drum remain in open old field-woody shrub cover types that developed following abandonment or displacement of agricultural activities over the past 50 to 100 years. Grass (Schizachyrium scoparium, Avenella flexuosa) - sedge (Carex pensylvanica, C. rugosperma and C. lucorum) communities with scattered clumps of pine (Eastern white pine-Pinus strobus, red pine-P. resinosa) and oak (northern red oak- Quercus rubra, white oak-Q. alba) dominate dry sand plains formed from glacial outwash and post- glacial windblown deposits (successional northern sandplain grassland, Ecological Communities of New York State, Edinger et al. 2014). In addition to occurring on the most xeric sites at Fort Drum, these communities are maintained in an open, early successional state through mowing and occasional ground fires that occur because of military training exercises. On more mesic old field sites with finer textured soils, early successional grass-sedge-herb and woody shrub communities are common with species composition varying depending on time since abandonment, type of disturbance and soil drainage (successional old field and shrub communities, Edinger et al. 2014). Common species include bluegrasses (Poa pratensis, P. compressa), several introduced grass species (e.g., Elymus repens, Bromus inermis, Dactylis glomerata), sedges (Carex spp.) and a diverse herb layer comprised of goldenrods (Solidago altissima, S. nemoralis, S. rugosa), New England aster (Sympyotrichum novae-angliae), evening primrose (Oenothera biennis), Queen-

Anne's-lace (Daucus carota), ragweed (Ambrosia artemisiifolia) and common chickweed (Cerastium arvense). Woody shrub cover ranges from 10 to 50 percent and includes species such as speckled alder (Alnus incana), shrub dogwoods (Cornus amomum, C. racemosa), sumacs (Rhus glabra, R. typhina), raspberries (Rubus spp.) and saplings of red maple (Acer rubrum), green ash (Fraxinus pennsylvanica) and willow species (Salix spp.). Other relatively open cover types at Fort Drum include recently harvested tracts dominated by early successional tree species such as trembling and big-tooth aspen (Populus tremuloides, P. grandidentata), gray birch (Betula populifolia), black cherry (Prunus serotina) and red maple (successional northern hardwoods, Edinger et al. 2014).

Broadleaved deciduous forests cover approximately 33 percent of Fort Drum and are comprised of two primary northern hardwood types: beech-maple forest and maple-basswood (Tilia americana) rich mesic forest (Braun 1950; Edinger et al. 2014). Beech-maple forests occur on upland sites with relatively shallow, coarse textured and frigid soils derived from acidic glacial till. Common associates include yellow birch, white ash, American hophornbeam (Ostrya virginiana) and red maple. Depending on the level of disturbance, understories are comprised of seedlings from the above species along with striped maple (Acer pennsylvanicum), American hornbeam (Carpinus caroliniana), viburnums (Viburnum lantanoides, V. acerifolium) and Eastern hemlock. Extensive diameter-limit timber harvests have occurred in the upland forests at Fort Drum over the past 10 to 20 years, which has apparently resulted in a much higher component of red maple, black cherry, white ash and northern red oak than might be expected for typical northern hardwoods at this latitude. In addition, beech bark disease (fungal pathogens Neonectria faginata and N. ditissima) has likely reduced the prevalence of American beech at Fort Drum as it has in other areas of the northeastern U.S. over the past 100 years (Morin et al. 2007). Maple-basswood forests occur at lower elevations and on finer textured, mesic soils derived from glacial lacustrine deposits and till derived from limestone bedrock. Dominant tree species include sugar and red maple, American basswood and white ash, but American elm (Ulmus americana), bitternut hickory (Carya cordiformis) and black cherry are also common associates. Understory vegetation is generally more diverse than upland forests due to base-rich soils, abundant soil moisture and warmer temperatures, and includes woody species such as American hophornbeam, alternate-leaved dogwood (Cornus alternifolia), mountain maple (Acer spicatum) and witch hazel (Hamamelis virginiana), ferns

(Athyrium filix-femina, Dryopteris marginalis) and a rich herb layer (Edinger et al. 2014). Red maple, black (Fraxinus nigra) and green ash and cottonwood (Populus deltoides) may increase in frequency on poorly drained floodplains, stream terraces and depressions within the maple- basswood type (red maple-hardwood swamp forest and floodplain forest types described in Edinger et al. 2014). Dutch elm disease (Ophiostoma spp.) has killed most mature American elm trees at Fort Drum, but elm regeneration remains abundant on mesic, fine textured soils throughout the installation.

Evergreen coniferous forests and evergreen-deciduous mixtures occur on approximately 25 percent of Fort Drum landscapes. Eastern white pine and Eastern hemlock form mixed stands with northern hardwoods on both upland and lowland sites (pine-northern hardwood and hemlock-northern hardwood forests, Edinger et al. 2014). Hemlock is significantly more common on lowland soils derived from coarse loamy glacio-fluvial deposits and in forested wetlands where it forms dense stands with northern white cedar (Thuja occidentalis), red maple and yellow birch (hemlock- hardwood swamp forest, Edinger et al. 2014). Pine plantations (Eastern white, red pine, Jack pine Pinus banksiana and Scotch pine Pinus sylvestris) and open stands of northern red oak predominate on excessively drained sand plains north and east of the airfield. A number of oak and oak-pine stands are mowed periodically to maintain an open understory for military training requirements.

Five tree species (Eastern white pine, red maple, sugar maple, Easter hemlock and black cherry) make up over 70 percent of the total basal area and are among the most abundant species on the installation. In conjunction with the above species, trembling aspen, gray birch, American elm, white ash and northern red oak comprise the top ten species in terms of relative abundance (84 percent of all species) and relative frequency at Fort Drum (Table 1).

Developing species-age cohorts Age and DBH for site trees were extracted from FIA plots for 13 species characteristic of major forest types at Fort Drum. Insufficient data were available for a few overstory tree species at Fort Drum (American beech, white oak Quercus alba) and several species are not generally considered as canopy dominants (e.g., American hophornbeam, black ash), so these species were excluded from the analysis. All plots had a forested condition class and elevation less than 500 m; the latter excluded trees located in higher elevations of the Adirondack Mountains and Tug Hill Plateau that 79

were less likely to reflect site conditions at Fort Drum. Plots were initially limited to those located in Saint Lawrence, Jefferson or Lewis counties to approximate site conditions on Fort Drum as much as possible. However, in order to obtain minimal samples sizes to support development of regression equations, data from additional counties were required for three species (black cherry, northern red oak and Eastern hemlock) and a statewide sample was required for two species (American basswood and American elm). All site tree records were filtered to remove duplicate records from multiple inventory years. Tree age (years) and DBH (mm) were analyzed using simple linear regression (SAS JMP 13.1.0, SAS Institute 2016); diameter distributions for all species or species groups were not significantly different from normal based on a Shapiro-Wilk W test (W < 0.05) and all outliers were retained in the analysis.

Age-diameter regression equations were applied to forest stands on Fort Drum by using the mean DBH for the most dominant species in each stand as the explanatory variable. The diameter distribution for all trees was negatively skewed (Figure 2c) and over 80 percent of the stands had a mean DBH less than 218.4 mm (8.6 in). I therefore assumed that mean DBH might provide a more meaningful basis for estimating stand ages than median DBH by reducing the influence of large numbers of smaller diameter stems. Smaller trees are less likely to be members of dominant and co- dominant crown classes and would therefore be less comparable with site trees on FIA plots. Beginning with the species with the highest relative basal area (RBA) in a stand, I used mean DBH and a species-specific regression equation to calculate age and rounded values to the next highest 20-year class (e.g., ages 1-19 assigned to the 20-year age class, ages 21-39 assigned to the 40-year age class, etc.). The number of species used to estimate age and forest type for each stand was guided by the following general rules:

1) Stands where SPP1RBA > 70%; mean DBH for SPP1 used to calculate age classes and

SPP1RBA used to determine forest type;

2) Stands where SPP1RBA > 50% and < 70% and SPP2RBA > 20%; mean DBH for both species used to calculate age classes and both species used to define forest type; 3) Mean DBH for SPP1, SPP2 and SPP3 used to determine age classes and forest type for all other stands;

4) Where SPP1 was not considered a characteristic overstory species (e.g., American hophornbeam), the mean diameter for the second most dominant species was used as a starting point for age calculations.

The above stand type classification rules are similar to those used on National Forest lands in the northeastern U.S. as outlined in the NED-2 User’s Guide (Twery et al. 2011). For most stands, one or two species comprised the majority of the basal area and were used to define forest type. When present, certain tree species with lower relative basal area were used to help define characteristic forest types. For example, red maple and Eastern hemlock were dominant species in more than one forest type and often comprised large proportions of total basal area, but based on the presence of species such as black ash and northern white cedar that are indicative of poorly drained sites, stands were typed as “Wet Forest” instead of “Hemlock” or “Mixed hemlock-hardwood” types. Forest type classifications followed community descriptions by Erdinger et al. (2014). Using all possible combinations of age classes (7) and species types (9) would have created a very large number of categories to process within LANDIS-II and would not likely improve succession models in any biologically meaningful manner. Therefore, to improve processing time, simplify preparation of parameter files and maximize interpretability of results from initial simulations, stands were coded into two broad age classes (stand age < 40 = “young” stands; stand age > 40 = “mature” stands) for each forest type where these age classes occurred.

Vegetation types for areas on the installation not covered by the forest inventory were derived from a land cover dataset developed from 1-meter resolution aerial photography in 2006 (Chris Dubony 2011, personal communication) and county soil surveys. These cover types were primarily abandoned agricultural fields comprised of early successional grass-forb and woody shrub species, and open grass fields maintained for military training activities. Species were not defined for grass- forb cover types and subtypes were classified as mesic or xeric based on soil texture and drainage. Shrub cover types were combined into a single class comprised of woody deciduous shrub species (Alnus incana, Salix spp.) and early successional hardwoods (Ulmus americana, Betula populifolia, Acer rubrum and Populus tremuloides) with mesic and xeric subtypes defined by soil texture and drainage. Open fields were assigned age class 20 and shrub types contained age classes 20 and 40. Once the stand attribute table was fully coded by type and age classes, stand features (polygons)

were converted to a raster format with 30-meter cell size using an integer value to represent each age-type condition.

Model parameterization The Landis-II Age-Only Succession module incorporates life history information on tree species (e.g., longevity, reproductive traits, fire and shade tolerance), spatial distribution of initial species- age cohorts and establishment probabilities for each species by site type to model species regeneration, growth, colonization and mortality for designated time periods (Mladenoff and He 1999). Information from Silvics of North America (Burns and Honkala 1990) and previous research utilizing LANDIS-II (Thompson et al. 2011; Duveneck et al. 2014) was used to define species attributes (Appendix A, Table A-1). Establishment probabilities for 25 woody species and 2 herbaceous cover types (Appendix A, Table A-2) were assigned based on known adaptations to site conditions (Burns and Honkala 1990), community descriptions (Ferree and Anderson 2013; Erdinger et al. 2014) and their association with ecological site types derived from soil properties and physiographic variables at Fort Drum (Odom 2018). Input files were generated using a simple text editor and executed to simulate community succession at Fort Drum over 100 years using a 20- year time step.

Results Species composition and relative basal areas were similar for FIA site tree plots and Fort Drum forest inventory plots; however, the installation contained a significantly larger amount of white pine and oak species and lower basal area in species associated with poorly drained sites such as northern white cedar and black ash (Figure 2a). This may be an artifact of under sampling of forested wetlands in the Fort Drum inventory, a larger relative proportion of excessively drained, sandy soils on the installation that favor pine and oak species, or a combination of both. Relative basal area for the most common northern hardwood species were approximately equivalent. Diameter distributions of trees on FIA versus Fort Drum plots were also similar with both datasets exhibiting an inverse J-shaped distribution, although there was a larger proportion of smaller diameter trees (DBH < 300 m) on FIA plots (0.71) than on the installation (0.48) (Figure 2, b and c).

Two hundred and twenty one unique community type-age cohorts (see Appendix A, Table A-1 for a complete listing of cohorts) were identified at a stand level based on the relative basal area of dominant tree species and age classes derived from age-diameter equations. Linear models of age- diameter relationships were all statistically significant (p < 0.05) with r2 values ranging from 0.42 for sugar maple to 0.63 for white ash (Table 2). Predicted ages for individual trees ranged from less than 20 (American elm and poplar species) to over 110 years (white ash and Eastern hemlock) with mean ages for all species ranging from 33.8 (SE ± 2.5) to 69.6 (SE ± 4.5) years. Age distributions for the most common species were consistent with a priori assumptions that forests at Fort Drum are relatively young (Figure 3). Age-diameter curves for sugar maple (Figure 4) and northern red oak (Figure 5) compared reasonably well with previous research in terms of growth rates (slope of best fit lines), although stands represented by FIA site trees appeared to be younger than those described in the literature (except for sugar maple stands described by Kenefic and Nyland 1999 on the Allegheny Plateau).

Changes in forest types and individual species over time largely followed expected successional trends (Figure 6). Early successional species such as quaking aspen and black cherry increased substantially (13 and 8 percent respectively) over the first 40-60 years and then declined as more shade tolerant species (sugar and red maple) increased in the landscape. The impacts of Dutch elm disease were not simulated in initial succession models and therefore American elm appeared to increase substantially in importance by colonizing many of the grass-forb and shrub communities on mesic sites with fine textured and relatively base-rich soils. The young maple-elm forest type, which also included significant components of aspen, gray birch and black cherry, increased more than any other community type on the installation (> 400 percent) and continued its expansion throughout the 100-year simulation. Much of the increase in the aforementioned types occurred in post-agricultural, old field communities that declined substantially in the first 20 to 40 years and were almost entirely absent by year 60 of the simulation. Oak species and Eastern hemlock did not increase substantially (approximately +3 percent), but were able to maintain their relative abundance in the landscape despite increasing competition from sugar and red maple on all but the most xeric sites. However, in lieu of disturbance, oak woodland, oak-maple and oak-pine forest types all declined substantially as composition shifted to more shade tolerant maple species. Forested wetlands remained relatively constant over time in terms areal extent. Species changes in 83

this type that covered a very small proportion of the landscape were not evaluated in detail, but would likely experience some loss of shade intolerant species such as yellow birch and black ash while maintaining an overstory comprised of Eastern hemlock, northern white cedar and red maple. Despite its current dominant position in the landscape (> 30 percent relative basal area), the relative abundance of white pine remained constant over the duration of the simulation at a landscape level, but pine dominance in mixed stands declined at a stand level as more shade tolerant hardwoods increased in younger age classes.

Discussion Simulation of forest succession over 100 years on a landscape in northern New York State produced results largely consistent with our general understanding of how tree species composition changes over time in this ecoregion (Bormann and Likens 1979; Copenheaver 2008; Thompson et al. 2013; Duveneck et al. 2014). In the absence of major disturbances, shade tolerant mesophytic broadleaved species such as sugar maple, red maple and American beech increased in importance in established forest stands. Sugar and red maple are abundant and widespread at Fort Drum and appear to be able to rapidly colonize all but the most xeric sites. Sugar maple increased 200 percent in terms of its relative frequency in the landscape and maintained a constant rate of increase of 2 percent per year throughout the simulation period. Red maple increase at a similar rate until year 60 and then showed a modest decline of 3 percent over the last 40 years of the simulation, presumably due to increased competition from more shade tolerant species. American beech is a relatively minor component in the landscape as a whole (relative abundance and relative basal area < 2 percent), which may reflect an inability to compete on the relatively fertile sites that comprise much of the installation, impacts of beech bark disease that is prevalent in the region (Twery and Patterson 1984) or simply inadequate time for populations to recover from past land use disturbance (Thompson et al. 2013).

Large areas on Fort Drum remain in early successional grass-forb and shrub communities resulting from conversion of agricultural lands 50 to 60 years ago. Early successional tree species such as gray birch, quaking aspen and American elm appear to be slowly colonizing these areas and increased substantially over the first 60 years of the simulation. Birch and aspen declined slightly during latter stages of the simulation while elm continued to increase at a rapid rate. The latter may be attributable to elm being slightly more shade tolerant than other early successional species and

more competitive on the fertile and moderately poorly drained soils that characterize much of the former agricultural lands at Fort Drum. It is unclear why black cherry, which is also an important component of early successional stands and the second most common species in terms of relative frequency and abundance on the installation, declined precipitously after year 40 of the simulation (from 10 percent to less than 4 percent of the total landscape). Other early successional species have more abundant, wind-dispersed seeds and substantially larger effective seed dispersal distances that may account for the inability of black cherry to maintain a dominant position in the landscape. Old field community types (24 percent of the current landscape) were almost completely absent by the end of the simulation as they were overtaken by woody shrub and early successional tree species.

The proportion of the installation covered by mature stands of important conifer species such as Eastern hemlock and white pine remained fairly constant throughout the simulation, although mixed pine-hardwood stands tended to shift towards hardwood dominance as shade tolerant deciduous species, primarily sugar and red maple, increased in abundance. Young pine and mixed pine- hardwood types decreased by 83 and 74 percent respectively during the 100-year simulation. Oak woodlands that occur almost exclusively on excessively drained, coarse textured soils declined almost 99 percent. Establishment probabilities for maple species were almost half of those for oak and pine species on these sites, but high relative abundance, greater regeneration potential and shade tolerance apparently combined to overwhelm any potential adaptions to relatively low soil fertility and moisture. It may be that despite low stand densities (mean basal area < 17.5 m2/ha) and concomitant high light levels that would appear to favor oak reproduction, the cold wet climate at Fort Drum may limit acorn production and increase seedling mortality in oak species relative to the more prolific and cold-adapted maple and aspen species (Hannah 1987; Burns and Honkala 1990; Johnson, Shifley and Rogers 2009). Oak dominance is currently maintained in these stands by mechanical mowing and limited wildfires resulting from military training exercises, and it seems clear that some form of disturbance will be required in the future if northern red and white oaks are to remain even minor components of the landscape at Fort Drum. Loss of oak woodlands could be a concern for management of a growing wild turkey population on the installation.

The accuracy of age classes derived from FIA site trees was impossible to determine without undertaking a significant effort to collect and analyze tree increment cores. However, age-diameter

curves developed for two species, sugar maple and northern red oak, were consistent with relationships derived from published data in terms of growth rates and stand ages (Figures 4 and 5). Most of the stands in previous studies were described as old growth or mature and age-diameter curves developed from FIA site trees and applied to stands at Fort Drum consistently predicted younger age classes in comparison to stands analyzed in these studies. The one exception was for a sugar maple stand studied by Kenefic and Nyland (1999) in south, central New York State that had a similar diameter distribution, but was apparently younger and growing faster in comparison to sugar maple stands at Fort Drum. This stand had a much higher percentage of sugar maple in comparison to most stands at Fort Drum and was selectively harvested in 1973 and 1993 with the specific intent of modifying the diameter distribution, including the removal of poor quality and non-commercial stems. Trees in this study were approximately half the age of trees at Fort Drum for the same diameter. Timber harvests have also occurred over the past 10 years in stands with a sugar maple component at Fort Drum, but were likely not carried out with the express intention of creating balanced diameter distributions and increasing radial growth of remaining sugar maple trees. Therefore, it is not surprising that the Kenfic and Nyland stand contained stems with diameters twice those of Fort Drum trees at the same age.

Age-diameter relationships from previous studies were based on single, mature stands with known cutting histories. Mature sugar maple-dominated stands do occur at Fort Drum and in the surrounding landscape, but most of the installation is characterized by relatively younger stands with smaller mean DBH than the stands in the aforementioned studies. In addition, site conditions in any single stand should be less variable than the range of conditions in hundreds of northern hardwood stands at Fort Drum, which may mean that growth rates (and therefore age-diameter relationships) may be applicable to some stands, but not others. Overall, age-diameter curves for sugar maple and northern red oak showed a high degree of variability across the studies reviewed with diameter explaining as little as 39 percent to as much as 94 percent of the variation in age. Age-diameter relationships based on published studies may be more representative of older northern hardwood stands on upland sites at Fort Drum, but may not reflect growth rates for lower elevation species on mesic (maple-basswood type) and xeric (pine and pine-oak types) sites at Fort Drum.

Although the accuracy of age classes developed for forest stands at Fort Drum was not quantified, ages represented by 20-year classes should account for some level of variation in age-diameter relationships and have been used elsewhere to model successional trajectories at a landscape scale. Zhang et al. (2009) used FIA site trees to develop age-diameter equations and species age cohorts for oak (Quercus spp.)-hickory (Carya spp.) forests in the Missouri Ozarks. Details of their regression analyses were not presented, but r2 for age-diameter correlations were fairly low, ranging from 0.15 (northern hardwoods) to 0.35 for white oak. Duveneck et al. (2004) calculated tree ages using FIA plot data and site index curves to develop 5-year age classes to parameterize a LANDIS- II model for a landscape in Michigan. Other studies have referenced FIA data as the source of species age cohorts used to parameterize LANDIS-II models, but often do not clearly described how age data were derived, do not provide an accuracy assessment of age cohort estimates and do not include sensitivity analyses that might help to understand how variability in age cohorts might affect modeling results (Shifley et al. 2006; Rittenhouse et al. 2007; Yang et al. 2011). In addition, it is not clear from published FIA field guides if site trees are the only trees used to establish stand age or if additional stems are cored and analyzed. If a few trees on FIA plots are used to assign stand ages within the FIA database, estimates of stand age would be subject to a high degree of variation since each FIA plot represents approximately 6,000 acres of forest, which could contain hundreds of stands with varying histories, species compositions and site conditions. No assessments of the accuracy of tree or stand ages in the FIA Eastwide database have been published and age data are generally considered some of the least accurate in the entire FIA program (Northeast FIA staff, personal communication).

Conclusions While admittedly imperfect, developing broad age cohorts from species-specific age-diameter equations derived from FIA site trees is a relatively straightforward process and provides a means of minimizing site variability to some degree by selecting FIA plots and site trees from similar ecoregions as those being modeled. However, unknown stand histories and variability in growth rates, both within a species at different life stages and among species with different shade tolerances in mixed northern hardwood stands, creates a substantial level of uncertainty concerning the accuracy of FIA-based age cohorts used in LANDIS-II simulations. Over long simulation timeframes (> 100 years), inaccuracies in cohort ages may become somewhat unimportant since, in 87

lieu of major disturbance, shade tolerant species should eventually dominate most northern hardwood stands regardless of whether stand simulations begin at age 20 or 40. However, as exogenous disturbances are introduced to models to simulate more realistic future landscape conditions, especially over shorter time periods, the accuracy of cohort ages would appear to be a significant issue since response to various disturbances can be highly age-dependent (cite….). It would seem prudent therefore to invest some resources in an assessment of tree or stand ages for landscapes being modeled and to include some level of sensitivity analysis to more accurately assess how inaccuracies in cohort ages might influence model results.

References Arguez, A., I. Durre, S. Applequist, R. S. Vose, M. F. Squires, X. Yin, R. R. Heim, Jr., and T. W. Owen, 2012: NOAA's 1981-2010 U.S. Climate Normals: An Overview. Bulletin of the American Meteorological Society, 93, 1687-1697.

Bailey, R. G.; Avers, P. E.; King, T.; McNab, W. H., eds. 1994. Ecoregions and subregions of the United States (map). Washington, DC: USDA Forest Service. 1:7,500,000, with supplementary table of map unit descriptions, compiled and edited by W. H. McNab and R. G. Bailey.

Bishop, D. A., C. M. Beier, N. Pederson, G. B. Lawrence, J. C. Stella, and T. J. Sullivan. 2015. Regional growth decline of sugar maple (Acer saccharum) and its potential causes. Ecosphere 6(10): 179, 14 p.

Black Rock Forest. 2017. Long term plot data from The Calvin Whitney Stillman Research Archive, accessed 12/8/2018 at http://blackrockforest.org/environmental-data/forest-legacy- data/long-term-plot-data.

Bormann, F. H., and G. E. Likens. 1979. Pattern and Process of a Forested System. Springer- Verlag, New York. 253 pp.

Braun, E. L. 1950. Deciduous Forests of Eastern North America. Philadelphia: Blakiston. 596 p.

Burns, R. M. and B. H. Honkala, technical coordinators. 1990. Silvics of North America: 1. Conifers; 2. Hardwoods. Agriculture Handbook 654. U.S.Department of Agriculture, Forest Service, Washington, DC.

Canham, C. D. and R. Q. Thomas. 2010. Frequency, not relative abundance, of temperate tree species varies along climate gradients in eastern North America. Ecology 91(12): 3433-3440.

Cogbill, C. V. 2003. Commentary on “Age / Diameter Relationships”. Online discussion forum posted 1/1/2003, http://www.nativetreesociety.org/oldgrowth/age_determinations.htm.

Cogbill, C., J. Burk, and G. Motzkin. 2002. The forests of presettlement New England, USA: spatial and compositional patterns based on town proprietor surveys. J. Biogeogr. 29:1279–1304. 89

Copenheaver, C. A. 2008. Old-field succession in western New York: the progression of forbs and woody species from abandonment to mature forest. Rhodora 110(942):157-170.

Cronon, W. 1983. Changes in the land: Indians, colonists and the ecology of New England. Hill and Wang, New York.

Crookston N., G. Rehfeldt, G. Dixon and A. Weiskittel. 2010. Addressing climate change in the forest vegetation simulator to assess impacts on landscape forest dynamics. For Ecol Mgmt 260: 1198–1211.

Daly, C., D. Bachelet, J. Lenihan, R. Neilson, W. Parton, and D. Ojima. 2000. Dynamic simulation of tree-grass interactions for global change studies. Ecological Applications. 10(2): 449-469. de Bruijn, A., E. J. Gustafson, B. R. Sturtevant, J. R. Foster, B. R. Miranda, N. I. Lichti and D. F. Jacobs. 2014. Toward more robust projections of forest landscape dynamics under novel environmental conditions: Embedding PnET within LANDIS-II. Ecological Modelling Volume 287, 10 September 2014, Pages 44–57.

Dey, D. C., J. Dwyer, and J. Wiedenbeck. 2017. Relationship between tree value, diameter, and age in high-quality sugar maple (Acer saccharum) on the Menominee Reservation, Wisconsin. J. Foresry 115(5): 397-405.

Duveneck, M. J., R. M. Scheller, M. A. White, S. D. Handler, and C. Ravenscroft. 2014. Climate change effects on northern Great Lake (USA) forests: A case for preserving diversity. Ecosphere 5(2): 23.

Edinger, G. J., D. J. Evans, S. Gebauer, T. G. Howard, D. M. Hunt, and A. M. Olivero (editors). 2014. Ecological Communities of New York State. Second Edition. A revised and expanded edition of Carol Reschke’s Ecological Communities of New York State. New York Natural Heritage Program, New York State Department of Environmental Conservation, Albany, NY. 173 p.

Ferree, C and M. G. Anderson. 2013. A Map of Terrestrial Habitats of the Northeastern United States: Methods and Approach. The Nature Conservancy, Eastern Conservation Science, Eastern Regional Office. Boston, MA. Accessed 2/19/2014 from: http://easterndivision .s3.amazonaws.com/Terrestrial/CanadaHabitat/HABITAT_neUS_eCANADA.zip.

Gates, F. C. and G. E. Nichols. 1930. Relation between age and diameter in trees of the primeval northern hardwood forest. J. Forestry 28(3): 395-398.

Gibbs, C. B. 1963. Tree diameter a poor indicator o f age in West Virginia hardwoods. Research Note NE-11. Upper Darby, PA: U.S. Department of Agriculture, Forest Service, Northeastern Forest Experiment Station. 4 p.

Guisan, A. and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology, Ecological Modelling 135( 2–3): 147-186.

Gustafson, E.J. 2013. When relationships estimated in the past cannot be used to predict the future: using mechanistic models to predict landscape ecological dynamics in a changing world. Landscape Ecol (2013) 28: 1429.

Hannah, Peter R. 1987. Regeneration methods for oaks. Northern Journal of Applied Forestry. 4: 97-101

Hayhoe et al. 2008. Regional climate change projections for the Northeast USA. Mitig Adapt Strat Glob Change (2008) 13:425–436.

He, H.S., D.J. Mladenoff, and T.R. Crow. 1999. Linking an ecosystem model and a landscape model to study forest response to climate warming. Ecological Modelling 114(1999): 213-233.

Huntington, T. G., A. D. Richardson, K. J. McGuire, and K. Hayhoe. 2009. Climate and hydrological changes in the northeastern United States: recent trends and implications for forested and aquatic ecosystems. In: NE Forests 2100: A Synthesis of Climate Change Impacts on Forests of the Northeastern U.S. and Eastern Canada. Canadian Journal of Forest Research 39:199-212.

Iverson, L. R., E. D. Martin, C. T. Scott, and A. Prasad. 1997. A GIS-derived integrated moisture index to predict forest composition and productivity in Ohio forests. Landscape Ecology 12: 331-348.

Iverson, L. R., A. Prasad, and S. Matthews. 2008. Modeling potential climate change impacts on the trees of the northeastern United States. Mitig Adapt Strat Glob Change 13: 487-516.

Janowiak, M. K. et al. 2018. New England and northern New York forest ecosystem vulnerability assessment and synthesis: a report from the New England Climate Change Response Framework project. Gen. Tech. Rep. NRS-173. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 234 p.

Johnson, P.S., S.R. Shifley, and R. Rogers. 2009. The ecology and silviculture of oaks, 2nd ed. Wallingford, UK: CABI Publishing, CAB International. 580 p.

Kenefic, L. and R. Nyland. 1999. Sugar maple height-diameter and age-diameter relationships in an uneven-aged northern hardwood stand. Northern Journal of Applied Forestry 16(1): 43-47.

Leak, W. B. 1985. Relationships of tree age to diameter in old-growth northern hardwood and spruce-fir. USDA For. Serv. Res. Note NE-329, 4 p.

Little, E.L., Jr., 1971. Atlas of United States trees, volume 1, conifers and important hardwoods: Misc. Pub. 1146. Washington, D.C.: U.S. Department of Agriculture. 9 p., 200 maps.

Loehle, C. 2000. Forest ecotone response to climate change: sensitivity to temperature response functional forms. Can. J. For. Res. 30:1632-1645.

Loehle, C. and D. Leblanc. 1996. Model-based assessments of climate change effects on forests: A critical review. Ecological Modelling 90: 1-31.

Loewenstein, E.F., P. S. Johnson, and H. E. Garrett. 2000. Age and diameter structure of a managed uneven-aged oak forest. Canadian Journal of Forest Research 30: 1060-1070.

Marquis, D.A. 1991. Independent effects and interactions of stand diameter, tree diameter, crown class, and age on tree growth in mixed-species, even-aged hardwood stands. In: McCormick, Larry H.; Gottschalk, Kurt W., eds. Proceedings, 8th Central Hardwood Forest Conference; 92

1991 March 4-6; University Park, PA. Gen. Tech. Rep. NE-148. Radnor, PA: U.S. Department of Agriculture, Forest Service, Northeastern Forest Experiment Station: 442-458.

Martin-Benito, D. and N. Pederson. 2015. Convergence in drought stress, but a divergence of climatic drivers across a latitudinal gradient in a temperate broadleaf forest. J. Biogeogr. 42: 925–937.

Matthews, S. N., L. R. Iverson, A. M. Prasad, M. P. Peters, and P. G. Rodewald . 2011. Modifying climate change habitat models using tree species-specific assessments of model uncertainty and life history-factorsForest Ecology and Management 262 (2011): 1460–1472.

McDowell, L. 1989. Soil Survey of Jefferson County, New York. Washington, D.C.: USDA Soil Conservation Service. 210p.

McKenney, D. W., J. H. Pedlar, K. Lawrence, K. Campbell, and M. F. Hutchinson. 2007. Potential Impacts of Climate Change on the Distribution of North American Trees. BioScience 57:939- 948.

Melillo, J.M., T. C. Richmond, and G. W. Yohe. 2014. Climate change impacts in the United States: the third national climate assessment. U.S. Global Change Research Program. U.S. Government Printing Office, Washington, DC.

Miller, Gary W. 1997. Effect of crown growing space and age on the growth of northern red oak. In: Spiecker, H.; Rogers, R.; Somogyi, Z., comps. IUFRO Proceedings: advances in research in intermediate oak stands; 1997 July 27-30; Freiburg, Germany. Freiburg, Germany: University of Freiburg: 140-159.

Mladenoff, D.J. and H.S. He. 1999. Design, behavior and applications of LANDIS, an object- oriented models of forest landscape disturbance and succession. Pages 125-162 in Spatial modeling of forest landscape change, Mladenoff, D.J. and Baker, W.L. (eds.), Cambridge University Press, Cambridge, UK.

Mladenoff D.J. 2004. LANDIS and forest landscape models. Ecol Mod 180: 7–19.

Morin, X., and W. Thuiller. 2009. Comparing niche- and process-based models to reduce prediction uncertainty in species range shifts under climate change. Ecology 90:1301-1313.

O’Connell, B.M. et al. 2014. The Forest Inventory and Analysis database: Database description and user guide version 6.0.1 for Phase 2. U.S. Department of Agriculture, Forest Service. 748 p.

Odom, R.H. 2018. Simulated effects of climate change on soil moisture deficits, species distributions and biomass in a northern hardwood forest (doctoral dissertation). Unpublished manuscript, Virginia Polytechnic Institute and State University, Blacksburg, VA.

Oliver, C. D. 1975. The development of northern red oak (Quercus rubra L.) in mixed species, even-aged stands in central New England. Ph.D. Dissertation, Yale University, 1975. Xerox University Microfilms Ann Arbor, Michigan. 223 p.

Pearson, R. G. and T. P. Dawson. 2003. Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography 12: 361–371.

Pederson, N. et al. 2013. Is an epic pluvial masking the water insecurity of the greater New York City region? Journal of Climate, 26, 1339–1354.

Prasad, A. M. and L. R. Iverson. 2003. Little's range and FIA importance value database for 135 eastern U.S. tree species. http://www.fs.fed.us/ne/delaware/4153/global/littlefia/index.html, Northeastern Research Station, USDA Forest Service, Delaware, Ohio.

Rentch, J.S., 2001. Stand dynamics and disturbance history of five oak-dominated old-growth stands in the unglaciated Appalachian Plateau. Ph.D. Thesis. West Virginia University, Morgantown, WV, 227 p.

Rittenhouse, C. D., W. D. Dijak, F. R. Thompson, and J. J. Millspaugh. 2007. Development of landscape-level habitat suitability models for ten wildlife species in the central hardwoods region. Gen. Tech. Rep. NRS-4. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 47 p.

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 spatial and temporal resolution. Ecol Mod 201: 409–419.

Shifley; Stephen R. 1987. A generalized system of models forecasting Central States tree growth. Res. Pap. NG279.St. Paul, MN: U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station. 10 p.

Shifley, S. R., F. R. Thompson, W. D. Dijak, M. A. Larson, and J. J. Millspaugh. 2006. Simulated effects of forest management alternatives on landscape structure and habitat suitability in the Midwestern United States. Forest Ecology and Management 229 (2006) 361–377.

Soil Survey Staff. 1999. Soil taxonomy: A basic system of soil classification for making and interpreting soil surveys. 2nd edition. Natural Resources Conservation Service. U.S. Department of Agriculture Handbook 436.

Stanton B. F. and N. L. Bills. 1996. The return of agricultural lands to forest: Changing land use in the twentieth century. Extension Bulletin 96-03. Ithaca, NY: Cornell University Dept. Agric. Resource and Managerial Econ. 140 p.

Stover, M. E. and P. L. Marks. 1998. Successional vegetation on abandoned cultivated and pastured land in Tompkins County, New York. Journal of the Torrey Botanical Society 125: 150-164.

Tardif, J., F. Conciatori, P. Nantel, and D. Gagnon. 2006. Radial growth and climate responses of white oak (Quercus alba) and northern red oak (Quercus rubra) at the northern distribution limit of white oak in Quebec, Canada. Journal of Biogeography 33(9): 1657-1669.

Teck, Richard M.; Hilt, Donald E. 1991. Individual tree-diameter growth model for the Northeastern United States. Res. Pap. NE-649. Radnor, PA: U.S.. Department of Agriculture, Forest Service, Northeastern Forest Experiment Station. 11 p.

Thompson, J.R., D.R. Foster, R. Scheller, and D. Kittredge. 2011. The influence of land use and climate change on forest biomass and composition in Massachusetts, USA. Ecological Applications 21(7): 2425-2444.

Thompson J.R., D. N. Carpenter, C. V. Cogbill, and D. R. Foster. 2013. Four centuries of change in Northeastern United States forests. PLoS ONE 8(9): e72540. https://doi- org.ezproxy.lib.vt.edu/10.1371/journal.pone.0072540

Thuiller, W. 2004. Patterns and uncertainties of species' range shifts under climate change. Global Change Biology 10: 2020–2027.

Tubbs, C. H. Age and structure of a northern hardwood selection forest, 1929-1976. Journal of Forestry 75(1): 1977.

Twery, M. J., P. D. Knopp, S. A. Thomasma, and D. E. Nute. 2011. NED-2 User's Guide. Gen. Tech. Rep. NRS-85. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 193 p. https://doi.org/10.2737/NRS-GTR-85.

Twery, M.J. and W. A. Patterson, W.A. 1984. Variations in beech bark disease and its effects on species composition and structure of northern hardwood stands in central New England. Canadian Journal of Forest Research 14: 565–574.

U.S. Forest Service. 2017. Forest inventory and analysis national core field guide [NorthernResearch Station edition] volume I supplement: field data collection procedures for Phase 2+ plots. Ver. 7.2. Washington, DC: U.S. Department of Agriculture, Forest Service. Available at http://www.nrs.fs.fed.us/fia/data-collection/ (accessed January 29, 2018).

Van de Gevel, S., J. Hart, M. Spond, P. White, M. Sutton, and H. Grissino-Mayer. 2012. American chestnut to northern red oak: a forest dynamics study in the Blue Ridge Mountains, North Carolina, U.S.A. Botany 90: 1263-1275.

Walker, K. V., M. B. Davis, and S. Sugita. 2002. Climate change and shifts in potential tree species range limits in the Great Lakes region. Journal of Great Lakes Research 28(4): 555-567.

Wang, W. J., H. S. He, F. R. Thompson, J. S. Fraser, and W. D. Dijak. 2017. Changes in forest biomass and tree species distribution under climate change in the northeastern United States. Landscape Ecology 32:1399-1413.

Woodall, С., B. Conkling, M. Amacher, et al. 2010. The Forest Inventory and Analysis database. Version 4.0: Database description and user’s manual for Phase 3. Gen. Tech. Rep. NRS-61. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 180 p.

Woodall, C., C. Oswalt, J. Westfall, C. Perry, M. Nelson, and A. Finley. 2009. An indicator of tree migration in forests of the eastern United States. Forest Ecology and Management. 257: 1434- 1444.

Yang, J., H. S. He, S. R. Shifley, F. R. Thompson, and Y. Zhang. 2011. An innovative computer design for modeling forest landscape change in very large spatial extents with fine resolutions. Ecological Modelling 222 (15): 2623-2630.

Zhang, Y., H.S. He, W.H. Dijak, J. Yang, S.R. Shifley, and B.J. Palik. 2009. Integration of satellite imagery and forest inventory in mapping dominant and associated species at a regional scale. Environmental Management 44, 312–323.

Appendix A. Species life histories and establishment probabilities used to parameterize the LANDIS-II Age- Only Successional Model.

Table A-1. Life history traits for species and cover types included in LANDIS-II Age-Only Succession model. Effective Max Probability Min Max Sexual Shade Fire Fire Species/cover type Longevity seeding seeding of sprout sprout maturity tolerance tolerance strategy distance distance sprouting age age (yr) (yr) (1-5) (1-5) (m) (m) (0-1) (yr) (yr) Mesic field 40 1 1 2 100 1000 0.5 0 40 none Xeric field 40 1 1 4 200 1000 0.5 0 40 none Abies balsamea 200 25 5 1 30 160 0 0 0 none Acer rubrum 150 10 4 1 100 200 0.75 0 100 none Acer saccharum 300 40 5 1 100 200 0.1 10 60 none Alnus indica 40 10 1 1 100 500 0.9 0 40 none Betula alleghaniensis 300 40 3 2 100 400 0.1 10 180 none Betula populifolia 100 30 2 2 200 5000 0.5 10 70 none Carya cordiformis 300 30 3 3 30 100 0.75 10 100 none Fagus grandifolia 300 40 5 1 30 100 0.75 0 100 none Fraxinus americana 150 20 3 1 100 200 0.75 10 100 none Fraxinus nigra 150 20 3 1 100 200 0.75 10 100 none Ostrya virginiana 100 25 5 1 100 500 0.7 0 40 none Pinus banksiana 100 15 1 3 30 100 0 0 0 serotiny Pinus resinosa 200 35 2 4 30 275 0 0 0 none Pinus strobus 400 40 3 3 60 210 0 0 0 none Populus deltoides 100 20 1 1 1000 5000 0.9 10 100 none Populus tremuloides 100 20 1 1 1000 5000 0.9 10 100 none Populus grandidentata 100 20 1 1 1000 5000 0.9 10 100 none Prunus serotina 150 20 1 1 30 100 0.75 0 100 none Quercus alba 300 25 3 2 30 1000 0.75 20 100 none Quercus rubra 250 25 3 2 30 1000 0.75 20 100 none Salix spp. 150 20 1 1 200 5000 0.75 10 70 none Thuja occidentalis 400 20 3 1 30 60 0.1 10 100 none Tilia americana 250 30 4 1 30 120 0.1 10 200 none Tsuga canadensis 450 30 5 2 30 100 0 0 0 none Ulmus americana 200 40 3 1 100 400 0.75 0 100 none 99

Table A-2. Establishment probabilities for species included in LANDIS-II Age-Only Succession model listed by ecological site type.

1 2 3 4 5 6 7 Hydric Subhydric Mesic glacio- Mesic glacio- Mesic, acidic Xeric sand Mesic, basic Species depressional alluvial lacustrine fluvial sand glacial till plain glacial till wetland deposit plain terrace Mesic field 0.9 0.9 0.7 0.3 0.2 0.1 0.3 Xeric field 0.1 0.4 0.3 0.3 0.3 0.9 0.5 Abies balsamea 0.9 0.7 0.2 0.1 0.1 0 0.1 Acer rubrum 0.7 0.8 0.9 0.9 0.9 0.5 0.7 Acer saccharum 0.2 0.3 0.9 0.5 0.9 0.5 0.9 Alnus incana 0.9 0.9 0.5 0.1 0.1 0.1 0.1 Betula alleghaniensis 0.9 0.7 0.5 0.7 0.9 0.2 0.5 Betula populifolia 0.3 0.7 0.9 0.3 0.5 0.5 0.5 Carya cordiformis 0.1 0.7 0.6 0.2 0.2 0.1 0.9 Fagus grandifolia 0.3 0.5 0.5 0.6 0.9 0.3 0.4 Fraxinus americana 0.3 0.7 0.9 0.3 0.9 0.3 0.7 Fraxinus nigra 0.9 0.9 0.5 0.2 0.1 0.1 0.1 Ostrya virginiana 0.2 0.5 0.5 0.4 0.8 0.1 0.4 Pinus banksiana 0.1 0.1 0.2 0.3 0.3 0.9 0.2 Pinus resinosa 0.1 0.1 0.2 0.3 0.5 0.9 0.3 Pinus strobus 0.4 0.7 0.5 0.7 0.9 0.9 0.5 Populus deltoides 0.3 0.9 0.5 0.2 0.1 0.1 0.4 Populus tremuloides 0.3 0.9 0.9 0.5 0.5 0.5 0.5 Populus grandidentata 0.3 0.7 0.8 0.5 0.7 0.5 0.5 Prunus serotina 0.3 0.9 0.9 0.5 0.9 0.5 0.7 Quercus alba 0.1 0.2 0.4 0.4 0.3 0.8 0.5 Quercus rubra 0.1 0.2 0.5 0.5 0.5 0.9 0.5 Salix spp. 0.9 0.9 0.5 0.1 0.1 0.1 0.1 Thuja occidentalis 0.9 0.5 0.3 0.2 0.1 0.1 0.1 Tilia americana 0.3 0.5 0.9 0.4 0.5 0.2 0.9 Tsuga canadensis 0.9 0.9 0.9 0.9 0.7 0.5 0.4 Ulmus americana 0.3 0.7 0.9 0.2 0.4 0.2 0.7

100

Table 1. Relative frequency, density and abundance of the 22 most common tree species at Ft. Drum, NY, USA.

Number of plots Species Relative frequency Relative basal area Relative abundance where present Acer rubrum 3734 55.8 13.7 20.0 Prunus serotina 2983 44.6 8.9 11.0 Pinus strobus 2385 35.6 31.4 16.7 Populus tremuloides 1702 25.4 5.0 7.2 Acer saccharum 1624 24.3 9.2 8.2 Ulmus americana 1168 17.4 1.4 3.3 Fraxinus americana 1080 16.1 1.9 3.0 Betula populifolia 1057 15.8 0.6 3.7 Tsuga canadensis 1018 15.2 9.2 8.2 Quercus rubra 754 11.3 5.2 3.2 Betula alleghaniensis 691 10.3 1.2 1.9 Fagus grandifolia 583 8.7 1.2 1.5 Tilia americana 505 7.5 1.6 2.0 Populus grandidentata 433 6.5 1.7 1.9 Ostrya virginiana 345 5.2 0.1 0.9 Quercus alba 332 5.0 1.0 0.8 Amelanchier laevis 279 4.2 0.1 0.7 Fraxinus nigra 273 4.1 0.3 1.0 Pinus resinosa 229 3.4 1.7 1.2 Betula papyrifera 194 2.9 0.2 0.5 Thuja occidentalis 163 2.4 0.6 1.2 Carya cordiformis 117 1.7 0.3 0.4

101

Table 2. Age-diameter relationships developed from site trees (n = 395) extracted from the New York FIA database.

Dbh (mm) Age (yrs) Linear Regression Model Species n (SE) (SE) Range r2 RMSE P Equation

Fraxinus americana 35 249.7 (12.2) 47.7 (3.9) 21-118 0.63 14.1 <0.001 A = -14.77259 + 0.250115 D Tilia americana** 35 299.5 (13.6) 50.5 (3.3) 24-102 0.56 13.1 <0.001 A = -4.309491 + 0.1830496 D Prunus serotina* 15 273.6 (19.9) 46.9 (5.1) 23-77 0.54 13.8 0.002 A = -4.579108 + 0.1880035 D Ulmus americana** 62 243.0 (10.2) 33.8 (2.5) 17-120 0.52 13.6 <0.001 A = -9.319295 + 0.1774207 D Acer rubrum 81 255.0 (6.9) 52.0 (1.8) 20-87 0.50 11.7 <0.001 A = 5.0184657 + 0.1840787 D Betula alleghaniensis 10 230.6 (29.4) 57.6 (6.0) 26-88 0.48 14.4 0.025 A = 24.976281 + 0.1414536 D Quercus rubra* 13 282.3 (21.0) 52.2 (4.4) 27-77 0.48 11.9 0.009 A = 11.665456 + 0.1434076 D Populus tremuloides, P. 17 256.5 (21.1) 35.1 (3.1) 19-67 0.45 9.8 0.003 A = 9.7495765 + 0.0986561 D grandidentata Tsuga canadensis* 22 313.1 (17.8) 69.6 (4.5) 33-115 0.44 16.3 0.001 A = 16.623835 + 0.1693081 D Pinus strobus, P.resinosa 44 292.9 (9.6) 42.2 (2.2) 19-74 0.42 11.2 <0.001 A = -0.836716 + 0.1473024 D Acer saccharum 61 288.4 (8.2) 61.3 (2.1) 27-95 0.42 12.6 <0.001 A = 11.430045 + 0.1734929 D

* Additional data from FIA plots < 500m elevation located in Clinton, Franklin, Oneida and Oswego counties included to increase sample size.

** Additional data from FIA plots < 500m elevation in all New York counties included to increase sample size

102 Figure 1. Study site location (inset) and location of military infrastructure (training areas shown in green outline) at Fort Drum, New York, USA. Training areas are comprised of over 1,500 forest stands managed to support military training requirements, timber and fiber production, game and non-game wildlife management programs and ecosystem sustainability. The “Impact Zone” and developed areas were excluded from the study.

103 Fort Drum FIA relative relative Tree species basal area basal area (percent) (percent) Mean 188.9 Std Dev 103.0 Pinus strobus 31.4 15.4 Std Err Mean 0.89 Range 1137.9 Acer rubrum 13.7 13.6 Interquartile Range 104.1 n 13327 Tsuga canadensis 9.2 9.6

Acer saccharum 9.2 7.8 Prunus serotina 8.9 5.3

Quercus rubra 5.2 1.9 Populus tremuloides 5.0 3.7

(b) FIA plots (n = 395) Fraxinus americana 1.9 4.3 Populus grandidentata 1.7 1.3 Pinus resinosa 1.7 2.0

Tilia americana 1.6 2.0 Ulmus americana 1.4 4.4 Mean 292.6 Std Dev 173.0 Fagus grandifolia 1.2 2.3 Std Err Mean 0.62 Range 1196.3 Betula alleghaniensis 1.2 1.6 Interquartile Range 226.1 n 78946 Quercus alba 1.0 0.2 Thuja occidentalis 0.6 6.4 Betula populifolia 0.6 1.1

Salix spp. 0.4 0.9 Fraxinus nigra 0.3 1.8 Carya cordiformis 0.3 1.0 (c) Fort Drum plots (n = 6,695)

(a)

Figure 2. Relative basal area (a) for the twenty most frequent trees at Fort Drum and for the same species on site tree plots extracted from the New York FIA database. Overall, species composition and relative density are similar, but significant differences are highlighted for several species (bold type and gray shading). Diameter distributions (Dbh) are shown for all trees for FIA site tree plots (b) and the Fort Drum forest inventory plots (c).

104

Figure 3. Age distributions for the 12 most common tree species at Ft. Drum, New York, USA based on age-diameter equations developed from FIA site trees.

105

Figure 4. Age-diameter relationships for sugar maple derived from previous studies and analysis of FIA site trees.

MI – Managed stand with several selection harvests (Tubbs 1977, Table 1). WI – Managed stands, selectively harvested, Wisconsin (Dey et al. 2017, Figure 1, *diameter measured from top of cut stump). NH – Old-growth stand, New Hampshire (Leak 1985, Table 1, *diameter measured at top of root swell). FIA – Site trees from New York FIA database (this study). NY – Uneven-aged managed stand, Allegheny Plateau, NY (Kenefic and Nyland 1999, Figure 2).

106

Figure 5. Age-diameter relationships for northern red oak derived from previous studies and analysis of FIA site trees.

QU – Based on mean radial growth averages for 12 stands in southern Quebec, Canada (Tardif et al. 2006, Table 2). NC – Old-growth stand, southern Appalachian Mountains, NC (van de Gevel et al. 2012, Figure 3). NY - Black Rock Forest, NY (unpublished data, Black Rock Forest Consortium, The Calvin Whitney Stillman Research Archive, downloaded and compiled 12/8/2018 from http://blackrockforest.org/environmental-data/forest-legacy-data/long-term-plot-data). OH – Old-growth stands (5), Allegheny Plateau, WV-OH-PA (Rentch 2001, Figure 4.7). WV – Average for 16, 55 and 80 year old managed stands, Monongahela National Forest, WV (Miller 1997, Table 5). CT – Average diameter and age for 7 managed stands (Oliver 1975, Table 1). FIA – Site trees from New York FIA database (this study).

107

Figure 6. Landscape level change in the 12 most common overstory tree species at Fort Drum, New York, USA over a 100-year simulation of forest succession under a no disturbance scenario.

108

a) 2015

b) 2115

Figure 7. Current (a) and simulated (b) community type-age cohorts at Ft. Drum, New York, USA. Maple-dominated stands in maple-elm (bright green) and northern hardwood (dark brown) types increased substantially over the 100-year simulation while open grass-forb communities, oak woodlands (orange) and mixed pine (dark olive) stands declined. Forest stand data were not available for “INACTIVE” (gray) areas. In the map legend, “YNG” denotes stands less than 40 years of age and “HWD” denotes northern hardwood species.

109 Chapter 3 Simulated Effects of Climate Change on Soil Moisture Deficits, Species Distributions and Biomass in a Northern Hardwood Forest

Abstract Communities that occur in the ecotones between major forest biomes in eastern North America are expected to undergo significant changes under rapid and unprecedented climatic change in the next century. However, past land use histories, successional status and the broad environmental tolerances of many tree species make it difficult to accurately predict which species will be most affected. To better understand species-site relationships as they relate to the potential impacts of rapid climate change in the northeastern U.S., tree species distributions and environmental gradients were studied at Fort Drum, a 43000-ha U.S. Army installation located at the intersection of temperate and boreal forests in northern New York State, USA. Forests at Fort Drum are comprised primarily of cool temperate tree species typical of northern hardwood forests (e.g., Fagus grandifolia, Betula alleghaniensis, Acer saccharum, Tsuga canadensis), but also contain boreal species at the southern end of their ranges (e.g., Populus tremuloides, P. grandidentata, Thuja occidentalis, Abies balsamea, Picea rubens) and elements of warm temperate oak-hickory forests whose ranges extend far to the south (e.g., Quercus rubra, Q. alba, Carya cordiformis, Ulmus Americana, A. rubrum). Previous research has suggested that warm temperate forest species may increase in abundance in the northeastern U.S. if increasing temperatures result in higher mortality or decreased regeneration in more cold tolerant species. Site characteristics at Fort Drum related to soil moisture, soil temperature, soil fertility and microclimate were analyzed using k-means cluster and non-parametric multivariate correlation analysis to identify biophysical conditions thought to be important to regeneration and growth that may be altered under a warmer climate. Mean importance values for the 23 most frequently occurring tree species were mapped in a geographic information system (GIS) and related to ecological site types using indicator species analysis to quantify plant-site relationships. Change in soil water deficits relative to current conditions was calculated for mid- and late-century dates under low (Representative Concentration Pathway 4.5) and high (Representative Concentration Pathway 8.5) emissions climate scenarios using a water budget model and climate parameters derived from the Community Climate System Model (CCSM 3.0). Projected changes in water deficits were used to modify species establishment probabilities for ecological site types and 110

model change in abundance and biomass using the LANDIS-II forest landscape simulation model. Percent soil sand content, soil drainage, soil pH and elevation explained much of the variability in a regression tree model (n = 5,416; r2 = 0.86, RMSE = 0.22) incorporating seven ecological site types as the response variable. Water deficits increased up to 300 percent for excessively drained sites dominated by sandy or shallow, rocky soils under warmer (+2.0 oC to +4.7 oC mean annual temperature) and wetter (+19 mm to +74 mm mean annual precipitation) future climates, but as expected, deeper soils with finer textures and larger capacities to store water showed moderate to no effects from projected increases in evapotranspiration. Simulated declines in species establishment probabilities on dryer sites led to decreased importance and biomass in mesophytic species (Populus spp., Acer spp.) and a relative increase in importance and biomass of more drought-adapted species (Quercus rubra, Pinus strobus). However, overall results suggest that soil moisture availability will remain adequate to support successful regeneration of most tree species at Fort Drum under warmer and slightly wetter climate conditions projected over the next several decades.

Introduction Over the next century, rapid warming of the Earth’s atmosphere due to accumulation of carbon dioxide (CO2) and other greenhouse gases from anthropogenic sources is expected to create a plethora of challenges for forest communities and individual species in North America and throughout the world. The most drastic changes in atmospheric temperatures are projected to occur at high latitudes and so increased attention has focused on the boreal and cool temperate forests that cover large areas of North America and northern Eurasia (Amber et al. 2007; Beckage et al. 2008; Shuman and Shugart 2009; Canham and Thomas 2010; Boisvert-Marsh et al. 2014). High altitude montane forests that are also comprised of cold adapted species are also expected to be adversely affected (Lee et al. 2005; McCullough et al. 2016; Wason and Dovciak 2017), and in general, boreal forest species are predicted to be displaced by more southerly and lower altitude species as increasing evapotranspiration, drought, forest fires and insect and disease outbreaks lead to higher mortality and decreased regeneration in these forests. In the northeastern U.S., mean annual temperature is expected to rise by 1.4o C to 4.4o C by the end of the century (Hayhoe et al. 2008), potentially leading to significant and widespread impacts on forest ecosystems in the region (Rustad et al. 2012; Janowiak et al. 2018). Although there

111

appears to be general agreement that atmospheric temperatures are rising and will continue to do so without reduction in fossil fuel use, numerous studies and simulations conducted in eastern North America have yielded somewhat mixed results when projecting the effects of climate change on tree species. Well-known species distribution models (SDM) developed by Iverson et al. (2008, 2011), McKenney et al. (2007, 2011) and others based on correlations between historical bioclimatic patterns and extant species distributions have identified “winners and losers” in terms of successful adaptation to climate stress by certain tree species, at least in terms of range expansion or contraction at a macro-scale. These studies suggest that environmental conditions that support widespread boreal and cool temperate species such as Populus tremuloides, Acer saccharum, Fagus grandifolia and Abies balsamea may contract by 70 to 90 percent in the northeastern U.S. by the end of the century, although the authors recognize that substantial lag times may occur between changes in climate and actual contraction in species distributions (Iverson and McKenzie 2013). Concomitantly, SDM’s predict that southern species (e.g., Quercus spp., Carya spp., southern Pinus spp., Liriodendron tulipifera) will expand into the Northeast and colonize sites left vacated by declining boreal species. Tang et al. (2012) used a dynamic ecosystem model (LPJ-GUESS) to assess the impacts of climate change on New England forests and reported that 60 percent of the region would be dominated by Quercus spp.- Carya spp. forests by 2099 with boreal species relegated to northern Maine and isolated mountain tops. The evidence of rapid northward and upslope shifts in temperate-boreal forest ecotones may cause novel species assemblages as warm temperate species colonize sites currently dominated by cool temperate forest types (Beckage et al. 2008). This work, and others (Janowiak et al. 2018) describe a dire future for northeastern forests as currently assembled and suggest similar impacts on wildlife populations, hydrologic processes and resource-based economies throughout the region.

In contrast, Foster and D’Amato (2015) showed that some boreal species (e.g., Picea rubens) are increasing in abundance at lower elevations in the northern Appalachian Mountains, apparently due to recovery from extensive logging in the early part of the 20th century and the effects of acidic deposition in the 1970’s and 1980’s. Wason and Dovciak (2017) also found conflicting results when studying demography of tree saplings versus adults in the northern Appalachians, with some species expanding their range upslope (Fagus grandifolia), some downslope (P.

112 rubens, A. saccharum) and no regional trend in either direction for Abies balsamea. These authors nonetheless concluded that climate plays a dominant role in determining regional distributions across all species studied, but noted that past land use and soil characteristics are also are important determinants for individual species and sites. Nowacki and Abrams (2015) observed that land use legacies are a major factor in determining current and future tree species distributions in the eastern U.S. Widespread forest cutting and burning in the 18th and 19th centuries increased the abundance of oak (Quercus spp.), pine (Pinus spp.) and other disturbance-adapted species throughout much of the region. Reforestation and fire suppression over the past ca. 100 years has led to increased abundance of more mesophytic, shade tolerant hardwood species (primarily Acer spp.) at the same time that atmospheric warming has intensified, and contrary to predictions of increased dominance of oak-pine communities due to CO2-induced global warming. These studies and others (Vadeboncoeur et al. 2011; Hanberry 2013; Thompson et al. 2013; Goring et al. 2016) suggest that forest communities in the East may not as yet be responding to a warming climate because equilibrium with long term temperature and moisture patterns has not been reached due to the effects of past disturbances. Furthermore, it is likely that current successional trajectories, in lieu of major disturbances, will continue to exert a primary influence on species distributions for several decades unless increasing atmospheric temperatures result in widespread mortality or reduced regeneration of tree species in the region (Loehle 2000; Thompson et al. 2013; Duveneck et al. 2017).

A growing body of research increasingly suggests that soil water deficits resulting from increased evapotranspiration rates and/or decreased precipitation are a primary area of concern for North American forests under a rapidly warming climate (Hanson and Weltzin 2000; Ollinger et al. 2008; Allen et al. 2010; Dymond et al. 2014; Clark et al. 2016; Vose et al. 2016). Decreases in available soil water can result in decreased photosynthetic efficiency, decreased nutrient uptake and hydraulic stress, often resulting in lower growth rates and increased susceptibility to secondary pathogenic complexes (McDowell et al. 2008). Many tree species appear to be able to recover from short-duration or low intensity water shortages, but long-term growth declines (Payette et al. 1996; Brzostek et al. 2014; Martin-Benito and Pederson 2015; Nolet and Kneeshaw 2018) and widespread mortality (McDowell et al. 2008; Klos et al. 2009; Worrall et al. 2013) have been documented in the eastern U.S. and Canada from multi-year and

113 intense droughts, even in areas that exhibit relatively high rainfall and mesophytic conditions such as the southern Appalachians (Clinton et al. 1993). However, precipitation forecasts from a number of global climate models indicate that the northeastern U.S. may receive 5 to 10 percent more precipitation over the next several decades (Hayhoe et al. 2008; Huntingdon et al. 2009), which is consistent with the recent trend of decreased frequency and intensity of drought in the region. Gustafson and Sturtevant (2013) used mortality data from the USDA-Forest Service Forest Inventory and Analysis database (FIA; O’Connell et al. 2014) and the Palmer Drought Severity Index (Palmer 1965) to develop relationships between loss of forest biomass on FIA plots and drought in Wisconsin, USA. They concluded that drought-induced mortality could be detected in FIA data, that it varied with species drought tolerance and site type, and that projected changes in temperature and precipitation could be used to predict the effects of future drought on forest biomass and species abundance. However, when a similar approach was applied in the northeastern U.S., model performance was poor and little of the variation in biomass was explained by drought. The cause of poor model performance was attributed to the relatively low incidence and severity of drought in the northeast during the time period studied (1969 – 2007) in comparison to the upper Midwest (Gustafson 2014).

Given the level of uncertainty about how regional climate change will affect specific landscapes, ecological processes, and biological communities, resource managers are faced with developing management plans that incorporate a range of potential future landscape states (Clark et al. 2001; Becknell et al. 2015; Swanston et al. 2018). To assist resource managers with responsibility for forest and wildlife management on U.S. military installations, a multi-scale approach was developed to assess potential impacts from rapid climate change in the coterminous U.S. A coarse-scale analysis of over 600 installations identified locations that were most vulnerable to climate change and highlighted potential candidates within major ecoregions for more detailed analysis (Odom 2018a). A methodology was then developed to analyze specific impacts to forest ecosystems at a landscape scale using best available data and a stochastic forest landscape simulator (LANDIS-II, LANdscape DIsturbance and Succession; Scheller et al. 2007). LANDIS- II incorporates information on regeneration strategies and biological adaptions of tree species (e.g., fire tolerance, seed dispersal, nutrient requirements), disturbance effects and competitive interactions within a site-specific context to model forest succession and biomass over time. The

114

LANDIS-II suite of extensions has been applied at landscape scales in several regions of North America and has proven quite robust in modeling changes in species composition and distribution as a function of climate change while taking into account a variety of natural and anthropogenic disturbances (Mladenoff 2004). Initial landscape-level work focused on developing species-age cohorts and a preliminary forest site classification to create a baseline forest succession model (no disturbance events or site modifications from climate change) to better understand how species life history traits, site conditions and interspecific competition interacted to drive future changes in species distribution and biomass over 100 years at Ft. Drum, New York, USA (Odom 2018b). This work describes more fully the development of ecological site classifications and species-site relationships as the basis for simulating future landscape conditions, and an approach for modeling changes in soil water availability that could limit regeneration and growth of tree species under future climate change scenarios. Because mortality data to support temporal and spatial modeling of drought-biomass relationships at Fort Drum were not available at the scale required, a water balance model was used to project changes in available soil water and modify species establishment probabilities over time within the LANDIS-II Biomass Succession extension.

Methods Study Area Fort Drum is a U.S. Army installation covering over 43,000 ha near Watertown, New York, USA, approximately 25 km east of Lake Ontario (44.10o N, 75.65o W; Figure 1). The installation lies primarily in the Saint Lawrence Glacial Lake Plain physiographic region (Cleland et al. 2007) with elevations ranging from 126 m along West Creek to 280 m in the northeast portion of the base near Lake Bonaparte. The region has a humid, cool temperate climate with mean annual temperature of 8.0 degrees C and mean annual precipitation of 1029 mm over the past century. Mean monthly precipitation is highest in the late fall (114 mm in November) and lowest in mid- winter (70 mm in February), but is distributed evenly throughout the growing season. A substantial amount of precipitation falls as snow during the winter months averaging 2847 mm per year over the past 35 years. The average frost-free growing season runs from May 15 through September 25 (133 days), but frost can occur as late as June 2 and as early as September 9 (Arguez et al. 2012). Due to the greater distance from Lake Ontario and slightly higher

115 elevations, the northeastern portion of the base may experience cooler conditions, earlier frosts and slightly shorter growing seasons (Northeast Regional Climate Center, interpolated regional climate maps accessed 10/10/2016 at http://www.nrcc.cornell.edu/regional/climatenorms/ climatenorms.html). Overall, climate has changed relatively little over the past century, although there is a slight trend toward warmer and wetter weather and an absence of significant periods of drought over the past 30 to 40 years in comparison to the first half of the 20th century (Figure 2; Hayhoe et al. 2008; Huntington et al. 2009).

Three physiographic units characterize the majority of the installation: relatively flat, low elevation plains derived from fine textured glacial lacustrine deposits (28 percent of total area); slightly more elevated, coarse textured sand terraces and plains derived from glacial outwash (25 percent of total area); and bedrock controlled uplands covered by coarse glacial till in the northeastern third of the installation (27 percent of total area) (McDowell 1989). Topography is generally level to gently rolling and both alluvial and depressional wetlands are common throughout the installation. Soils in the western and southern portions of the base formed from post-glacial, fine textured lacustrine deposits and sandy outwash plains underlain by sandstone and limestone bedrock. Mesic to wet, circumneutral Hapludualfs and Endoaqualfs (e.g., Collamer, Hudson, Niagara and Rhinebeck series) are common on lowland plains while excessively drained Udipsamments (Plainfield and Windsor series) predominate on higher sand plains and terraces. Intermediate to the aforementioned are relatively narrow terraces of mesic, loamy fine sands (e.g., Wareham and Deerfield series). There are significant areas of calcareous glacial till covering lower elevation landscapes that have given rise to soils with relatively high base saturation and pH (Eutrudepts in the Amenia and Nellis series). The northeastern upland portion of the installation is dominated by acidic, coarse-loamy textured soils derived from glacial till underlain by bedrock comprised of gneiss, schist and granite. Most upland soils are mesic, relatively shallow to bedrock and have a frigid temperature regime (mean annual temperature < 8o C and difference between summer and winter temperature means > 6o C; Soil Survey Staff 1999). Predominant soil groups in the uplands are Lithic Haplorthods (e.g., Lyman- Abram complex) and Typic Dystrudepts (e.g., Insula-Millsite-Quetico complex). Organic soils (Haplosaprists such as Carlisle, Bucksport and Pondicherry series) associated with bogs and alluvial wetlands cover approximately 6 percent of the installation and are especially common in

116 the northeastern uplands as a result of glacial erosion of underlying bedrock during the latter stages of the Pleistocene (McDowell 1989).

Prior to European settlement in the 18th century, much of northern New York State outside of the Adirondack Mountains was covered by mixed conifer (Pinus strobus, P. resinosa, Tsuga canadensis) and northern hardwood (Fagus grandifolia, Betula alleghaniensis, Acer saccharum) forests typical of cool temperate regions in eastern North America (Braun 1950; Cogbill, Burke and Motzkin 2002). After the Revolutionary War, large numbers of settlers migrated from New England to western New York and began clearing land to establish farms and produce wood products for fuel, construction materials and commercial trade (Cronon 1983). Over the next 100 to 150 years, much of the land in the St. Lawrence River Valley, including the area that is now Fort Drum, was converted to agriculture with forests remaining in relatively isolated patches on soils either too wet or too rocky for agriculture. The U.S. Army began acquiring land in the early 1900’s and expanded the installation to over 35,000 ha during World War II. Fort Drum reached its current extent in the 1980’s as new facilities were developed to house and train the U.S. Army’s 10th Mountain Division (U.S. Army, Fort Drum web site, accessed 10/27/16 at http://www.drum.army.mil/AboutFortDrum/Pages/History_lv2.aspx).

Developing Ecological Site Types Ecological site types incorporate aspects of the physical environment that influence species establishment and growth such as soil moisture, fertility and temperature (Rowe 1962; Barnes et al. 1982; Host et al. 1987). In the LANDIS-II forest landscape modeling application (Mladenoff and He 1999), ecological types are primarily used to define establishment probabilities for species of interest and along with species life history traits drive much of the successional dynamics over time in lieu of disturbance. However, site characteristics are also important inputs to modules developed to model stand productivity and carbon dynamics (Wang et al. 2014; Dymond et al. 2016), disturbance events (He and Mladenoff 1999; Sturtevant et al. 2004; Thompson et al. 2016), timber harvesting regimes (Shiflet et al. 2006) and potential impacts of climate change (Gustafson and Sturtevant 2013; Duveneck et al. 2014; Lucash et al. 2017). Environmental characteristics presumed to be important in determining species composition and stand productivity at Fort Drum (Table 1) were derived from county soil surveys for Jefferson and Lewis counties (McDowell 1989; Soil Survey Staff 2013) and 10-meter resolution digital 117 terrain data (New York State GIS Clearinghouse, https://gis.ny.gov/elevation/ accessed 5/4/2012). Topographic exposure and other measures of landscape position are often correlated with site conditions and species distributions (McNab 1993; Iverson et al. 1997; Odom and McNab 2000; Villwock et al. 2011), therefore topographic exposure index (TEI) was computed by subtracting the average elevation for a circular region with 1-km radius around each plot location from the elevation at each plot. Large negative values for TEI indicate relatively sheltered conditions topographically (surrounding elevations higher) and large positive values indicate relatively exposed conditions (surrounding elevations the same or lower). Due to the known moderating effects of Lake Ontario on local temperature, distance to the Lake was calculated using a simple Euclidean distance proximity function in a geographic information system (ArcGIS 10.6, Environmental Systems Research Institute, Redlands, CA). Values for all environmental attributes were assigned to forest inventory plot locations (n = 6,740) established as part of a systematic inventory completed in 2012 (Chris Dobony, Fort Drum Natural Resources Manager, personal communication). Site variables were mapped to visualize spatial patterns and continuous variables were analyzed in an exploratory manner using k-means clustering (JMP 13.2, SAS Institute, Cary NC) to identify groups of plots with similar environmental characteristics. Variables were standardized and transformed in the k-means analysis to minimize effects of different measurement scales, non-normal distributions and outliers. Iterative clustering and multivariate, nonparametric correlation analysis (Spearman’s ρ correlation coefficient, JMP 13.2, SAS Institute, Cary NC) was used to eliminate variables that were highly correlated or showed similar directionality in principal component plots.

Soil taxonomic classes integrate important soil characteristics such as parent material, texture, moisture and temperature regime to define dominant components of soil series (Soil Survey Staff 1999). Therefore, the aforementioned exploratory analyses were used to guide the aggregation of soil series at Fort Drum into ecological site types based on taxonomy and soil series descriptions (i.e., similar soil series were selected in attribute tables and assigned the same site type class; see Appendix A for a complete list of series assignments). This approach was important to maintain consistency between published county soil surveys that are widely used by resource managers and ecological site classifications. It was also instrumental in delineating wetlands and alluvial areas that were not well represented by soil and topographic variables assigned to forest

118 inventory plots. These areas contained relatively low amounts of merchantable timber and have management restrictions to maintain water quality, and therefore, relatively few sample points were located in these site types. However, wetlands and alluvial areas were included in the development of ecological site types because of their importance to site biodiversity and wildlife, particularly the endangered Myotis sodalis (Menzel et al. 2001; Jachowski et al. 2016), and because they cover a significant portion of the installation (approximately 15 percent).

To assess the explanatory strength of environmental variables and provide insight into classification thresholds for each type, site type classes were assigned to inventory plot locations and analyzed using recursive partitioning (JMP 13.2, SAS Institute, Cary NC). Recursive partitioning is a form of decision tree or classification and regression tree analysis (CART) that can accommodate continuous and categorical variables and does not assume that variables have normal distributions or linear relationships (Vayssières et al. 2000; McCune and Grace 2002). Variables were recursively split into two groups using optimal thresholds based on a likelihood ratio chi-square statistic (G^2) and adjusted p-value or LogWorth (Sall 2002). For each node or decision point in the classification tree, all possible splits of explanatory variables into response classes are considered and the largest LogWorth value determines the best variable and split value to use. A classification tree was developed using seven environmental traits as explanatory variables: five continuous variables identified through k-means cluster analysis and two categorical variables (soil drainage class and soil temperature class) with site type class as the response variable. The classification tree was subdivided until groups were relatively homogeneous (low misclassification rate) and increase in the overall correlation statistic became asymptotic. Eighty percent of the sample was used as training data and 20 percent was randomly selected and withheld for validation.

Species-Site Relationships Soil moisture content, nutrient availability and microclimate conditions that influence regeneration are important site characteristics of forest stands that may be significantly altered by changes in regional climate regimes (Rustad et al. 2012). If the climate in the northeastern U.S. becomes warmer and wetter over the next century as forecasted by many global climate models, tree species are likely to respond differently across variable environmental gradients at a landscape scale and therefore may or may not reflect general shifts in geographic distribution as 119 predicted by some species distribution models. To better understand how tree species might respond to increased evapotranspiration and concomitant decreases in available soil water during the growing season, species-site relationships were analyzed at Fort Drum to support landscape level simulations of potential changes in site conditions in response to a warmer climate. The relative importance (importance value or IV) was calculated for the 23 most common tree species by averaging basal area and abundance data for all forest inventory plots. Species with occurrence (percent of plots where a species was present) less than 5 percent were removed from the analysis because they were deemed unlikely to be important in determining major site types, their spatial distributions were limited and the identification of sites supporting rare or infrequent species was not a goal of the study. However, despite being present on only 2 percent of the plots, Carya cordiformis was included due to on-going research on forest bat habitat at Fort Drum (Jachowski et al. 2016) as well as evidence that this species along with Ulmus americana might be an indicator of base-rich soils on the installation. Soils derived from post-glacial lacustrine deposits underlain by limestone bedrock cover a large proportion of the western half of Fort Drum and the relatively high pH of these soils are hypothesized to be an important site differentiator relative to more acidic soils that characterize till uplands and sand plains. Relationships between species importance values and environmental gradients were analyzed visually by mapping species distributions for all plots using a standard cartographic classification scheme and quantitatively using indicator species analysis (PC-ORD version 7.01 McCune and Mefford 2015). Indicator species analysis attempts to identify species that are characteristic of forest or site types by comparing the relative abundance and relative frequency of each species within each type. Species that are strongly associated with a particular site type have relatively large indicator values in comparison to other species (McCune and Grace 2002). A Monte Carlo randomization procedure (p < 0.05; n = 1000) was used to test the significance of maximum species indicator values (IVmax) relative to the null hypothesis that IVmax was no larger than would be expected by chance.

Modeling Soil Moisture Deficits A climatic water balance model (Web-based, Water-Budget, Interactive, Modeling Program- WebWIMP, Willmott et al. 1985, accessed April 2018 at: http://climate.geog.udel.edu/~wimp/) was used to estimate changes in available soil water for seven ecological site types based on

120 projected changes in mean annual temperature and precipitation. WebWIMP calculates mean monthly water balance for any location using a modified Thornthwaite procedure to estimate evapotranspiration, run-off and soil water deficit (WD) for soils with specific available water holding capacities. The WebWIMP application has been used to estimate impacts of water stress on forest communities in California (Young et al. 2017) and elsewhere in North America (Hember et al. 2017). Climate projections from the Community Climate System Model version 3 (CCSM 3.0, NCAR GIS Program 2012) and average water holding capacity for the top 100 cm of soil (AWC100, SSURGO database for Jefferson County, New York; Soil Survey Staff 2013) were used to estimate soil water availability for ecological site types at years 2055 and 2095 under two climate change scenarios (Representative Concentration Pathway (RCP) 4.5 and 8.5; Intergovernmental Panel on Climate Change (IPCC) 2014). The RCP 4.5 scenario projects moderately fast increases in greenhouse gas concentrations and atmospheric temperatures in the first half of the century followed by a leveling-off of emissions due to presumed success of global conservation efforts, and is similar to the B1 scenario used in previous IPCC assessments. The RCP 8.5 scenario projects a consistently increasing trend in emissions throughout the 21st century based on little change in the current rate of greenhouse gas emissions, represents an upper boundary for future temperature predictions, and is similar to the A2 scenario used in previous IPCC assessments. The upper 100 cm of soil contains the majority of tree fine roots responsible for most water and nutrient uptake (Hinckley et al. 1981) and soil moisture has been shown to be an important driver of ecosystem properties such as rates of biogeochemical cycling (Groffman et al. 2012), species composition and distribution (Host and Pregitzer 1992) and annual net primary production (Baribault et al. 2010). Mean values for AWC100 for ecological site types at Fort Drum were determined by averaging values for individual soil series within each type. Changes in mean annual temperature (MAT) and mean annual precipitation (MAP) projected by the CCSM model for latitude 44.0o N, longitude -76.0o W were input to the WebWIMP application and annual water deficits were calculated for six water holding capacity levels; 50, 75, 100, 125, 150 and 300 mm of soil water that approximated mean values for AWC100 for each site type at Fort Drum. The default declining water availability function “G” was used to model actual and potential evapotranspiration, water deficits and surpluses. Future water deficits for each site type were calculated at years 2055 and 2095 under both emission

121

scenarios by comparing annual water deficits calculated by WebWIMP with extant soil available water capacities in the following manner:

WDRe,y,c mm = WDe,y,c mm / AWC100e,y=0 mm Where: WDR = water deficit ratio WD = water deficit calculated by WebWIMP AWC100 = available water capacity for top 100 cm of soil column e = ecological site type (1-7) y = model year (0, 40, 80) c = climate condition (MAT, MAP derived from RCP 4.5 and RCP 8.5)

The WDR was used to modify species establishment probabilities that define how well species are adapted to site conditions in the LANDIS-II Biomass Succession extension. Projected water deficit ratios that represent change in available soil water under different climatic conditions were added or subtracted to baseline species establishment probabilities (SEP) determined from previous research (Odom 2018b) for each species or vegetation cover type based on whether the change was considered positive or negative relative to species-site relationships (Table 2).

SEPs,e,y,c = SEPs,e,y=0 - WDRe,y,c (SEP decreases with water loss)

SEPs,e,y,c = SEPs,e,y=0,c + WDRe,y,c (SEP increases with water loss) Where: SEP = species establishment probability (0-1; negative values set to 0 and values > 1 set to 1) s = species or cover type

For example, decreasing soil moisture would presumably have a positive effect on establishment and growth for A. saccharum on a poorly drained wetland or alluvial site. However, decreasing moisture would presumably have a negative effect for this species on sites characterized by excessively drained sandy outwash or shallow, rocky soils. Species adaptations to water stress were based on published literature (Pastor and Post 1986; Burns and Honkala 1990; Gustafson and Sturtevant 2013) and expert opinion.

122

Data from published literature (Burns and Honkala 1990; Thompson et al. 2011; Duveneck et al. 2014) on species life histories, maximum ANPP (maxANPP) and maximum achievable above ground biomass (maxAGB) were used to parameterize the LANDIS-II Biomass Succession extension (Table 3). LANDIS-II uses species life history traits, species-age cohorts, site limitations on establishment and growth, competition and age-dependent mortality to model changes in species composition and biomass over time (Scheller and Mladenoff 2004). Species establishment probabilities derived from projected water deficits were applied to initial SEP to simulate change in available soil water under the RCP 4.5 and 8.5 scenarios. Models were run for a 100-year time period using a 20-year time step to produce forest type maps and biomass estimates for mid- and late century climate conditions. Model output was in the form of raster datasets with 30-meter resolution for each tree species or cover type, climate scenario and 20- year time step. Routines in ArcGIS Model BuilderTM (©Environmental Systems Research Institute, Inc., Redlands, CA 2018) were created to automate georegistration and create tabular attribute data for each raster file. Changes in biomass and forest types for years 2055 and 2095 were compared to baseline projections for year 2015 that did not include site modification from climate change.

Results Site classifications and species associations appeared to reflect major temperature, moisture and fertility gradients in the landscape. However, some environmental variables were excluded from the analysis based on multivariate correlations (Table 4) and weak performance in cluster analyses. Percent clay and available water capacity were strongly correlated (Spearman’s U = -0.94 and -0.70, respectively) with percent sand and nominal logistic regression of these variables indicated that percent sand had a larger effect in determining ecological site types (Wald’s χ2 = 1,698.7 for percent sand, 1,389.7 for available water capacity, and 178.0 for percent clay; p < 0.0001). Therefore, percent sand was retained in the analysis as the primary indicator of soil moisture gradients. Elevation and distance to Lake Ontario also were strongly correlated (Spearman’s U = 0.81); both were deemed surrogates for soil and microclimate temperature, but logistic regression indicated that elevation (Wald’s χ2 = 1,359.8, p < 0.0001) explained more variation than distance to Lake Ontario (Wald’s χ2 = 948.1, p < 0.0001) so the latter was excluded from further analyses. Five preliminary upland site types based on cluster analysis of

123

percent sand content, soil pH, elevation and topographic exposure index were identified (Figure 3). Biplot rays in the final principal component plot were largely orthogonal to each other suggesting that variables represented different gradients in multivariate space (Figure 3a). Mean cluster values for percent sand content and elevation were all significantly different (Figure 3b; p = 0.05, Dunn nonparametric multiple comparisons). Mean soil pH was significantly different for all clusters except 1 and 4, but there was considerable overlap between clusters based on mean values for topographic exposure index. However, low mean TEI values for clusters 1 and 2 (-5.2 and -4.5 respectively, red and green colors on map, Figure 3c) that are indicative of concave or more sheltered conditions were significantly different than means for more exposed sites (Figure 3c; clusters 3, 4 and 5).

Seven ecological site types at Fort Drum were identified based on soil taxonomy, series descriptions and topographic characteristics (Figure 4). Indicator species analysis identified several relatively strong associations between tree species and ecological site types (Table 5), but also revealed the ubiquitous nature of most species in this generally mesic and highly disturbed landscape. Wetland sites (WET) are very poorly drained, organic, silty soils in depressions or bordering lakes and streams that support woody shrubs (Salix spp., Alnus spp.) and tree species such as Fraxinus nigra, Acer rubrum, Betula alleghaniensis, Thuja occidentalis and to some degree eastern hemlock (Tsuga canadensis). Alluvial sites (ALL) occur along major streams and the Black River, have a variety of soil textures depending on adjacent soils, and have drainage characteristics ranging from very poorly to excessively drained depending on seasonal water levels. Relatively few tree species were uniquely and strongly associated with alluvial sites, but Acer rubrum, Prunus serotina and Ulmus americana were moderately important site indicators. Typical associates included Populus deltoides, Salix nigra and Fraxinus pennsylvanica. Similarly, no single tree species was strongly associated with mesic, fine-textured glacio- lacustrine deposits (LAC) that characterize much of the western third of the installation. Betula populifolia, Populus tremuloides and Ulmus americana were moderately strong indicator species, but almost all other common tree species were present to some degree on this site type with the exception of those that prefer very wet or very dry conditions. Well-drained, rocky soils derived from circumneutral glacial till are interspersed with lacustrine sediments (BAS site type). Species that prefer base-rich soils such as Acer saccharum, Carya cordiformis, Tilia americana

124

and Ulmus americana were strongly associated with this site type relative to other species. The southern portion of Fort Drum is comprised of excessively drained, sandy glacial outwash (SND site type) and represents the driest end of the soil moisture gradient on the installation. Indicator species for this site type were Pinus strobus, Pinus resinosa, Quercus rubra and Quercus alba. Mesic to subxeric, coarse loamy, glacio-fluvial terraces (TER site type) were transitional between lowland lacustrine deposits and higher sand plains. Acer rubrum and Tsuga canadensis were strongly associated with this type, but species associated with more mesic (Acer saccharum, Fagus grandifolia) and more xeric (Pinus strobus, Quercus rubra) sites were also common associates. Relatively shallow, loamy and frigid soils derived from acidic glacial till (TUP site classification) predominate in the northeastern quarter of the installation and support typical northern hardwood species. Acer saccharum, Acer rubrum, Fagus grandifolia, Betula papyrifera), Fraxinus americana and Ostrya virginiana were associated with this site type, but again many other tree species were also present.

Recursive partitioning analysis of seven environmental variables provided additional support for our classification of soil series into ecological site types (Figure 5), although the initial model did not predict any alluvial sites correctly. Topographic exposure, percent slope and soil temperature regime did not contribute significantly to the decision tree, but soil drainage class, percent sand content, elevation and soil pH explained 86 percent of the variation in the model. Percent sand content and soil drainage class accounted for the majority of the total variation (Figure 5a) suggesting that interaction between water holding capacity and drainage were the most important site differentiators. Despite indications from cluster analysis that soil pH was an important indicator for site types, it accounted for less than 3 percent of the variation in the decision tree model. The model had an overall misclassification rate of 5.6 percent, but alluvial sites were completely misclassified in training and validation datasets with the majority (61 of 69 in the training set) misclassified as upland till sites (Figure 5b). However, after including two additional splits in the decision tree (sand content less than 54 percent and soil pH greater than 6.2), 51 alluvial sites were correctly classified and the misclassification rate was reduced to 4.4 percent. The final classification tree was not shown for brevity since the overall correlation coefficient and contribution of explanatory variables were essentially unchanged from our initial model.

125

Mean annual temperature and mean annual precipitation are expected to increase at Fort Drum based on projections from the CCSM 3.0 ensemble model (Table 6) relative to mean conditions observed from 1986 to 2005. By mid-century, mean annual temperature is expected to increase by 2.1 to 2.8o C and mean annual precipitation by 19 to 55 mm (RCP 4.5 and 8.5 respectively). Near the end of the century, mean annual temperature could increase by as much as 4.7o C and precipitation could increase by 74 mm (RCP 8.5). Annual evapotranspiration is estimated to increase by as much as 22 percent under the high emissions RCP 8.5 scenario for soils with 150mm AWC. Changes in water deficits for ecological site types resulting from projected changes in climate ranged from 0 mm for the WET site type to a decrease of 57 mm for the SND type at year 2095 under the RCP 8.5 emissions scenario (Figure 6). The latter represents a loss of 85 percent of the total soil water available for plant growth on excessively drained sites. Water deficits were highly variable across site types and therefore the magnitude and directionality of species establishment probabilities varied with species-site interactions under different emission scenarios. For example, Acer saccharrum SEP increased markedly under mid-century conditions on poorly drained sites (LAC and TER site types), but declined on drier sites (SND, TER and BAS site types) (Figure 7). After initial increase or declines, Acer saccharrum SEP remained relatively constant over the latter half of the century, presumably due to the lower rate of temperature increase under RCP 4.5 and increased rainfall under the higher-emissions RCP 8.5 scenario. Other species showed similar trends in SEP values over time depending on their respective site adaptations.

For the installation as a whole, mean above ground biomass increased during the first 40 years, remained constant for 40 years and then decreased slightly during the last 20 years of the modeling period. This pattern was consistent across all ecological site types, however declines in biomass were less pronounced for sites with higher available water capacities (WET, ALL and TER site types) in comparison to sites with lower available water (Figure 8). Total projected biomass at year 2115 decreased by 1.0 percent under the low emissions RCP 4.5 scenario and 3.9 percent under the high emissions RCP 8.5 scenario. Biomass for early successional mesophytic species such as Populus tremuloides, Betula populifolia, Populus deltoides and Salix nigra increased dramatically from 2015 to 2055 under all scenarios and then declined sharply in the last 40 to 60 years of the modeling period (Figure 9). Approximately 50 percent of the

126

installation currently supports open grass-forb and woody shrub communities that have developed on abandoned farmland and other disturbed areas over the past 50 to 100 years. These areas were projected to transition to forest cover over the next 40 to 50 years and will be comprised largely of mesophytic species that are abundant and widely distributed at Fort Drum. Most intermediate to shade tolerant species increased in biomass over the duration of the simulation, although there were notable exceptions (Table 7). Biomass of Ulmus americana, Quercus alba, Tilia americana, Betula alleghaniensis, Carya cordiformis and Acer saccharum increased substantially while Tsuga canadensis made more modest gains. On xeric sites, Pinus strobus and Quercus rubra biomass increased slightly whereas P. tremuloides and P. grandidentata (-53 and -64 percent), Acer saccharum (-19 percent) and Tsuga canadensis (-13 percent) decreased substantially. Fagus grandifolia and Ostrya virginiana were the only shade tolerant species that consistently decreased in biomass throughout the simulation. The reason for this was not clear, although these species were the least abundant shade tolerant species at Fort Drum and perhaps could not compete effectively with more abundant and prolific species.

The proportion of the landscape covered by open old-field and shrub communities was projected to shrink by over 90 percent and be replaced largely by young mesophytic hardwood cover types. Xeric hardwoods (Quercus spp.) and conifers (Pinus spp.) increased slightly in areal coverage through 2055, but then declined by the end of the simulation. Rich mesophytic (A. saccharum, A. rubrum, U. Americana, C. cordiformia, T. americana and Q. alba) and northern hardwood (A. saccharum, B. alleghaniensis, F. grandifolia) types dominated by shade tolerant species increased substantially in areal coverage, but a large portion of the landscape remained in early successional types after 100 years of simulated succession (Figure 10).

Discussion Despite a positive trend in mean annual precipitation over the past 50 years and future projections from climate models for a 5 to 10 percent increase in precipitation over the next century, mean annual temperatures in northern New York State could increase by as much as 4.7o C by 2095 under high CO2-emission scenarios. This would lead to substantial soil water deficits in some forest ecosystems. The degree to which this might result in increased intensity or duration of drought conditions is not clear, although it appears likely that the probability of larger water deficits late in the growing season will increase (Hayhoe et al. 2008). If extended periods 127 of low precipitation occur in the future during the growing season, higher summer temperatures will increase evapotranspiration rates and thereby decrease soil water availability. Water deficits and their impacts on forest communities would likely be more pronounced on excessively drained sites with low water holding capacity (Figure 11). Due to substantial lake-effect snow in winter (mean = 2900 mm per year) and adequate precipitation during the growing season, soil water availability in the region is generally more than adequate to maintain healthy tree growth. However, this research based on temperature and precipitation projections from the CCSM-3 project suggests that soil water available for tree growth and reproduction could be reduced by as much as 85 percent on sites characterized by coarse textured or shallow, rocky soils at Fort Drum. This level of water shortage could lead to reductions in tree growth rates, reduced reproductive capacity and mortality resulting directly from carbon deprivation or indirectly through stress-induced disease syndromes or vulnerability to insect attack (McDowell et al. 2008). Deficits and potential impacts should be lower for soils with greater water holding capacity (> 100 mm/100 cm of soil depth; e.g., silt loam, clay loam and organic soils) or on concave landforms where surface and subsurface water accumulates. In addition, the species composition, age and structure of vegetation will affect how soil water is utilized on different sites. Therefore, water deficits and their effects on individual trees and forest communities are not likely to be homogeneous within forest landscapes where past land use, established successional trajectories, site characteristics, species-site adaptations and interspecific competition for resources are highly variable (Thompson et al. 2013; Duveneck et al. 2017).

An approach was developed to assess large-scale (> 1:24,000) landscape variability at Fort Drum incorporating extant data from county soil surveys, digital elevation models, land cover maps and a commercial timber inventory to model ecological site types, species distributions, species- site relationships and water deficits in a highly disturbed 43,000-ha landscape. Soil taxonomic classes incorporate characteristics such as parent material, texture, drainage, temperature regime and depth that are important influences on site quality, regeneration and forest productivity. However, soil series and phases as mapped by soil scientists often need to be aggregated, reorganized or integrated with other ecological data to provide more meaningful information for ecological analyses (Grigal 2009; Jiang et al. 2015). Soil and topographic characteristics related to moisture, temperature and fertility gradients were integrated to identify biophysical landscape

128

units (ecological site types) that were related to tree species distributions, relative abundance and relative importance. Soil drainage and water holding characteristics were the most important determinants of ecological site types, but topography and soil pH were also important in identifying specific site types and species associations. The development of a predictive site classification model was not an initial goal of the study, however, the regression tree developed explained a high degree of variability among environmental variables, highlighted interactions between classification components and had a low misclassification rate. This approach to forest site classification could be modified for other ecoregions in the eastern U.S. and applied to forest landscapes managed by the Department of Defense at other installations to support climate change assessments that in turn can be related to other management and training components on the landscape. More importantly, we were able to relate soil water capacities of sites at Fort Drum to estimates of future water deficits using a water balance calculator and inputs from a downscaled climate model. This allowed the projection of potential water deficits into the future in a site-specific manner and potential species responses to changing environmental conditions based on perceived adaptations to soil moisture regimes.

Understanding species-site relationships within an ecoregional context is an important consideration for predicting how species will respond to future climate conditions . Adaptations to climate and other environmental conditions have evolved over thousands of years and are largely an expression of cumulative genotypic variation within a species (Davis and Shaw 2001). However, within eastern North America, populations of many tree species have been geographically distinct for thousands of years and may have developed local adaptations to environmental stresses, especially near their range limits where stress events would presumably be more frequent or more intense (e.g., more frequent exposure to temperature minima at northern range limits versus increased probability of severe moisture stress at southern limits) (Leites et al. 2012; Potter et al. 2012; Fei et al. 2017). Environmental gradients that influence local species distribution, abundance, growth and survival may interact differently or one factor may be more important than another in different ecological settings (Fisichelli et al. 2013). As a result, coarse-scale predictions of species declines or expansions across entire ranges may not accurately reflect regional responses of distinct populations to climate change (Bussotti et al. 2015). Uncertain outcomes of interspecific competition and the stochastic nature of many natural

129 disturbances presents a formidable challenge in quantifying species-site relationships in mixed species forests. This challenging task is made even more complex by the different land use legacies and on-going anthropogenic disturbances at Fort Drum. In addition to a large proportion of the landscape being in transition from post-agricultural old-field communities, forest stands in the northeastern portion of the installation have been subjected to heavy diameter-limit timber harvests over the past 10 to 20 years (C. Dobony, personal communication), that has shifted species demographics towards younger, shade intolerant species (mostly Prunus serotina, Betula alleghaniensis and Fraxinus americana) while leaving residual stems of more shade tolerant northern hardwoods (Nowacki and Abrams 2008; Deluca et al. 2009). These two-aged, mixed stands introduce additional variability in species-site analyses and may confound apparent relationships between environmental gradients and the spatial distribution of species.

In this analysis, a few tree species were moderately well correlated with ecological site types and appeared to be distributed in a recognizable pattern along moisture and fertility gradients. Quercus and Pinus species were more important on coarse textured and excessively drained sites and species such as Abies balsamea, Fraxinus nigra, Thuja occidentalis, Acer rubrum and Tsuga canadensis were more prevalent on poorly drained sites associated with non-forested wetlands and alluvial stream terraces. Species with affinities for base-rich soils such as Acer saccharum, Ulmus americana, Tilia americana and Carya cordiformis were relatively more abundant on well-drained glacial till derived from limestone or base-rich sandstone. However, A. saccharum was also relatively abundant on acidic upland till sites, although mean above ground biomass projected by LANDIS-II was almost four times greater on base-rich sites (1,917 g m2) than on glacial till derived from acidic parent materials (510 g m2). This could be an artifact of biomass reduction from timber harvests on upland sites, but mean diameter and basal area were similar for stands in both areas. Early successional species such Prunus serotina, Betula populifolia, Populus tremuloides, Fraxinus americana and Acer rubrum were common throughout the study site, except on excessively drained, sandy soils where mean importance and biomass were relatively low for these species. Tsuga canadensis importance and biomass were significantly higher on glacio-fluvial terraces located between xeric sand plains and fine textured lacustrine deposits. Predominant soil textures in this site class were relatively coarse sandy loams, but were also poorly to somewhat poorly drained according to soil surveys. This seems consistent with

130 site preferences of T. canadensis in other ecological settings (e.g., southern and central Appalachian Mountains) where soils are consistently moist, but have adequate internal and surface drainage (Burns and Honkala 1990). Other than on these sites, T. canadensis was less common than expected at Fort Drum in comparison to forests in the Adirondack foothills region as a whole.

The LANDIS-II forest landscape simulator requires a significant amount of data preparation, parameterization and interpretation of output to gain insight into complex ecological processes and patterns that characterize forest ecosystems. There are numerous opportunities to incorporate data with implicit and untested assumptions that are not fully supported by empirical observations and are difficult to independently verify. One of the more complex tasks is to formulate species-age cohorts at a stand, management unit or landscape level that approximates what is actually on the ground. To date, the majority of research utilizing LANDIS-II has been applied at regional or smaller scales and relied heavily on FIA datato define inputs such as spatially explicit species-age cohorts. Applying LANDIS-II at a management scale in a highly disturbed landscape while trying to accurately emulate mixed-species forest communities presented unique challenges. Simulations may have been confounded by an attempt to include most of the common tree species and cover types that occur at Fort Drum. This led to a large number of species-age cohorts being modeled, which may have inflated biomass estimates (Robert Scheller, personal communication, 9/4/2018) and certainly made interpretation of model results more difficult. Collapsing community types into fewer classes and arbitrarily limiting species membership to one or two types could improve interpretability, but could also result in important community types and competitive dynamics being excluded or obfuscated. In addition, the forest type reclassification procedure employed by LANDIS-II is not completely transparent and very different classifications can result from what appear to be relatively inconsequential changes in the reclassification schema. This does not affect the competitive processes simulated or biomass projections in LANDIS-II, but could easily result in misrepresentation of the composition and landscape proportions of forest type maps that are common output from LANDIS-II simulations. For example, in some cases, initial cover types modeled at the onset of succession and biomass computations did not match community types used to parameterize the model. This was most evident for old-field and shrub community types that were reclassified as

131

Acer spp. – Ulmus spp. forest at the first time step in the model. LANDIS-II may not incorporate algorithms necessary to accurately represent succession from open grass-shrub communities to forest in early stages of the modeling process and the inclusion of tree species as components of shrub cover types may have created a bias in the reclassification process. There also appears to be a tendency for the LANDIS-II Biomass Succession module to overestimate biomass during model spin-up and exceed maximum allowed biomass parameters for species over long time periods (Simons-Legaard et al. 2015). The model is sensitive to the maxANPP parameter that controls the rate of biomass accumulation and maximum allowed biomass for each species. Published parameters for maxANPP and maxAGB from other research in the region were used to parameterize models at Fort Drum (Thompson et al. 2011; Wang et al. 2017), but estimates appeared to be relatively high in comparison to these studies and other estimates of biomass accumulation in northern hardwood ecosystems (Bormann and Likens 1979).

Conclusions Agricultural land use legacies and on-going disturbances from military training and timber harvesting have resulted in the dominance of early successional, mesophytic woody shrub and tree species across much of the Fort Drum landscape. Relatively older (> 80 years), closed- canopy forest stands comprised of more shade tolerant species currently comprise only 10 percent of the landscape (Odom 2018b). Clearly, most forest stands at Fort Drum are in the aggrading stage of forest stand development and strong species-site relationships may not be evident for many decades (Bormann and Likens 1979). Reductions in available soil water resulting from projected increases in evapotranspiration should be related to future species establishment and growth that would in turn influence species distributions and biomass projections simulated by the LANDIS-II model. It was expected that mean species importance and mean biomass would differ between ecological site types that integrated soil moisture, temperature and fertility gradients in the landscape. However, with the exception of slight increases in Quercus and Pinus biomass on the most xeric sites, biomass trends over time under moderate- and high-emission climate scenarios were not significantly different than what would be expected under established patterns of community development and current climate conditions. Change in species abundance as reflected by the spatial and temporal distribution of community types and biomass accumulation over time, within the context of species site

132 requirements, appeared to follow established patterns of forest community development that are largely determined by time since disturbance and relative shade tolerance (Bormann and Likens 1979; Nowacki and Abrams 2008; Thompson et al. 2013).

Acute water deficits resulting from increased evapotranspiration or late-season droughts may create opportunities for Quercus and Pinus species to increase importance on well-drained sites by decreasing competition from mesophytic hardwoods. However, unless droughts are of sufficient intensity or duration to create significant mortality in the overstory, it seems unlikely that more shade intolerant species will displace Acer saccharum, A. rubrum and other shade tolerant mesophytic species (Host et al. 1987; Carleton 2003; Nowacki and Abrams 2008). These species appear to be adapted to a wide range of environmental conditions as evidenced by their large geographic ranges and it is likely that they will continue to dominate forest landscapes at Fort Drum over the next century even under a warmer climate. Military training activities, timber harvests and extant insect and disease problems (e.g., Emerald ash borer Agrilus planipennis); beech bark disease Neonectria faginata and N. ditissima; Dutch elm disease Ophiostoma spp.) have altered stand structure in many areas on the installation and created gaps for opportunistic species. However, research and monitoring is needed to assess how extant species are responding to these disturbances and how climate change may interact with them to influence stand and landscape dynamics over time.

133

References Allen, C.D., A.K. Macalady, H. Chenchouni, et al. 2010. A global overview of drought and heat- induced tree mortality reveals emerging climate change risks for forests. Forest Ecology and Management 259:660–684.

Amber J. S., N. M. Tchebakova, N.H.F. French, M. Flannigan, H. H. Shugart, et al. 2007. Climate-induced boreal forest change: predictions versus current observations. Global and Planetary Change. 56 (3–4):274-296.

Baribault, T. W., R. K. Kobe, and D. E. Rothstein. 2010. Soil calcium, nitrogen, and water are correlated with aboveground net primary production in northern hardwood forests. Forest Ecology and Management 260:723-733.

Barnes, B., K. Pregitzer, T. Spies, and V. Spooner. 1982. Ecological forest site classification. Journal of Forestry 80:493-498.

Bechtold, W.A.and P.L. Patterson, editors. 2005. The enhanced Forest Inventory and Analysis program – national sampling design and estimation procedures. Gen. Tech. Rep. SRS-80. Asheville, NC: U.S. Department of Agriculture Forest Service, Southern Research Station. 85 p.

Beckage, B., B. Osborne, D. G. Gavin, C. Pucko, T. Siccama, and T. Perkins. 2008. A rapid upward shift of a forest ecotone during 40 years of warming in the Green Mountains of Vermont. Proceedings of the National Academy of Sciences USA 105:4197–4202.

Becknell, J. M., A. R. Desai, M. C. Dietze, C. A. Schultz, G. Starr, P. A. Duffy, J. F. Franklin, A. Pourmokhtarian, J. Hall, P. C. Stoy, M. W. Binford, L. R. Boring, and C. L. Staudhammer. 2015. Assessing interactions among changing climate, management, and disturbance in forests: A macrosystems approach. BioScience 65:263-274.

Boisvert-Marsh, L., C. Périé, S. de Blois. 2014. Shifting with climate? Evidence for recent changes in tree species distribution at high latitudes. Ecosphere 5(7):83.

134

Bormann, F.H. and G.E. Likens. 1979. Pattern and process in a forested ecosystem. New York: Springer-Verlag, 253p.

Brzostek, E. R., D. Dragoni, H. P. Schmid, A. F. Rahman, D. Sims, C. A. Wayson, D. J. Johnson, and R. P. Phillips. 2014. Chronic water stress reduces tree growth and the carbon sink of deciduous hardwood forests. Global Change Biology 20:2531-2539.

Burns, R.M and B.H. Honkala [Technical coordinators]. 1990. Silvics of North America: Volume 1. Conifers. Volume 2. Hardwoods. United States Department of Agriculture (USDA), Forest Service, Agriculture Handbook 654.

Bussotti, F., M. Pollastrini, V. Holland, and W. Brüggemann. 2015. Functional traits and adaptive capacity of European forests to climate change. Environmental and Experimental Botany 111:91-113.

Canham, C. D. and R. Q. Thomas. 2010. Frequency, not relative abundance, of temperate tree species varies along climate gradients in eastern North America. Ecology 91:3433–3440.

Carleton, T J. 2003. Old growth in the Great Lakes forest. Environmental Reviews 11:S115- S134.

Clark, J.S., et al. 2001. Ecological Forecasts: An Emerging Imperative. Science 293(5530):657– 660.

Clark, J. S., L. Iverson, C. W. Woodall, C. D. Allen, et al. 2016. The impacts of increasing drought on forest dynamics, structure, and biodiversity in the United States. Global Change Biology 22:2329-2352.

Cleland, D.T., J.A. Freeouf, J.E. Keys, G.J. Nowacki, C.A. Carpenter, and W.H. McNab. 2007. Ecological Subregions: Sections and Subsections for the conterminous United States. Gen. Tech. Report WO-76D [Map on CD-ROM] (A.M. Sloan, cartographer). Washington, DC: U.S. Department of Agriculture, Forest Service, presentation scale 1:3,500,000; colored.

135

Clinton, B.D., L.R. Boring, and W.T. Swank. 1993. Canopy gap characteristics and drought influences in oak forests of the Coweeta basin. Ecology 74(5):1551-1558.

Davis, M. B., and R. G. Shaw. 2001. Range Shifts and Adaptive Responses to Quaternary Climate Change. Science 292:673. de Bruijn, A., E.J. Gustafson, B.R. Sturtevant, J. R. Foster, B.R. Miranda, N.I. Lichti, and D.F. Jacobs. 2014. Toward more robust projections of forest landscape dynamics under novel environmental conditions: Embedding PnET within LANDIS-II. Ecological Modelling 287:44–57.

Deluca, T., M.A. Fajvan, and G. Miller. 2009. Diameter-limit harvesting: effects of residual trees on regeneration dynamics in Appalachian hardwoods. Northern Journal of Applied Forestry 26 (2):52-60.

Duveneck, M.J., R.M. Scheller, M.A. White, S.D. Handler, and C. Ravenscroft. 2014. Climate change effects on northern Great Lake (USA) forests: a case for preserving diversity. Ecosphere 5:1–26.

Duveneck, M.J., J.R. Thompson, E.J. Gustafson, Y. Liang, and A. de Bruijn. 2017. Recovery dynamics and climate change effects to future New England forests. Landscape Ecol 32: 1385.

Dymond, S.F., R.K. Kolka, P.V. Bolstad, and S.D. Sebestyen. 2014. Long-term soil moisture patterns in a northern Minnesota forest. Soil Science Society of America Journal 78 (S1):S208-S216.

Dymond, S.F., A.W. D'Amato, R.K. Kolka, P.V. Bolstad, S.D. Sebestyen, and J.B. Bradford. 2016. Growth-climate relationships across topographic gradients in the northern Great Lakes. Ecohydrology. 9(6):918-929.

Fei, S., J. M. Desprez, K. M. Potter, I. Jo, J. A. Knott, and C. M. Oswalt. 2017. Divergence of species responses to climate change. Science Advances 3(5) e1603055 DOI: 10.1126/sciadv.1603055. 136

Fisichelli, N., L. Frelich, and P. Reich. 2013. Climate and interrelated tree regeneration drivers in mixed temperate–boreal forests. Landscape ecology 28:149-159.

Foster, J. R., and A. W. D'Amato. 2015. Montane forest ecotones moved downslope in northeastern USA in spite of warming between 1984 and 2011. Global Change Biology 21:4497-4507.

Goring, S. J., D. J. Mladenoff, C. V. Cogbill, S. Record, C. J. Paciorek, S. T. Jackson, M. C. Dietze, A. Dawson, J. H. Matthes, J. S. McLachlan, and J. W. Williams. 2016. Novel and lost forests in the upper midwestern United States, from new estimates of settlement-era composition, stem density, and biomass. PLoS ONE 11:1-34.

Grigal, D. F. 2009. A soil-based aspen productivity index for Minnesota. Forest Ecology and Management 257:1465-1473.

Groffman P.M. et al. 2012. Long-term integrated studies show complex and surprising effects of climate change in the northern hardwood forest. BioScience 62:1056-1066.

Gustafson, E. J. 2014. Applicability of predictive models of drought-induced tree mortality between the Midwest and Northeast United States. Forest Science 60(2):327–334

Gustafson, E. J., and B. R. Sturtevant. 2013. Modeling forest mortality caused by drought stress: implications for climate change. Ecosystems 16:60–74.

Gustafson, E. J., A. M. G. De Bruijn, B. R. Miranda, and B. R. Sturtevant. 2016. Implications of mechanistic modeling of drought effects on growth and competition in forest landscape models. Ecosphere 7(4):e01253. 10.1002/ecs2.1253.

Hanberry, B. B. 2013. Changing eastern broadleaf, southern mixed, and northern mixed forest ecosystems of the eastern United States. Forest Ecology and Management 306:171-17.

Hanson, P. J., and J. F. Weltzin. 2000. Drought disturbance from climate change: response of United States forests. Science of The Total Environment 262:205-220.

137

Hayhoe, K., C. Wake, B. Anderson, X.-Z. Liang, E. Maurer, J. Zhu, J. Bradbury, A. DeGaetano, A. Stoner, and D. Wuebbles. 2008. Regional climate change projections for the Northeast USA. Mitigation and Adaptation Strategies for Global Change 13:425-436.

He, H. S., and D. J. Mladenoff. 1999. Spatially explicit and stochastic simulation of forest landscape fire disturbance and succession. Ecology 80:81-99.

Hember, R.A., W.A. Kurz, and N.C. Coops. 2017. Relationships between individual-tree mortality and water-balance variables indicate positive trends in water stress-induced tree mortality across North America. Global Change Biology 23:1691-1710.

Hinckley, T. M., R. O. Teskey, F. Duhme, and H. Richter. 1981. Temperate Hardwood Forests. Chapter 3, pp.153-208 in Water Deficits and Plant Growth: Vol. VI Woody Plant Communities. T. T. Kozlowski, Ed. New York: Academic Press, 1981.

Host, G.E., K.S. Pregitzer, C.W. Ramm, J.B. Hart, and D.T. Cleland. 1987. Landform-mediated differences in successional pathways among upland forest ecosystems in northwestern Lower Michigan. Forest Science 33( 2): 445–457.

Host, G.E. and K.S. Pregitzer. 1992. Geomorphic influences on ground-flora and overstory composition in upland forests of northwestern lower Michigan. Canadian Journal of Forest Research 22(10):1547-1555.

Intergovernmental Panel on Climate Change (IPCC). 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp.

138

Iverson, L.R., M.E. Dale, C.T. Scott, and A. Prasad. 1997. A GIS-derived integrated moisture index to predict forest composition and productivity of Ohio forests (U.S.A.) Landscape Ecology (1997) 12(5): 331-348.

Iverson, L. R., A. M. Prasad, S. N. Matthews, and M. Peters. 2008. Estimating potential habitat for 134 eastern U.S. tree species under six climate scenarios. Forest Ecology and Management. 254:390-406.

Iverson, L. R., A. M. Prasad, S. N. Matthews, and M. Peters. 2011. Lessons learned while integrating habitat, dispersal, disturbance, and life-history traits into species habitat models under climate change Ecosystems 14:1005-1020.

Iverson, L. R. and D. McKenzie. 2013. Tree-species range shifts in a changing climate: detecting, modeling, assisting. Landscape Ecology 28:879–889.

Jachowski, D.S., C.T. Rota, C.A. Dobony, W.M. Ford, and J.W. Edwards. 2016. Seeing the forest through the trees: considering roost-site selection at multiple spatial scales. PLoS ONE 11(3):e0150011. doi:10.1371/journal.pone.0150011.

Janowiak, M.K., A.W. D'Amato, C.W. Swanston, et al. 2018. New England and northern New York forest ecosystem vulnerability assessment and synthesis: a report from the New England Climate Change Response Framework project. Gen. Tech. Rep. NRS-173. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 234 p.

Jiang, H., P. Radtke, A. Weiskittel, J. Coulston, and P. Guertin. Climate- and soil-based models of site productivity in eastern U.S. tree species. Canadian Journal of Forest Research 45(3):325-342.

Klos, R. J., G. G. Wang, W. L. Bauerle, and J. R. Rieck. 2009. Drought impact on forest growth and mortality in the southeast USA: an analysis using Forest Health and Monitoring data. Ecological Applications 19:699-708.

139

Lee, T.D., J.P. Barrett, and B. Hartman. 2005. Elevation, substrate, and the potential for climate- induced tree migration in the White Mountains, New Hampshire, USA. Forest Ecology and Management 212 (1–3):75-91.

Leites, L. P., G. E. Rehfeldt, A. P. Robinson, N. L. Crookston, and B. Jaquish. 2012. Possibilities and limitations of using historic provenance tests to infer forest species growth responses to climate change. Natural Resource Modeling 25:409-433.

Loehle, C. 2000. Forest ecotone response to climate change: sensitivity to temperature response functional forms. Canadian Journal of Forest Research 30:1632-1645.

Lucash, M.S., R.M. Scheller, E.J. Gustafson, and B.R. Sturtevant. 2017. Spatial resilience of forested landscapes under climate change and management. Landscape Ecology 32: 953.

Martin-Benito,D. and N. Pederson. 2015. Convergence in drought stress, but a divergence of climatic drivers across a latitudinal gradient in a temperate broadleaf forest. Journal of Biogeography 42:925–937.

McCullough, I.M., F.W. Davis, J.R. Dingman, et al. 2016. High and dry: high elevations disproportionately exposed to regional climate change in Mediterranean-climate landscapes Landscape Ecology 31:1063.

McCune, B., and J.B. Grace . 2002. Analysis of ecological communities. Gleneden Beach, Oregon: MjM Software Design. 300p.

McCune, B. and M. J. Mefford. 2015. PC-ORD. Multivariate Analysis of Ecological Data. Version 7.01. MjM Software, Gleneden Beach, Oregon, U.S.A.

McDowell, L. 1989. Soil Survey of Jefferson County, New York. Washington, D.C.: USDA Soil Conservation Service. 210p.

McDowell, N., W.T. Pockman, C.D. Allen, D.D. Breshears, N. Cobb, T. Kolb, J. Plaut, J. Sperry, A. West, D.G. Williams and E.A. Yepez. 2008. Mechanisms of plant survival and

140

mortality during drought: why do some plants survive while others succumb to drought? New Phytologist (2008) 178: 719–739.

McKenney, D.W., J.H. Pedlar, K.M. Lawrence, K.L. Campbell, and M.F. Hutchinson. 2007. Potential impacts of climate change on the distribution of North American trees. BioScience 57:939 - 948.

McKenney, D.W., J.H. Pedlar, R.B. Rood, and D. Price. 2011. Revisiting projected shifts in the climate envelopes of North American trees using updated general circulation models. Global Change Biology, 17:2720–2730.

McNab, W.H. 1993. A topographic index to quantify the effect of mesoscale landform on site productivity. Canadian Journal of Forest Research. 23: 1100-1107.

Menzel, M.A., J.M. Menzel, T.C. Carter, W.M. Ford, and J.W. Edwards. 2001. Review of the forest habitat relationships of the Indiana bat (Myotis sodalis). Gen. Tech. Rep. NE-284. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northeastern Research Station. 21 p.

Mladenoff, D.J. 2004. LANDIS and forest landscape models. Ecological Modelling 180 (2004) 7–19.

National Center for Atmospheric Research (NCAR) GIS Program. 2012. Climate Change Scenarios, version 2.0. Community Climate System Model, June 2004 version 3.0. http://www.cesm.ucar.edu/models/ccsm3.0/ was used to derive data products. Downscaling procedure in: T. Hoar and D. Nychka, 2008, Statistical Downscaling of the Community Climate System Model (CCSM) monthly temperature and precipitation projections,

141

unpublished white paper available at http://gisclimatechange.ucar.edu/sites/default/ files/users/Downscaling.pdf. Accessed 04/07/2018.

Nolet, P., and D. Kneeshaw. 2018. Extreme events and subtle ecological effects: lessons from a long-term sugar maple–American beech comparison. Ecosphere 9(7):e02336.

Nowacki, G.J., and M.D. Abrams. 2008. The demise of fire and the “mesophication” of forests in the eastern united states. Bioscience 58:123–138.

Nowacki, G.J., and M.D. Abrams. 2015. Is climate an important driver of post-European vegetation change in the Eastern United States? Global Change Biology 21(1):314-334.

O’Connell, B.M. et al. 2014. The Forest Inventory and Analysis database: Database description and user guide version 6.0.1 for Phase 2. U.S. Department of Agriculture, Forest Service. 748 p.

Odom, R.H. 2018a. Assessing the vulnerability of military installations in the coterminous United States to potential biome shifts resulting from rapid climate change (doctoral dissertation). Unpublished manuscript, Virginia Polytechnic Institute and State University, Blacksburg, VA.

Odom, R.H. 2018b. Developing species-age cohorts from forest inventory and analysis data to parameterize a forest landscape model (doctoral dissertation). Unpublished manuscript, Virginia Polytechnic Institute and State University, Blacksburg, VA.

Odom, R.H., and W.H. McNab. 2000. Using digital terrain modeling to predict ecological types in the Balsam Mountains of western North Carolina. USDA Forest Service Research Note SE-8. 12p.

Ollinger, S.V., C.L. Goodale, K. Hayhoe, and J.P. Jenkins. 2008. Potential effects of climate change and rising CO2 on ecosystem processes in northeastern U.S. forests. Mitigation and Adaptation Strategies for Global Change 13:467-485.

Palmer, W. C. 1965: Meteorological drought. U.S. Weather Bureau Research Paper 45, 58 pp. 142

Payette, S., M.-J. Fortin, and C. Morneau. 1996. The recent sugar maple decline in southern Quebec: probable causes deduced from tree rings. Canadian Journal of Forest Research 26:1069-1078.

Pastor, J., and W.M. Post. 1986. Influence of climate, soil moisture, and succession on forest carbon and nitrogen cycles. Biogeochemistry 2:3–27.

Potter, K. M., R. M. Jetton, W. S. Dvorak, V. D. Hipkins, R. Rhea, and W. A. Whittier. 2012. Widespread inbreeding and unexpected geographic patterns of genetic variation in eastern hemlock (Tsuga canadensis), an imperiled North American conifer. Conservation Genetics 13:475-498.

Rowe, J.S. 1962. Soil, site and land classification. Forestry Chronicle 38:420–432.

Rustad, L., J. Campbell, J.S. Dukes, T. Huntington, K.F. Lambert, J. Mohan, and N. Rodenhouse. 2012. Changing climate, changing forests: The impacts of climate change on forests of the northeastern United States and eastern Canada. Gen. Tech. Rep. NRS-99. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northern Research Station. 48 p.

Sall, J. 2002. Monte Carlo calibration of distributions of partition statistics. SAS Institute. Retrieved July 29, 2015 from http://www.jmp.com/content/dam/jmp/documents/en/white- papers/montecarlocal.pdf.

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.

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 spatial and temporal resolution. Ecol Mod 201:409-419.

143

Shuman, J. K. and H. H. Shugart. 2009. Evaluating the sensitivity of Eurasian forest biomass to climate change using a dynamic vegetation model. Environmental Research Letters, 4 045024.

Simons-Legaard, E., K. Legaard, and A. Weiskittel. 2015. Predicting aboveground biomass with LANDIS-II:A global and temporal analysis of parameter sensitivity. Ecological Modelling 313 (2015) 325–332.

Soil Survey Staff. 2013. Soil Survey Geographic (SSURGO) Database for Jefferson and Lewis Counties, New York. United States Department of Agriculture, Natural Resources Conservation Service. Available at https://sdmdataaccess.sc.egov.usda.gov. Accessed 04/28/2013.

Soil Survey Staff. 1999. Soil Taxonomy: A Basic System of Soil Classification for Making and Interpreting Soil Surveys. United States Department of Agriculture, Natural Resources Conservation Service, Agriculture Handbook Number 436, 886p.

Swanston, C., L.A. Brandt, M.K. Janowiak, S.D. Handler, P. Butler-Leopold, L. Iverson, F.R. Thompson III, T.A. Ontl, and P.D. Shannon. 2018. Vulnerability of forests of the Midwest and Northeast United States to climate change. Climatic Change 146:103-116.

Tang, G., B. Beckage, and B. Smith. 2012. The potential transient dynamics of forests in New England under historical and projected future climate change. Climatic Change (2012) 114:357.

Thompson, J.R., D.N. Carpenter, C.V. Cogbill, and D.R. Foster. 2013. Four centuries of change in northeastern United States forests. PLoS ONE 8(9): e72540.

144

Thompson J.R., E. Simons-Legaard, K.R. Legaard, and J.B. Domingo. 2016. A LANDIS-II extension for incorporating land use and other disturbances. Environmental Software and Modeling. 75:202-205

Vadeboncoeur M.A., S.P. Hamburg, C.V. Cogbill, and W.Y. Sugimura. 2011. A comparison of presettlement and modern forest composition along an elevation gradient in central New Hampshire. Canadian Journal of Forest Research 42:190–202.

Vayssières, M. P., R. E. Plant, and B. H. Allen-Diaz. 2000. Classification trees: An alternative non-parametric approach for predicting species distributions. Journal of Vegetation Science 11:679-694.

Vose, J.M., J.S. Clark, C.H. Luce, and T. Patel-Weynand, eds. 2016. Effects of drought on forests and rangelands in the United States: a comprehensive science synthesis. Gen. Tech. Rep. WO-93b. Washington, DC: U.S. Department of Agriculture, Forest Service, Washington Office. 289 p.

Wang, W. J., H. S. He, M. A. Spetich, S. R. Shifley, F. R. Thompson, W. D. Dijak, and Q. Wang. 2014. A framework for evaluating forest landscape model predictions using empirical data and knowledge. Environmental Modelling & Software 62:230-239.

Wang, W.J., H.S. He, F.R. Thompson III, J.S. Fraser, and W.D. Dijak. 2017. Changes in forest biomass and tree species distribution under climate change in the northeastern United States. Landscape Ecology (2017) 32:1399–1413.

Wason, J. W. and M. Dovciak. 2017. Tree demography suggests multiple directions and drivers for species range shifts in mountains of Northeastern United States. Glob Change Biol, 23:3335–3347.

Willmott, C.J., C.M. Rowe and Y. Mintz, 1985. Climatology of the terrestrial seasonal water cycle. Journal of Climatology 5:589-606.

Worrall, J., G.E. Rehfeldt, A. Hamann, et al. 2013. Recent declines of Populus tremuloides in North America linked to climate. Forest Ecology and Management. 299:35–51. 145

Xu, C., G.Z. Gertner, and R.M. Scheller. 2012. Importance of colonization and competition in forest landscape response to global climatic change. Climatic Change 110:53–83.

Young, D.J., J.T. Stevens, J.M. Earles, J. Moore, A. Ellis, A.L. Jirka, A.M. Latimer, and F. Lloret, Long‐term climate and competition explain forest mortality patterns under extreme drought. 2017. Ecology Letters, 20, 1, (78-86), (2016).

146

Appendix A.

Table A-1. Reclassification of soil series into ecological site types (Jefferson and Lewis counties, New York, USA).

Ecological site Soil series Ecological site description type code Abram-Lyman complex Mesic frigid acidic glacial till TUP Adams, 0 to 8 percent slopes Mesic basic glacial till BAS Adams, 8 to 15 percent slopes Mesic basic glacial till BAS Adams, rocky, 0 to 8 percent slopes Mesic basic glacial till BAS Adams, rocky, 8 to 15 percent slopes Mesic basic glacial till BAS Adjidaumo, 0 to 3 percent slopes, rocky Mesic glacio-lacustrine plain LAC Agawam fine sandy loam, 0 to 3 percent slopes Mesic glacio-fluvial sand terrace TER Agawam fine sandy loam, 3 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Allis silt loam, 0 to 3 percent slopes Mesic basic glacial till BAS Alton gravelly loam, 25 to 45 percent slopes Mesic basic glacial till BAS Alton gravelly loam, 3 to 8 percent slopes Mesic basic glacial till BAS Alton gravelly loam, 8 to 15 percent slopes Mesic basic glacial till BAS Amenia loam, 0 to 3 percent slopes Mesic basic glacial till BAS Amenia loam, 3 to 8 percent slopes Mesic basic glacial till BAS Arkport fine sandy loam, 3 to 8 percent slopes Mesic basic glacial till BAS Arkport fine sandy loam, 8 to 15 percent slopes Mesic basic glacial till BAS Benson channery silt loam, very rocky, 25 to 50 percent slopes Mesic basic glacial till BAS Benson-Galoo complex, very rocky, 0 to 8 percent slopes Mesic basic glacial till BAS Bice fine sandy loam, 15 to 25 percent slopes Mesic frigid acidic glacial till TUP Bice fine sandy loam, 25 to 50 percent slopes Mesic frigid acidic glacial till TUP Bice fine sandy loam, 3 to 8 percent slopes Mesic frigid acidic glacial till TUP Bice fine sandy loam, 8 to 15 percent slopes Mesic frigid acidic glacial till TUP Bice very stony fine sandy loam, 0 to 15 percent slopes Mesic frigid acidic glacial till TUP Bombay loam, 0 to 3 percent slopes Mesic basic glacial till BAS Bombay loam, 3 to 8 percent slopes Mesic basic glacial till BAS

147

Ecological site Soil series Ecological site description type code Bonaparte gravelly loamy fine sand, 0 to 8 percent slopes Mesic basic glacial till BAS Bucksport and Wonsqueak soils, ponded Hydric depressional wetland WET Bucksport, 0 to 3 percent slopes Hydric depressional wetland WET Bucksport-Pondicherry complex, occasionally flooded Hydric depressional wetland WET Bucksport-Wonsqueak association, 0 to 2 percent slopes, frequently ponded Hydric depressional wetland WET Canandaigua mucky silt loam Mesic glacio-lacustrine plain LAC Canandaigua silt loam Mesic glacio-lacustrine plain LAC Carlisle muck Hydric depressional wetland WET Chatfield loam, rocky, 0 to 8 percent slopes Mesic frigid acidic glacial till TUP Chatfield-Rock outcrop complex, rolling Mesic frigid acidic glacial till TUP Chatfield-Rock outcrop complex, steep Mesic frigid acidic glacial till TUP Chaumont silty clay, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Chaumont silty clay, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Claverack loamy fine sand, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Claverack loamy fine sand, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Collamer silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Collamer silt loam, 8 to 15 percent slopes Mesic glacio-lacustrine plain LAC Collamer silt loam, bedrock substratum, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Covington silty clay Mesic glacio-lacustrine plain LAC Croghan, 0 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Deerfield loamy fine sand, 0 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Deinache, 0 to 3 percent slopes Mesic basic glacial till BAS Dumps Xeric sand plain SND Elmridge fine sandy loam, 0 to 3 percent slopes Mesic basic glacial till BAS Elmridge fine sandy loam, 3 to 8 percent slopes Mesic basic glacial till BAS Endoaquents Mesic glacio-lacustrine plain LAC Endoaquents, disrupted-Hailesboro-Wegatchie complex, 0 to 8 percent slopes Mesic glacio-lacustrine plain LAC Endoaquents, disrupted-Wonsqueak-Adjidaumo complex, 0 to 3 percent slopes Hydric depressional wetland WET Ensley very stony silt loam Hydric depressional wetland WET 148

Ecological site Soil series Ecological site description type code Farmington loam, 0 to 8 percent slopes Mesic basic glacial till BAS Fluvaquents-Udifluvents complex, frequently flooded Subhydric alluvial deposits ALL Galen fine sandy loam, 0 to 3 percent slopes Mesic glacio-fluvial sand terrace TER Galen fine sandy loam, 3 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Galoo, acid-Rock outcrop complex, 0 to 8 percent slopes Mesic basic glacial till BAS Galoo-Rock outcrop complex, 0 to 8 percent slopes Mesic basic glacial till BAS Galway silt loam, 0 to 3 percent slopes Mesic basic glacial till BAS Galway silt loam, 3 to 8 percent slopes Mesic basic glacial till BAS Galway silt loam, 8 to 15 percent slopes Mesic basic glacial till BAS Galway very stony silt loam, 0 to 15 percent slopes Mesic basic glacial till BAS Granby mucky loamy fine sand Mesic glacio-fluvial sand terrace TER Groton gravelly loam, 0 to 3 percent slopes Mesic basic glacial till BAS Groton gravelly loam, 25 to 35 percent slopes Mesic basic glacial till BAS Groton gravelly loam, 3 to 8 percent slopes Mesic basic glacial till BAS Groton variant gravelly loam, 0 to 8 percent slopes Mesic basic glacial till BAS Hailesboro, 3 to 8 percent slopes Hydric depressional wetland WET Hailesboro-Wegatchie-Insula association, 0 to 15 percent slopes, rocky Mesic glacio-lacustrine plain LAC Halsey mucky loam Mesic basic glacial till BAS Halsey mucky loam Hydric depressional wetland WET Hamlin silt loam Mesic basic glacial till BAS Heuvelton silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Heuvelton silt loam, 8 to 15 percent slopes Mesic glacio-lacustrine plain LAC Heuvelton-Millsite-Rock outcrop complex, undulating Mesic glacio-lacustrine plain LAC Heuvelton-Muskellunge-Millsite complex, 0 to 15 percent slopes, very rocky Mesic glacio-lacustrine plain LAC Hinckley gravelly sandy loam, 0 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Hinckley-Hoosic cobbly sandy loams, 0 to 8 percent slopes Mesic glacio-fluvial sand terrace TER Hollis-Galoo, acid, complex, rocky, 0 to 8 percent slopes Mesic basic glacial till BAS Hollis-Rock outcrop complex, 0 to 8 percent slopes Mesic basic glacial till BAS Hudson and Vergennes soils, 15 to 35 percent slopes, severely eroded Mesic glacio-lacustrine plain LAC 149

Ecological site Soil series Ecological site description type code Hudson silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Hudson silt loam, 8 to 15 percent slopes Mesic glacio-lacustrine plain LAC Hudson-Chatfield-Rock outcrop complex, undulating Mesic glacio-lacustrine plain LAC Insula-Millsite-Quetico-Rock outcrop complex, 3 to 15 percent slopes, very bouldery Mesic frigid acidic glacial till TUP Insula-Quetico complex, rocky, 0 to 8 percent slopes Mesic frigid acidic glacial till TUP Insula-Rock outcrop complex, 0 to 8 percent slopes Mesic frigid acidic glacial till TUP Junius loamy fine sand Mesic glacio-fluvial sand terrace TER Kings Falls, 3 to 8 percent slopes, rocky Mesic basic glacial till BAS Kingsbury silty clay, 0 to 2 percent slopes Mesic glacio-lacustrine plain LAC Kingsbury silty clay, 2 to 6 percent slopes Mesic glacio-lacustrine plain LAC Kingsbury-Livingston complex Mesic glacio-lacustrine plain LAC Lamson fine sandy loam Mesic glacio-fluvial sand terrace TER Livingston mucky silty clay Mesic glacio-lacustrine plain LAC Livingston silty clay loam, frequently flooded Mesic glacio-lacustrine plain LAC Lyman-Abram complex, 15 to 35 percent slopes, very bouldery, very rocky Mesic frigid acidic glacial till TUP Lyman-Abram complex, 3 to 15 percent slopes, very bouldery, very rocky Mesic frigid acidic glacial till TUP Lyman-Abram complex, 3 to 25 percent slopes, very rocky Mesic frigid acidic glacial till TUP Madalin silt loam Mesic glacio-lacustrine plain LAC Madrid sandy loam, 3 to 8 percent slopes Mesic basic glacial till BAS Madrid sandy loam, 8 to 15 percent slopes Mesic basic glacial till BAS Massena silt loam, 0 to 3 percent slopes Mesic basic glacial till BAS Massena silt loam, 3 to 8 percent slopes Mesic basic glacial till BAS Massena very stony loam, 0 to 8 percent slopes Mesic basic glacial till BAS Medomak, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Millsite loam, rocky, 0 to 8 percent slopes Mesic frigid acidic glacial till TUP Millsite-Rock outcrop complex, rolling Mesic frigid acidic glacial till TUP Millsite-Rock outcrop complex, steep Mesic frigid acidic glacial till TUP Minoa fine sandy loam Mesic glacio-fluvial sand terrace TER Muskellunge silt loam, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC 150

Ecological site Soil series Ecological site description type code Muskellunge silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Naumburg (p), 0 to 3 percent slopes Mesic glacio-fluvial sand terrace TER Naumburg (swp), 0 to 3 percent slopes Mesic glacio-fluvial sand terrace TER Naumburg-Lyman complex, 0 to 15 percent slopes, rocky Mesic glacio-fluvial sand terrace TER Nehasne-Kings Falls complex, 15 to 35 percent slopes, very bouldery, very rocky Mesic basic glacial till BAS Nellis and Madrid soils, steep Mesic basic glacial till BAS Nellis loam, 0 to 3 percent slopes Mesic basic glacial till BAS Nellis loam, 15 to 25 percent slopes Mesic basic glacial till BAS Nellis loam, 3 to 8 percent slopes Mesic basic glacial till BAS Nellis loam, 8 to 15 percent slopes Mesic basic glacial till BAS Newstead silt loam Mesic glacio-lacustrine plain LAC Niagara silt loam, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Niagara silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Niagara silt loam, bedrock substratum, 2 to 6 percent slopes Mesic glacio-lacustrine plain LAC Nicholville, 2 to 8 percent slopes Mesic glacio-lacustrine plain LAC Onjebonge, 0 to 3 percent slopes Hydric depressional wetland WET Palms muck Hydric depressional wetland WET Phelps gravelly loam, 0 to 3 percent slopes Mesic basic glacial till BAS Phelps gravelly loam, 3 to 8 percent slopes Mesic basic glacial till BAS Pits, quarry Developed and disturbed INA Pits, sand and gravel Mesic basic glacial till BAS Plainfield and Windsor soils, hilly Xeric sand plain SND Plainfield sand, 0 to 8 percent slopes Xeric sand plain SND Plainfield sand, altered surface, 0 to 8 percent slopes Xeric sand plain SND Plainfield sand, altered surface, 0 to 8 percent slopes Developed and disturbed INA Plainfield sand, altered surface, rolling Xeric sand plain SND Plainfield sand, rolling Xeric sand plain SND Pondicherry and Searsport soils, ponded Hydric depressional wetland WET Pondicherry, 0 to 3 percent slopes Hydric depressional wetland WET 151

Ecological site Soil series Ecological site description type code Pootatuck fine sandy loam Subhydric alluvial deposits ALL Quetico-Rock outcrop complex, 2 to 8 percent slopes Mesic frigid acidic glacial till TUP Rhinebeck silt loam, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Rhinebeck silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Rhinebeck-Chatfield-Rock outcrop complex, rolling Mesic glacio-lacustrine plain LAC Rock outcrop-Abram-Knob Lock complex, 15 to 35 percent slopes, very bouldery Mesic frigid acidic glacial till TUP Roundabout (p), 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Roundabout (swp), 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Roundabout-Onjebonge-Lyman association, 0 to 15 percent slopes, rocky Mesic glacio-lacustrine plain LAC Roundabout-Onjebonge-Lyman association, 0 to 15 percent slopes, rocky Hydric depressional wetland WET Rumney, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Ruse gravelly loam, rocky Mesic basic glacial till BAS Saprists and Aquents, ponded Hydric depressional wetland WET Scarboro mucky loamy fine sand Hydric depressional wetland WET Searsport, 0 to 3 percent slopes Hydric depressional wetland WET Shaker fine sandy loam Mesic basic glacial till BAS Sun silt loam Mesic glacio-lacustrine plain LAC Sun very stony silt loam Mesic glacio-lacustrine plain LAC Teel silt loam Mesic glacio-lacustrine plain LAC Tonowanda silt loam, 0 to 3 percent slopes Mesic glacio-lacustrine plain LAC Tunbridge-Lyman complex, 15 to 35 percent slopes, very bouldery, very rocky Mesic frigid acidic glacial till TUP Tunbridge-Lyman complex, 3 to 15 percent slopes, very bouldery, very rocky Mesic frigid acidic glacial till TUP Udorthents, disrupted-Insula-Rock outcrop complex, 3 to 15 percent slopes Mesic frigid acidic glacial till TUP Udorthents, refuse substratum Developed and disturbed INA Udorthents,smoothed Developed and disturbed INA Urban land Developed and disturbed INA Vergennes silty clay loam 8 to 15 percent slopes Mesic glacio-lacustrine plain LAC Vergennes silty clay loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Wareham loamy fine sand Mesic glacio-fluvial sand terrace TER 152

Ecological site Soil series Ecological site description type code Water Developed and disturbed INA Wayland silt loam Mesic glacio-lacustrine plain LAC Wayland-Teel-Palms association, 0 to 3 percent slopes, frequently flooded Subhydric alluvial deposits ALL Wegatchie, 0 to 3 percent slopes Hydric depressional wetland WET Whately fine sandy loam Mesic basic glacial till BAS Willette muck Hydric depressional wetland WET Williamson silt loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Wilpoint silty clay loam, 3 to 8 percent slopes Mesic glacio-lacustrine plain LAC Wilpoint silty clay loam, 8 to 15 percent slopes Mesic glacio-lacustrine plain LAC Windsor loamy fine sand, 0 to 8 percent slopes Xeric sand plain SND Windsor loamy fine sand, 8 to 15 percent slopes Xeric sand plain SND Wonsqueak and Onjebonge soils, ponded Hydric depressional wetland WET Wonsqueak, 0 to 3 percent slopes Hydric depressional wetland WET Wonsqueak-Adjidaumo association, 0 to 2 percent slopes, frequently ponded Hydric depressional wetland WET Wonsqueak-Onjebonge association, 0 to 3 percent slopes, frequently ponded Hydric depressional wetland WET

153

Table 1. Biophysical characteristics used to develop and evaluate ecological site types at Fort Drum, New York, USA.

Environmental variable (abbreviation) Range of values and units Data source K-means cluster analysis Mean soil available water capacity for top 100 cm (awc100cm) 67 - 267 mm SSURGO Mean soil sand content (pctsand)* 0 - 93 percent SSURGO Mean soil clay content (pctclay) 0 - 90 percent SSURGO Mean soil pH (soilph)* 4.5 - 7.2 SSURGO Elevation (elev)* 126 m - 280 m National Elevation Data (NED) Topographic exposure index (tei)* -22.4 - 33.8 Derived from NED

Slope gradient (pctslope) 0 - 83 percent Derived from NED Proximity to Lake Ontario (dist2lake) 22.8 km - 61.2 km Calculated in GIS Categorical variables included in recursive partitioning Soil drainage class (drcls) 1 very poor 2 poor 3 somewhat poor 4 somewhat well SSURGO 5 well 6 somewhat excessive 7 excessive Soil temperature regime (stmpcls) 0 mesic 1 frigid SSURGO

154

Table 2. Effect of projected soil moisture deficits on species establishment probabilities by ecological site type at Fort Drum, New York, USA. Species adaptation to drought is based on literature (Pastor and Post 1986; Burns and Honkala 1990; Gustafson and Sturtevant 2013) and species- site analyses.

Ecological site type Hydric Subhydric Mesic Mesic Mesic Xeric sand Mesic basic Species depressional alluvial lacustrine fluvial sand acidic plain glacial till wetland deposits plain terrace glacial till (n = 1493) (n = 627) (n = 167) (n = 85) (n = 1471) (n = 1624) (n = 1319) Abies balsamea + ------Acer rubrum + + + + + - - Acer saccharum + + - - - - - Amelanchier laevis + + - - - - - Betula alleghaniensis + + - - - - - Betula papyrifera + + - - - - - Betula populifolia + + + - - - - Carya cordiformis + + + + + - - Fagus grandifolia + + + - - - - Fraxinus americana + + + - - - - Fraxinus nigra + ------Ostrya virginiana + + - - - - - Pinus resinosa + + + + + + + Pinus strobus + + + + + + + Populus grandidentata + + - - - - - Populus tremuloides + + - - - - - Prunus serotina + + + + - - - Quercus alba + + + + + + + Quercus rubra + + + + + + + Thuja occidentalis + ------Tilia americana + + - - - - - Tsuga canadensis + + + - - - - Ulmus americana + + + - - - -

155

Table 3. Characteristics of 23 woody species and 2 non-forest cover types used in LANDIS-II simulations. Species Longevity Shade tolerance maxANPP1 maxAGB2 Species Common name code (yr) (1 = low; 5 = high) (g/m2/yr) (g/m2) Abies balsamea ABBA balsam fir 200 5 1000 26000 Acer rubrum ACRU red maple 200 4 1000 26000 Acer saccharum ACSA3 sugar maple 300 5 1000 26000 Alnus spp. ALNUS alder 40 1 1000 26000 Betula alleghaniensis BEAL yellow birch 300 3 1000 26000 Betula populifolia BEPO gray birch 150 2 1000 26000 Carya cordiformis CACO bitternut hickory 300 3 1000 26000 Fagus grandifolia FAGR American beech 350 5 1000 26000 Fraxinus americana FRAM2 white ash 300 3 1000 26000 Fraxinus nigra FRNI black ash 150 3 1000 26000 Ostrya virginiana OSVI hophornbeam 100 4 1000 26000 Pinus resinosa PIRE red pine 300 2 1000 26000 Pinus strobus PIST eastern white pine 400 3 1000 26000 Populus grandidentata POGR4 bigtooth aspen 110 1 1000 26000 Populus tremuloides POTR5 quaking aspen 110 1 1000 26000 Prunus serotina PRSE2 black cherry 200 2 1000 26000 Quercus alba QUAL white oak 400 3 1000 26000 Quercus rubra QURU northern red oak 250 3 1000 26000 Salix nigra SANI black willow 40 1 1000 26000 Thuja occidentalis THOC2 northern white cedar 400 3 1000 26000 Tilia americana TIAM American basswood 250 4 1000 26000 Tsuga canadensis TSCA eastern hemlock 450 5 1000 26000 Ulmus americana ULAM American elm 200 3 1000 26000 Mesic grass-forbs MM Mesic meadow 40 1 500 10000 Xeric grass-forbs XM Xeric meadow 40 1 500 10000

1 Maximum annual net primary productivity for species and cover types. 2 Maximum achievable aboveground biomass (AGB) for species and cover types.

156

Table 4. Multivariate correlation of environmental variables used to guide classification of ecological site types at Fort Drum, New York, USA. All correlation statistics were significant except where noted by “ns” (Spearman’s U, D = 0.05, p < 0.0001).

Environmental variable awc100cm pctsand pctclay soilph elev tei dist2lake pctslope

Soil available water capacity (awc100cm) 1.00 -0.70 0.75 0.33 -0.17 -0.06 ns ns Soil sand content (pctsand) -0.70 1.00 -0.94 -0.44 0.33 ns 0.07 ns Soil clay content (pctclay) 0.75 -0.94 1.00 0.38 -0.31 ns -0.05 0.02 Soil pH (soilph) 0.33 -0.44 0.38 1.00 -0.62 -0.10 -0.55 -0.31 Elevation (elev) -0.17 0.33 -0.31 -0.62 1.00 0.28 0.81 0.16 Topographic exposure index (tei) -0.06 ns ns -0.10 0.28 1.00 0.16 ns Distance to Lake Ontario (dist2lake) ns 0.07 -0.05 -0.55 0.81 0.16 1.00 0.15 Slope gradient (pctslope) ns ns 0.02 -0.31 0.16 ns 0.15 1.00

157

Table 5. Indicator values based on relative frequency of 23 tree species at Fort Drum, New York, USA. Maximum indicator values (IV) are statistically different than expected by chance except where noted * (p < 0.05, Monte Carlo randomization, n = 1000). Indicator values shown in bold font are more strongly associated with respective site types. See text for descriptions of ecological site types.

Ecological site type 1-WET 2-ALL 3-LAC 4-TER 5-TUP 6-SND 7-BAS Species Avg Max Max Hydric Subhydric Mesic Mesic Mesic, Xeric Mesic, depressional alluvial glacio- glacio-fluvial acidic sand basic IV IV Grp wetland deposits lacustrine plain sand terrace glacial till plain glacial till (n = 167) (n = 85) (n = 1471) (n = 1624) (n = 1319) (n = 1493) (n = 627) Abies balsamea 3 21 1 21 0 0 1 0 1 1 Acer rubrum 56 77 4 75 61 41 77 64 49 25 Acer saccharum 25 48 5 13 24 18 14 48 14 44 Amelanchier laevis 4 13 5 1 4 3 2 13 1 2 Betula alleghaniensis 13 35 1 35 7 4 13 18 7 5 Betula papyrifera 3 9 5 5 0 1 2 9 1 0 Betula populifolia 13 23 3 10 11 23 22 6 15 8 Carya cordiformis 2 12 7 1 1 2 0 1 0 12 Fagus grandifolia 7 17 5 4 5 4 9 17 9 4 Fraxinus americana 16 29 5 14 14 20 10 29 5 21 Fraxinus nigra 7 24 1 24 6 5 5 3 1 3 Ostrya virginiana 5 12 5 1 4 4 1 12 3 9 Pinus resinosa 2 10 6 1 0 1 3 1 10 0 Pinus strobus 33 54 6 41 21 19 46 27 54 22 Populus grandidentata* 6 11* 1 11 1 3 5 10 8 4 Populus tremuloides 23 35 3 17 20 35 26 23 21 20 Prunus serotina 45 66 2 28 66 39 51 44 44 45 Quercus alba 4 12 6 1 1 5 3 1 12 3 Quercus rubra 9 29 6 4 8 5 11 4 29 2 Thuja occidentalis 6 32 1 32 1 1 2 1 1 3 Tilia americana 8 14 7 4 9 9 4 13 3 14 Tsuga canadensis 14 28 4 17 15 9 28 7 17 7 Ulmus americana 21 38 3,7 11 36 38 11 7 5 38 158

Table 6. Soil water deficit (mm) for seven ecological site types at Fort Drum, New York, USA for mid- and late-century climate conditions. Deficits were derived using the WebWIMP water balance modeling tool (Wilmott et al. 1985, http://climate.geog.udel.edu/~wimp/, accessed 4/8/2018). Changes in mean annual temperature (o C) and mean annual precipitation (mm) for Representative Concentration Pathway (RCP) 4.5 and 8.5 are based on ensemble projections from the Community Climate System Model 3.0 (CCSM-3, NCAR GIS Program 2012). See text for descriptions of ecological site types. RCP 4.5 RCP 4.5 RCP 8.5 RCP 8.5 Ecological site Soil available water Base year 2055 2095 2055 2095 classification capacity 2015 +2.1oC; +19mm +2.5oC; +46mm +2.8oC; +55mm +4.7oC; +74mm (mm / 100 cm) (mm) SND 67 -17 -37 -26 -38 -57 BAS 76 -10 -24 -15 -24 -38 TUP 77 -10 -24 -15 -24 -38 TER 86 -5 -15 -9 -15 -27 ALL 120 -2 -9 -5 -9 -18 LAC 159 0 -6 -2 -5 -12 WET 267 0 0 0 0 0

159

Table 7. Above ground biomass projections for 23 tree species at Fort Drum, New York, USA under a high emissions (RCP 8.5) climate change scenario.

Biomass Biomass Change from Biomass Change from Species 2015 2055 2015 2115 2015 RCP 8.5 RCP 8.5 (g m2) (g m2) (percent) (g m2) (percent) Abies balsamea 121 181 50 419 246 Acer rubrum 12,735 17,016 34 13,539 6 Acer saccharum 13,216 15,715 19 23,537 78 Betula alleghaniensis 1,511 2,658 76 4,577 203 Betula populifolia 3,266 4,850 49 1,592 -51 Carya cordiformis 1,883 2,312 23 4,128 119 Fagus grandifolia 7,188 6,461 -10 4,597 -36 Fraxinus americana 4,197 3,479 -17 881 -79 Fraxinus nigra 121 152 26 69 -43 Ostrya virginiana 537 281 -48 358 -33 Pinus resinosa 2,375 1,805 -24 641 -73 Pinus strobus 9,047 10,099 12 10,933 21 Populus deltoides 5,017 11,438 128 2,167 -57 Populus grandidentata 1,308 1,903 46 696 -47 Populus tremuloides 9,070 14,626 61 3,226 -64 Prunus serotina 7,025 6,254 -11 1,372 -80 Quercus alba 872 1,289 48 2,210 153 Quercus rubra 3,977 4,770 20 4,306 8 Salix nigra 4,907 11,731 139 9,100 85 Thuja occidentalis 2,274 2,242 -1 2,080 -9 Tilia americana 1,817 2,162 19 3,894 114 Tsuga canadensis 11,410 12,639 11 15,487 36 Ulmus americana 6,782 12,004 77 20,319 200

160

Figure 1. Study site location (inset) and location of military infrastructure (training areas shown in green outline) at Fort Drum, New York, USA.

161

Figure 2. Historical mean annual temperature (oC), precipitation (mm) and Palmer Drought Severity Index (PDSI, Palmer 1965) for Watertown, New York, USA (Sources: NOAA National Center for Environmental Information, Global Summary of the Year data, https://www.ncdc.noaa.gov/cdo-web/; U.S. Drought Risk Atlas, National Drought Mitigation Center, http://droughtatlas.unl.edu/). Fitted dashed lines illustrate trends over time. 162162 b) Mean values for clusters. a) Cluster n pctsand soilph elev tei (%) (m)

1 Q 808 41.4 6.64a 169.9 -5.2a 2 Q 1309 84.0 5.50 192.2 -4.5a 3 Q 1246 56.8 4.85 239.7 2.8b,c 4 Q 1709 47.9 6.67a 181.9 1.5***c 5 Q 1669 86.7 5.60 211.2 1.9b

Figure 3. K-means cluster analysis (n = 6,741) of four environmental variables at Ft. Drum, New York, USA. Circle sizes in the principal components plot (a) are proportional to the number of sample units in each cluster and color-shaded regions contain 90% of the values in respective clusters. Cluster means were all significantly different (p = 0.05, nonparametric multiple comparisons, Dunn 1964) except where noted by superscripts (b). The map at right (c) illustrates the spatial distribution of each cluster on the installation.

163 Æ WET – hydric organic and alluvial deposits (6%) Æ ALL – subhydric alluvial stream terrace (2%) Æ TUP – mesic, frigid, loamy acidic glacial till (27%) Æ LAC – mesic, fine glacio-lacustrine deposits (28%) Æ TER – mesic, coarse loamy glacio-fluvial terrace (10%) Æ SND – xeric, coarse glacial outwash (16%) Æ BAS – mesic, loamy, basic glacial till (10%)

Æ Water (1%)

● 280m

● 126m ● 208m

Figure 4. Ecological site types at Fort Drum, New York, USA derived from 1:12,000 county soil series (Soil Survey Geographic (SSURGO) Database for Jefferson and Lewis counties, New York; available at https://sdmdataaccess.sc.egov.usda.gov. Accessed 04/28/2013). Areal proportions for each type shown in parentheses. See text for complete descriptions of ecological site types.

164 Variable Splits G^2 Proportion explained c) Spatial pctsand 2 6843.9 0.436 distribution of drcls 3 6741.9 0.429 sample points elev 2 1674.7 0.107 color-coded by soilph 2 453.0 0.029 groups in the classification a) Variation explained by environmental variables. tree (d). Map coordinates are in Actual Predicted by model (number of plots) UTM, zone 18N, Ecological meters. ALL BAS LAC SND TER TUP WET site type ALL 0 3 3 1 1 61 0 BAS 0 372 58 20 4 47 2 LAC 0 7 1145 0 1 24 0 SND 0 2 1 1153 13 8 4 TER 0 0 6 11 1285 9 2 TUP 0 0 0 1 4 1046 0 drcls WET 0 0 6 0 2 3 111

Total 0 384 1219 1186 1310 1198 119 drcls drcls 5,6,7 % correct 0 96.9 93.9 97.2 98.1 87.3 93.3 1,2,3,4 b) Confusion matrix for the training dataset color-coded < 53 > 53 < 85 > 85 to map (c) and classification tree (d). pctsand pctsand pctsand pctsand < 216m > 216m < 207m > 207m drcls 1 drcls 2,3,4 Figure 5. Recursive partitioning of seven ecological site elev elev elev elev types based on four environmental variables at Fort < 6.9 drcls 1,2 < 7 soilph Drum, New York, USA. The classification tree model soilph explained r2 = 0.86 (n = 5416; RMSE = 0.22) of the > 6.9 variation with a misclassification rate of 5.6 percent. drcls 3,4 > 7 soilph soilph Twenty percent of the observations were withheld for validation (r2 = 0.859; n = 1324; RMSE = 0.23). d) Classification tree color-coded to map (c) showing split values for variables. 165 Figure 6. Change in species establishment probability (SEP) for Acer saccharum at Fort Drum, New York, USA over time by ecological site type and emissions scenario. The Representative Concentration Pathway (RCP) 4.5 scenario assumes an increase in mean annual temperature in the early part of the century (+ 2.1o C) that is maintained through the end of the century by conservation measures. The RCP 8.5 scenario assumes continued production of greenhouse gases at current rates and increasing mean annual temperatures to the end of the century (+4.7o C). Climate projections are based on the Community Climate System Model 3.0 (National Center for Atmospheric Research (NCAR) GIS Program. 2012). See text for descriptions of ecological site types. 166 Figure 7. Predicted above ground biomass (AGB) for all species and cover types by ecological site type at Fort Drum, New York, USA under current climate conditions (base model, top) and the high emissions RCP 8.5 scenario (bottom). Results for the RCP 4.5 scenario were intermediate and not shown for clarity of presentation. See text for descriptions of ecological site types.

167

Figure 8. Total above ground biomass (AGB) of five tree species with different shade and drought tolerances shown for two ecological site types at Fort Drum, New York, USA. The SND site type had an estimated soil water deficit approximately twice that of the TER site type at year 2095 under the high emissions RCP 8.5 scenario. See text for descriptions of ecological site types. 168

Figure 9. Total projected above ground biomass (AGB) for 16 tree species at Fort Drum, New York, USA under a high emissions (RCP 8.5) climate change scenario based on an ensemble Community Climate System Model 3.0 (NCAR GIS Program 2012). Solid lines represent relatively shade tolerant species and dashed lines represent relatively shade intolerant species. Biomass projections based on the lower emissions RCP 4.5 scenario were only slightly higher than those under the RCP 8.5 scenario and are not shown for clarity of presentation. 169 2015 2055 2095

Figure 10. Projected change in community type over time under a high emissions (RCP 8.5) climate change scenario at Fort Drum, New York, USA. Area shown in dark gray on map was not classified into a forest type due to the lack of forest inventory data. 170 2015 2095 AWC 150 CLAY LOAM AWC 150 CLAY AWC 100LOAM SANDY

Æ Surplus Æ Moderate surplus Æ Moderate deficit Æ Deficit AWC 50 SAND

Figure 11. Monthly water balance (mm) for three soils at Ft. Drum, New York, USA for years 2015 and 2095. Available water holding capacity (AWC) is derived from SSURGO data for the top 100 cm of soil (Soil Survey Staff 2013).Water balances are based on projected mean annual temperature and mean annual precipitation (CCSM-3 ensemble RCP 8.5 scenario, NCAR GIS Program 2012) and calculated using the WebWIMP water balance model (Wilmott et al. 1985). 171

172