Late-Successional Biomass Development in Northern Hardwood-Conifer Forests of the Northeastern United States

Abstract

Managing the contribution of forest ecosystems to global carbon cycles requires accurate predictions of biomass dynamics in relation to stand development. Our study evaluated competing hypotheses regarding late-successional biomass dynamics in northern hardwood-conifer forests using a data set spanning the northeastern United States, including 48 mature and 46 old-growth stands. Continuous data on dominant tree ages were available for 29 of these and were used as an indicator of stand development. Aboveground live biomass was significantly (P < 0.001) different between mature (195 Mg/ha) and old-growth (266 Mg/ha) sites. Aboveground biomass was positively (P < 0.001) and logarithmically correlated with dominant tree age; this held for live trees (r2 = 0.52), standing dead trees (r2 = 0.36), total trees (r2 = 0.63), and downed woody debris (r2 = 0.24). In a Classification and Regression Tree analysis, stand age class was the strongest predictor of biomass, but ecoregion and percent conifer accounted for ∼25-33% of intraregional variability. Biomass approached maximum values in stands with dominant tree ages of ∼350-400 years. Our results support the hypothesis that aboveground biomass can accumulate very late into succession in northern hardwood-conifer forests, recognizing that early declines are also possible in secondary forests as reported previously. Empirical studies suggest a high degree of variability in biomass development pathways and these may differ from theoretical predictions. Primary forest systems, especially those prone to partial disturbances, may have different biomass dynamics compared with those of secondary forests. These differences have important implications for both the quantity and temporal dynamics of carbon storage in old-growth and recovering secondary forests.

Full text

Late-Successional Biomass Development in Northern Hardwood-Conifer
Forests of the Northeastern United States
William S. Keeton, Andrew A. Whitman, Gregory C. McGee, and Christine L. Goodale
Abstract: Managing the contribution of forest ecosystems to global carbon cycles requires accurate predictions
of biomass dynamics in relation to stand development. Our study evaluated competing hypotheses regarding
late-successional biomass dynamics in northern hardwood-conifer forests using a data set spanning the north-
eastern United States, including 48 mature and 46 old-growth stands. Continuous data on dominant tree ages
were available for 29 of these and were used as an indicator of stand development. Aboveground live biomass
was significantly (P⬍0.001) different between mature (195 Mg/ha) and old-growth (266 Mg/ha) sites.
Aboveground biomass was positively (P⬍0.001) and logarithmically correlated with dominant tree age; this
held for live trees (r
2
⫽0.52), standing dead trees (r
2
⫽0.36), total trees (r
2
⫽0.63), and downed woody debris
(r
2
⫽0.24). In a Classification and Regression Tree analysis, stand age class was the strongest predictor of
biomass, but ecoregion and percent conifer accounted for ⬃25–33% of intraregional variability. Biomass
approached maximum values in stands with dominant tree ages of ⬃350 – 400 years. Our results support the
hypothesis that aboveground biomass can accumulate very late into succession in northern hardwood-conifer
forests, recognizing that early declines are also possible in secondary forests as reported previously. Empirical
studies suggest a high degree of variability in biomass development pathways and these may differ from
theoretical predictions. Primary forest systems, especially those prone to partial disturbances, may have different
biomass dynamics compared with those of secondary forests. These differences have important implications for
both the quantity and temporal dynamics of carbon storage in old-growth and recovering secondary forests. FOR.
SCI. 57(6):489 –505.
Keywords: aboveground biomass, northern hardwoods, stand development, carbon cycles, old-growth
MANAGING THE CONTRIBUTION OF FOREST ECO-
SYSTEMS to global carbon budgets requires ac-
curate predictions of biomass dynamics in rela-
tion to stand development and management. Accurate
predictions have become particularly pertinent with the
development of domestic and international carbon markets
and proposals for forest sector participation in both volun-
tary and compliance systems (Ruddell et al. 2007, Ray et al.
2009b). There remain critical issues regarding the long-term
net effects of different forest management approaches, such
as intensive versus less intensive management, on carbon
sequestration and storage (Harmon et al. 1990, Thornley
and Cannell 2000, Harmon and Marks 2002). Consideration
of alternate management systems often entails a projection
of long-term in situ forest biomass accumulation under a
“no management” scenario, and comparison of this trajec-
tory with carbon storage in multiple sinks (e.g., forest, wood
products, or landfills) under active harvest scenarios (Har-
mon et al. 2009, Nunery and Keeton 2010). Estimating the
carbon storage capacity of recovering secondary forests,
such as those now dominant in the northeastern United
States, similarly necessitates estimation of biomass dynam-
ics in relation to forest stand development processes (Brown
et al. 1999). Consequently, improving our understanding of
carbon storage dynamics in primary (i.e., never cleared),
old-growth, and unmanaged forests has become increas-
ingly important for characterizing these baseline, reference,
and future potential conditions (Hudiburg et al. 2009, Keith
et al. 2009, Rhemtulla et al. 2009).
Forest Biomass and Stand Development
For decades, models relating biomass development to
forest age for northern hardwoods have been based on
ground-breaking research performed in the 1970s at the
Hubbard Brook Experimental Forest in New Hampshire,
USA. This widely cited model, developed by Bormann and
Likens (1979), predicts peaks in biomass after less than two
centuries (approximately 170 years) of stand development,
followed by declining biomass in stands 200 –350 years of
age, and “steady-state” biomass dynamics in stands ⬎350
years of age. Bormann and Likens broke this timeline into
four phases of development, which they called “reorgani-
zation,” “aggregation,” “transition,” and “steady state.” Net
ecosystem productivity (NEP) was predicted to decline to
zero “at full maturity” (Odum 1969, Whittaker et al. 1974).
Biomass accumulation curves for the first two development
stages were based on empirical data collected in secondary
William S. Keeton, University of Vermont, Rubenstein School of Environment and Natural Resources, 343 Aiken Center, Burlington, VT 05405—Phone:
(802) 656-2518; Fax: (802) 656-2623; william.keeton@uvm.edu. Andrew A. Whitman, Manomet Center for Conservation Sciences—
awhitman@manomet.org. Gregory C. McGee, State University of New York— ggmcgee@esf.edu. Christine L. Goodale, Cornell University—
clg33@cornell.edu.
Acknowledgments: This research was supported in part by grants from the Northeastern States Research Cooperative, the US Forest Service McIntire-Stennis
Forest Research Program, the University of Maine Cooperative Forestry Research Unit, the Merck Family Fund, the Pierce Charitable Trust, the National
Fish and Wildlife Foundation, the Laird Norton Endowment Fund, and the Maine Outdoor Heritage Fund. Sabina Burrascano and three anonymous reviewers
provided helpful reviews of the manuscript.
Manuscript received July 31, 2009, accepted January 18, 2011 Copyright © 2011 by the Society of American Foresters
Forest Science 57(6) 2011 489

stands located at Hubbard Brook and elsewhere. Biomass
dynamics projected for the latter stages of development
were largely theoretical, although grounded in observational
studies and derived from predictions made by the JABOWA
model (Botkin et al. 1972), one of the early gap-based forest
succession simulators.
Bormann and Likens (1979) cited a lack of old-growth
forests in the northeastern United States as the reason for
relying on theoretical projections for late-successional bio-
mass development models. But since the 1970s researchers
have discovered and mapped more old-growth in the North-
east than was previously known to exist (McMartin 1994,
Davis 1996, D’Amato et al. 2006). For instance, by some
estimates there are more than 80,000 ha of old-growth forest
in the Adirondack region of New York alone, one of the
largest concentrations east of the Mississippi River (Dun-
widdie et al. 1996). These forests span a wide range of sites
and biophysical diversity and thus are not anomalous (e.g.,
restricted to inaccessible or low-productivity sites) but are
in fact representative of landscape-scale ecological variabil-
ity (McMartin 1994). More than 30 years of research has
described the structural characteristics (e.g., Tyrrell et al.
1998, Hale et al. 1999, McGee et al. 1999, Ziegler 2000,
Keeton et al. 2007, Whitman and Hagan 2007), successional
dynamics (e.g., Foster 1988, Abrams and Orwig 1996,
Goodburn and Lorimer 1999, McLachlan et al. 2000), and
disturbance regimes (e.g., Runkle 1982, Ziegler 2002,
Lorimer and White 2003, D’Amato and Orwig 2008) of
late-successional/old-growth northern hardwood-conifer
forests. These studies allow us to revise our understanding
of late-successional forest dynamics using real-world refer-
ence stands and empirical rather than theoretical data.
Competing Hypotheses Describing Biomass
Dynamics in Relation to Forest Age
Studies conducted in primary (never cleared, but con-
taining a mix of successional stages) and old-growth (a
stand development condition) forests have provided differ-
ent and sometimes conflicting perspectives on late-succes-
sional biomass dynamics. For example, research in the US
Upper Midwest (Tyrrell and Crow 1994) showed basal area
peaks somewhat later than predicted by Bormann and Lik-
ens (1979), achieving maximum values in 230- to 260-year-
old forests, but with subsequent declines in older forests. In
contrast with Tyrrell and Crow (1994), Ziegler (2000)
showed continued basal area increases in a chronosequence
spanning forest ages in excess of 400 years, with little
evidence of asymptotic relationships or declines as age
increased. There is also growing recognition of the potential
for continued net positive carbon uptake (NEP) very late
into stand development in temperate forests globally (Pre-
gitzer and Euskirchen 2004, Luyssaert et al. 2008, Keith et
al. 2009), although this has been poorly explored in northern
hardwoods specifically. Taken together, this body of liter-
ature supports a hypothesis that there may be potential for
biomass accumulation in northern hardwoods of both
greater magnitude and duration than previously recognized.
A competing hypothesis is offered by recent work at
Hubbard Brook, based on long-term plot remeasurements in
the unmanipulated (or control) watershed (W6). These have
shown biomass accumulations slowing much earlier than
expected, leveling off after only 80 years of stand develop-
ment (Fahey et al. 2005, Siccama et al. 2007). Live tree
biomass actually decreased over the most recent remeasure-
ment period (1997–2002); total biomass (including standing
dead trees) continued to increase. Fahey et al. (2005) sug-
gested a number of possible explanations for the trends they
detected, including growth reductions due to acid deposi-
tion, declines caused by beech bark disease (Nectria spp.),
altered stand development processes related to land use
history, and the possibility of climate change effects. How-
ever, they could not reject the possibility that biomass
accumulation potential in secondary forests may be lower
than previously predicted. Remaining unexplored is
whether the biomass dynamics observed in primary forests
reflect fundamentally different stand development processes
than those in secondary forests.
Biomass curves based on the Bormann and Likens
(1979) model continue in use as the basis for regional
carbon budget assessments and methodologies for estimat-
ing future forest carbon storage potential (e.g., Turner et al.
1995, Smith et al. 2006). The curves also underpin algo-
rithms used in individual tree-based forest growth and de-
velopment models (Dixon 2002, Ray et al. 2009a), which
are increasingly relied on to predict carbon storage under
alternate forest management strategies (Nunery and Keeton
2010). Because of their wide use and importance to carbon
budget estimation, ensuring their accuracy has taken on new
significance. In this study, we examine whether generalized
biomass curves are sufficiently representative or predictive
of biomass dynamics playing out across the northern hard-
wood region. We evaluate the competing hypotheses de-
scribed above and include our own empirical evidence,
exploring biomass relationships in a data set spanning most
of northern New York and New England.
Determination of Tree and Stand Ages
In this study we examine stand age-related trends for a
subset of our data points. However, assigning an “age” to
primary northern hardwood forests can be problematic. Un-
like western temperate coniferous forests, in which stand
origin often can be traced to a single high severity or
“stand-replacing” fire event (Keeton and Franklin 2004,
2005), return intervals for stand-replacing disturbances,
such as hurricanes, in northern hardwoods often exceed
1,000 years (Seymour et al. 2002, Lorimer and White 2003).
These are more than double the maximum lifespans of the
constituent late-successional tree species. Structure and age
distributions in primary temperate hardwoods are shaped,
rather, over centuries by high- to moderate-frequency par-
tial disturbances, such as gap-forming events, that result in
fine- to intermediate-scaled regeneration and release effects
(Hanson and Lorimer 2007, D’Amato and Orwig 2008).
Methods for assigning “stand age” in primary northern
hardwood systems typically require an estimation of the age
of canopy trees, either as an average for all codominant and
dominant trees over a given minimum diameter (e.g.,
Woods and Cogbill 1994) or as an average of the largest,
490 Forest Science 57(6) 2011

dominant trees (e.g., Keeton et al. 2007, Stovall et al. 2009).
For all of these methods, the result is an estimate of canopy
tree age, rather than time since high-intensity disturbance.
Determination of tree ages in old-growth stands is challeng-
ing because of the prevalence of rotten heartwood. Repeated
coring is usually required to obtain a single intact core.
However, coring the largest trees increases the precision of
stand age estimates from relatively small samples because
of the close correspondence between age and diameter for
the larger trees in uneven-aged stands (Leak 1985). In this
study we determined the average ages of the largest trees,
and thus our ages are weighted toward the maximums
encountered in the sample trees. We use the term “dominant
tree age,” recognizing that our values are not true estimates
of stand age but rather more probably are an indicator of
variation in disturbance history, large tree survivorship, and
related degree of stand structural development. We propose
that this is as a useful analytical approach for comparing old
forest sites and assessing development potential and related
ecological processes (see, e.g., Stovall et al. 2009, Warren et
al. 2009).
Because of the difficulties associated with age determi-
nation, many forest inventory data sets have assigned sites
or plots to age classes (e.g., mature versus old-growth).
Previous studies (e.g., McGee et al. 1999, Goodale and Aber
2001), consequently, often have used categorical compari-
sons, rather than modeling age relationships using continu-
ous variables. In this study, we use a combination of both
methods depending on the available age data. Using mature
stands as a benchmark for comparison against old-growth
helps contrast the conditions encountered in different age
classes. These contrasts probably would be even more dis-
tinct if old-growth structure were compared with young and
partially harvested forest stands (McGee et al. 1999, Crow
et al. 2002, Angers et al. 2005).
Methods
Overview of Data Sets and Study Region
Our study combined data from four existing data sets,
hereafter referred to as the “Goodale,” ”Keeton,” “McGee,”
and “Whitman” data sets. The data set includes 94 indepen-
dent data points, 46 old-growth and 48 mature sites (Table
1). It thus provides a robust sampling of the remaining
old-growth forests in the Northeast. For the purposes of our
study, we defined “old-growth” as primary (never cleared)
forest with dominant canopy trees ⬎150 years of age and
uneven-aged (more than three age classes) structure. Our
“mature” forest designation refers to stands that are approx-
imately 80 –150 years of age, exhibiting even to multiaged
(two or three age classes) structure. The mature stands
originated after logging and human-caused wildfires in the
late 19th and early 20th centuries and have had little or no
logging since establishment, which minimizes variability
associated with management history. Our sites included
both northern hardwood-dominated as well as mixed-wood
stands (see Table 1 for data on conifer percentages by site),
with varying components of Betula alleghaniensis (yellow
birch), Fagus grandifolia (American beech), Acer saccha-
rum (sugar maple), Acer rubrum (red maple), Picea rubens
(red spruce), Tsuga canadensis (eastern hemlock), and
Abies balsamea (balsam fir). There were scattered Pinus
strobus (eastern white pine), including remnant old-growth
trees, at some sites.
Collectively our data set provides a well-distributed,
representative sample of the northern hardwood region of
the northeastern United States. The data focus on three
subregions within this area, the Adirondack Mountain re-
gion of upstate New York, the White Mountains of New
Hampshire, and the west-central highlands to northern hills
of Maine. For clarity throughout the article we refer to these
respectively as “New York” (Keeton and McGee data sets),
“New Hampshire” (Goodale and Whitman data sets), and
“Maine” (Whitman data set). These subregions share a
moist temperate climate characterized by cold winters and
warm summers, with even distribution of precipitation
throughout the year. Surficial geology within our study
areas is dominated by postglacial plateaus, hills, and mon-
tane landforms with relatively fertile soils derived primarily
from glacial and alluvial tills and deposits.
Data Collection
Site selection and sampling methods differed for each of
the four data sets, and thus are described separately. Sample
sizes by state and biomass variable are provided in Table 2.
Keeton Data Set
The Keeton data set includes data from 29 sites, 19 of
which were described in Keeton et al. (2007); 10 sites were
sampled subsequently. These sites were selected from the
available known occurrences of old-growth and comparable
secondary forests within the Adirondack region based on
site-matching criteria described in Keeton et al. (2007). The
sites, sampled in 2002–2006, are located along 200- to
300-m-long, first- and second-order stream reaches. Eleva-
tions ranged from 450 to 600 m. At each site forest structure
and composition (live and standing dead trees ⬎5 cm dbh)
were sampled within 6 –10 (proportionate to reach length)
variable radius (2.3 metric basal area factor) prism plots
randomly placed within 30 m of the stream bank (well
distributed, even ratio per side). Tree heights were measured
in alternating plots with an Impulse 200 Laser Rangefinder;
snags were assigned to decay stages (1–9). Downed coarse
woody debris (DCWD) (downed logs ⱖ10 cm diameter at
intercept, ⱖ1 m length) volume by decay class (1–5) was
measured along multiple, systematically placed transects
using the line-intercept method and layout scheme de-
scribed in Keeton et al. (2007). Transects were 200- to
300-m-long, depending on reach length. Age at breast
height was determined from increment cores for four to six
randomly selected, large dominant B. alleghaniensis,P.
rubens,orT. canadensis at each site. Cored trees were
randomly selected from among the larger canopy domi-
nants. Ages were estimated in the field following McGee et
al. (1999), with one core per site randomly selected and
returned to the laboratory for examination, after sanding and
mounting, under a dissecting microscope to assess field
error (mean ⫽⫾9 years; SD ⫽⫾5).
Forest Science 57(6) 2011 491

McGee Data Set
McGee collected data from seven old-growth and four
mature study sites in upland Adirondack northern hardwood
old-growth forests, sampled in 1995–1997. Sites were lo-
cated on mid to upper slope positions at elevations ranging
from 470 to 770 m. Stand structural and compositional data
for live and standing dead trees ⬎10 cm dbh were collected
on two 0.1-ha (20 ⫻50 m) randomly established sample
plots at each site. Length and end diameters were measured
for downed log sections and stumps ⬎10-cm diameter
within nested subplots or across the whole 0.1-ha area
depending on log diameter (see McGee et al. 1999). Struc-
tural and compositional data for six of the seven old-growth
stands and the four maturing (postfire, 90- to 100-year-old)
stands are reported by McGee et al. (1999). Insufficient data
are available to estimate dominant tree age for the McGee
sites.
Whitman Data Set
Whitman collected data from 20 old-growth and 20 ma-
ture study sites in upland northern hardwood forests in
Table 1. Site information, including data on stand structure and composition.
Site
Data
set State Ecoregion
1
Age class
Percent
conifer
Aboveground
live tree
biomass
(Mg/ha)
Basal
area
(live)
(m
2
/ha)
Stem
density
(live)
(trees/ha)
Mountain Pond Whitman ME M212A Old-growth 0.00 305.09 35.10 467
Deadriver4 Whitman ME M212A Mature 29.20 390.58 51.70 683
Deadriver5 Whitman ME M212A Mature 41.75 209.67 32.00 600
Little Bigelow Whitman ME M212A Old-growth 5.92 225.19 31.83 600
Borestone Mtn. 1 Whitman ME M212A Old-growth 23.45 224.15 29.98 517
Borestone Mtn. 2 Whitman ME M212A Old-growth 0.00 226.26 26.99 417
Greatpond9 Whitman ME M212A Mature 0.00 205.32 29.74 583
Kibby1244 Whitman ME 212B Mature 3.51 191.44 24.86 333
Kibby1508 Whitman ME 212B Mature 11.90 202.27 26.42 333
Merrill1 Whitman ME 212B Mature 0.00 121.30 15.77 267
PiercePond1 Whitman ME M212A Mature 0.00 284.32 36.84 717
Skinner153 Whitman ME M212A Mature 0.00 189.19 25.90 550
Skinner267 Whitman ME 212A Mature 17.10 175.43 26.17 633
T11R15 WELS0 Whitman ME M212A Mature 12.38 339.51 44.85 933
T11R15 WELS1 Whitman ME M212A Old-growth 8.18 333.28 37.65 383
T11R15 WELS4 Whitman ME M212A Old-growth 52.35 245.97 37.70 633
Yankeetuladi Whitman ME M212A Old-growth 2.53 266.05 32.33 617
Yankeetuladi Whitman ME M212A Old-growth 0.00 163.69 22.34 433
T28MD27 Whitman ME M212A Mature 0.00 121.08 15.57 300
T29MD30 Whitman ME M212A Mature 12.97 224.59 33.26 917
T2R4BKPWKR Whitman ME M212A Mature 9.52 224.83 29.84 500
T3R8 WELS Whitman ME 212C Old-growth 3.06 232.90 31.20 617
Boody Brook Whitman ME 212C Old-growth 8.09 568.85 64.17 733
Big Reed 1 Whitman ME M212A Old-growth 2.10 370.67 43.48 583
Big Reed 2 Whitman ME M212A Old-growth 15.74 318.03 41.20 667
Big Reed 3 Whitman ME M212A Old-growth 35.12 410.30 54.57 717
Big Reed 4 Whitman ME M212A Old-growth 50.63 104.21 16.45 367
Big Reed 5 Whitman ME M212A Old-growth 64.17 158.81 27.26 533
Big Reed 6 Whitman ME M212A Old-growth 0.00 348.49 38.93 383
T8R16 WELS9 Whitman ME M212A Mature 0.00 317.33 39.30 583
T9R10 WELS13 Whitman ME M212A Old-growth 7.48 314.68 40.30 633
T9R8 WELS10 Whitman ME M212A Mature 56.18 215.04 34.11 700
T9R8 WELS14 Whitman ME M212A Mature 21.56 145.78 20.08 400
WELD5 Whitman ME M212A Mature 0.00 169.46 22.90 383
WYMAN130 Whitman ME M212A Mature 23.56 343.11 39.58 500
Bartlett26 Whitman NH M212A Old-growth 20.77 361.67 48.25 717
Bartlett40 Whitman NH M212A Old-growth 5.06 297.50 35.92 417
Carrabasset11 Whitman NH M212A Mature 13.70 221.55 63.60 1,350
The Bowl 1 Whitman NH M212A Old-growth 10.77 256.34 34.15 550
Lafayette Brook Goodale NH M212D Old-growth 7.46 287.57 33.85 363
Gibbs Brook Goodale NH M212D Old-growth 10.21 200.58 24.76 375
Glen Boulder Goodale NH M212D Old-growth 15.43 251.14 29.61 600
Spruce Brook Goodale NH M212D Old-growth 0.00 316.92 35.10 775
The Bowl 2 Goodale NH M212D Old-growth 0.00 250.25 30.46 588
Mt. Bickford Goodale NH M212D Mature 1.77 170.45 31.58 313
Zealand Valley Goodale NH M212D Mature 0.00 186.70 28.74 550
George’s Gorge Goodale NH M212D Mature 0.00 233.05 32.25 663
Wild River Goodale NH M212D Mature 0.00 204.51 29.37 1,213
(Continued)
492 Forest Science 57(6) 2011

Maine (Whitman and Hagan 2007). Sampling was con-
ducted in 2003 and 2004. Sites occurred on elevations
ranging from 150 to 700 m and on mesic slopes and hilltops.
Previous researchers had visited the old-growth sites and
established the lack of human and catastrophic disturbance
and the fact that tree ages exceeded 150 years (Maine
Critical Areas Program 1985). Whitman sampled all old-
growth northern hardwood stands in the study area that
could be relocated using existing reports. Mature stands
were randomly selected from stand maps for selected large
(⬎100,000 ha) landowners dispersed across central and
northern Maine. Live and standing dead trees (ⱖ10 cm dbh,
at a height of 1.37 m) were measured in a randomly estab-
lished 0.06 ha (3 ⫻200 m) plot at each site. DCWD (ⱖ10
cm intercept diameter, ⱖ1 m in length) was sampled using
the line intercept method (Van Wagner 1968) on a 200-m
transect. Tree age data were not collected at the Whitman
sites.
Goodale Data Set
Study site selection, location, and sampling design for
the Goodale data set are described in detail in Goodale and
Aber (2001). Five study areas in the White Mountains of
Table 1. Site information, including data on stand structure and composition (cont).
Site
Data
set State Ecoregion
1
Age class
Percent
conifer
Aboveground
live tree
biomass
(Mg/ha)
Basal
area
(live)
(m
2
/ha)
Stem
density
(live)
(trees/ha)
Mt. Chocorua Goodale NH M212D Mature 0.00 124.93 23.48 613
Cascade Brook Goodale NH M212D Mature 4.82 227.80 29.77 400
Mt. Tom Goodale NH M212D Mature 2.18 139.67 21.20 425
Lost Pond Goodale NH M212D Mature 0.00 188.96 29.12 613
Carter Dome Tr. Goodale NH M212D Mature 13.10 265.25 35.07 488
Mt. Paugus Goodale NH M212D Mature 10.30 180.38 32.59 875
Ampersand Mt. McGee NY M212D Old-growth 12.99 357.57 39.55 380
Gill Brook McGee NY M212D Old-growth 15.33 258.01 31.20 405
Whalestail Mt. McGee NY M212D Old-growth 3.24 356.50 38.25 345
Huntington 1 McGee NY M212D Old-growth 4.85 234.93 28.30 405
Sucker Brook McGee NY M212D Old-growth 3.26 267.56 31.90 470
Mason Lake McGee NY M212D Old-growth 4.88 328.56 36.70 370
Moose Mountain McGee NY M212D Old-growth 2.57 358.70 38.95 405
Bigsby Pond McGee NY M212D Mature 7.80 220.21 29.90 500
Van Hoevenberg McGee NY M212D Mature 0.00 198.00 24.55 390
Hennessy Mt. McGee NY M212D Mature 0.00 240.84 34.10 650
Gooseberry Mt. McGee NY M212D Mature 0.00 207.25 27.85 490
Constable Inlet Keeton NY M212D Mature 61.00 130.17 23.30 714
Darby Brook Keeton NY M212D Mature 19.00 145.62 22.60 555
Pigeon Lake Keeton NY M212D Mature 84.00 135.26 26.40 1,380
Witchopple 1 Keeton NY M212D Old-growth 34.00 165.99 26.80 827
Witchopple 2 Keeton NY M212D Mature 45.00 160.08 27.20 1041
Combs Brook Keeton NY M212D Mature 13.00 150.15 21.90 563
Clearlake Outlet Keeton NY M212D Old-growth 36.00 188.26 29.30 1,398
Clearlake 2 Keeton NY M212D Old-growth 29.00 201.04 31.80 1,627
Panther Keeton NY M212D Mature 29.00 178.27 27.30 848
Oxbow Keeton NY M212D Mature 34.00 126.62 21.60 951
Limekiln Keeton NY M212D Old-growth 33.00 241.52 37.50 1,348
Limekiln Trail Keeton NY M212D Old-growth 14.00 176.86 27.10 1,062
Little Moose Trib 1 Keeton NY M212D Old-growth 44.00 195.66 29.30 609
Otter Brook Keeton NY M212D Mature 32.00 190.52 31.40 733
Little Moose Outlet Keeton NY M212A Old-growth 43.00 232.17 35.30 710
Little Moose Trib 6 Keeton NY M212A Old-growth 70.00 341.76 51.40 705
Little Moose Trib 5 Keeton NY M212A Mature 68.00 191.19 32.50 748
Upper Sylvan Keeton NY M212A Old-growth 59.00 214.97 33.30 1,010
Little Moose Trib 3 Keeton NY M212A Mature 88.00 89.81 19.10 820
Dutton Brook Keeton NY M212A Old-growth 69.00 262.03 37.90 602
Beth’s Brook Keeton NY M212A Mature 18.00 138.60 21.40 497
Camp Nine Keeton NY M212A Mature 57.00 127.63 24.10 661
Arbutus Outlet Keeton NY M212A Mature 3.00 195.54 29.50 731
Wolf Lake Outlet Keeton NY M212A Mature 45.00 171.06 32.10 2,572
Huntington 2 Keeton NY M212A Old-growth 27.00 193.66 26.30 472
Huntington 3 Keeton NY M212A Old-growth 68.00 206.81 32.10 406
Pico Keeton NY M212A Mature 7.00 147.46 23.00 1,222
McKenna Keeton NY M212A Old-growth 50.00 212.70 31.10 843
MellonBerry Keeton NY M212A Old-growth 75.00 217.35 33.50 792
1
M212 ⫽New England-Adirondack Province: M212A, White Mountains section; M212D, Adirondack Highlands section; 212 ⫽Laurentian Mixed Forest
Province: 212A, Aroostook Hills and Lowlands section; 212B, Maine-New Brunswick Foothills and Lowlands section; 212C, Fundy Coastal and Interior
Section
Forest Science 57(6) 2011 493

New Hampshire were identified from historical maps, each
containing sites originating from fires or logging (the ma-
ture sites) or in old-growth. There was one location at each
study area for each of the three site histories, for a total of
15 sites (10 mature and 5 old-growth). Elevations ranged
from 600 to 700 m; the sites were sampled in 1996. At each
site, two randomly placed, 20 ⫻20-m (0.04-ha) plots were
established on drained, midslope topographic positions. dbh
and species were recorded for all live trees ⬎9.5 cm dbh and
1.37 m in height. The data set does not include data for
standing dead trees, DCWD, or tree ages, and thus sample
sizes in the regional data set are lower for those variables.
Data Analysis
Our analysis consisted of four elements designed to
evaluate the competing hypotheses regarding late-succes-
sional biomass development in northern hardwoods. These
were 1) processing and standardization of field data to
generate a regional-scale set of biomass estimates and stand
structure metrics, 2) comparisons of age classes to provide
a regional perspective based on the available categorical
data (see Table 2 for sample sizes), 3) regression modeling
of biomass in relation to dominant tree age for the subset of
sites (n⫽29) from which continuous age data were avail-
able, and 4) a multivariate analysis to test the sensitivity of
biomass estimates to multiple sources of variability (n⫽
94). These elements are described sequentially below. All
statistical analyses were run in TIBCO Spotfire S⫹8.1.
Tests were significant at
␣
⫽0.05.
Field data from the four data sets were processed in
either Microsoft Excel (Goodale, McGee, and Whitman
data sets) or the Northeast Ecosystem Management Deci-
sion Model (Keeton data set) (Twery et al. 2005) to generate
a suite of stand structure biometrics. All tree-related metrics
were standardized to a ⬎10-cm dbh minimum size thresh-
old. Aboveground tree biomass estimates were calculated
from dbh measurements using species-specific allometric
equations. For the Keeton, McGee, and Whitman data sets,
the equations followed Jenkins et al. (2003). The Goodale
data were processed using equations from Whittaker et al.
(1974) with modifications by Siccama et al. (1994) for
yellow birch, sugar maple, American beech, and red spruce;
equations from Hocker and Early (1983) were used for
paper birch (Betula papyrifera), red maple, aspen (Populus
tremuloides), and eastern hemlock.
DCWD volume estimates were calculated differently for
line-intercept versus fixed-area plot surveys. The former
relied on equations from Warren and Olsen (1964) as mod-
ified by Shivers and Borders (1996), whereas the latter used
geometric equations described in McGee et al. (1999). Our
procedure for converting DCWD volumes to biomass in-
volved two steps. First, we conducted an analysis using the
subset of sites (New York) for which decay class data were
available. Decay class-specific conversion factors (i.e., spe-
cific gravity values) from Harmon et al. (2008) were applied
to the volume data sorted by age class. Then we calculated
an average conversion factor that accurately estimated bio-
mass for each age class, given the differences in decay class
distributions (see Results below). These values were 0.325
and 0.285 g/cm
3
for mature and old-growth, respectively,
and were consistent with conversion factors generated from
bulk density measurements of DCWD in the Adirondacks
(McGee et al. 1999). In the second step, volume data for all
the sites were converted to biomass using these age class-
specific values.
Table 2. Results of Tukey tests comparing biomass characteristics in mature (80- to 150-year-old, even to multiaged) versus
old-growth (>150 years old, uneven-aged) northern hardwood-conifer forests.
Aboveground biomass
Sample size (n) Statistical resultsMeans
95% Confidence
Intervals (⫾)
Mature Old-growth Mature Old-growth Mature Old-growth Tcritical Tstatistic P
...............(Mg/ha) ...............
Live Trees
Region 194.95 266.33 17.56 23.65 48 46 1.663 ⫺4.749 ⬍0.001
ME 226.13 283.33 36.05 51.93 18 17 1.699 ⫺1.774 0.043
NH 194.84 277.75 24.47 34.09 11 8 1.761 ⫺3.872 ⬍0.001
NY 165.49 248.22 17.22 27.17 19 21 1.692 ⫺5.041 ⬍0.001
Standing dead
Region
1
21.49 41.44 4.07 7.89 37 38 1.673 ⫺4.407 ⬍0.001
ME 20.82 36.78 7.06 12.81 18 17 1.708 ⫺2.139 0.021
NY 22.12 45.21 4.43 9.79 19 21 1.701 ⫺4.212 ⬍0.001
Total
Region
1
216.48 305.37 22.43 29.93 37 38 1.668 ⫺4.659 ⬍0.001
ME 246.95 320.11 37.45 54.62 18 17 1.699 ⫺2.165 0.019
NY 187.61 293.43 18.38 31.70 19 21 1.694 ⫺5.660 ⬍0.001
Downed coarse woody debris
Region
1
27.10 35.94 5.07 4.79 37 38 1.666 ⫺2.569 0.006
ME 24.16 28.15 6.00 6.12 18 17 1.692 ⫺0.985 0.166
NY 29.88 42.25 8.47 6.21 19 21 1.691 ⫺2.467 0.009
Sample sizes varied because of differences among the four subregional data sets. Comparisons were significant at
␣
⫽0.05.
1
Includes Maine and New York only; sufficient sample sizes were not available for New Hampshire for these variables.
494 Forest Science 57(6) 2011

We used linear regression modeling to analyze relation-
ships between dominant tree age and biomass/stand struc-
ture variables. Alternate curve-fitting techniques were used
to assess the relative fit of linear, logarithmic, polynomial
(second to fourth order), and negative exponential curves,
applied as transformations of the dependent variable. Re-
siduals were plotted and examined for evidence of het-
eroscedasticity. Where nonlinear curves explained equal or
greater variance compared with a linear trend and where
asymptotic relationships were the most mechanistically
plausible, we selected the nonlinear (in all cases logarith-
mic) curve as the final model.
We used Tukey honestly significant difference tests,
assuming unequal variance for all categorical comparisons
of biomass and stand structure variables by age class (ma-
ture versus old-growth) and state. We also used Tukey tests
and one-way analysis of variance to examine differences in
basal area estimates between data sets as a function of
sampling intensity and subregion. Basal area was selected as
the indicator metric for these analyses because of its sensi-
tivity to sampling intensity in spatially heterogeneous sys-
tems (Shivers and Borders 1996).
For the final step in our analytical procedure we con-
ducted a classification and regression tree (CART) analysis.
This allowed us to examine the relative influence of forest
age class versus ecoregion and the proportion of conifers
(by basal area) at each site (independent variables) on
aboveground biomass (dependent variable). Sites were as-
signed to ecological units (sections) according to the Na-
tional Hierarchy of Ecological Units (Table 1) following
Smith and Carpenter (1996). Our rationale was that subre-
gional ecological variability (e.g., climate, soils, productiv-
ity, and disturbance regime) might account for variability in
biomass levels independent of forest age class (McNab et al.
2007). Likewise, the shade-tolerant conifer component in
mixed stands or increasing late in succession has the poten-
tial to influence biomass development (Stoy et al. 2008).
Because our sites encompass a range of percent conifer
values, it was important to distinguish this source of vari-
ability from the effects of forest age. CART is robust, a
nonparametric, binary procedure, accommodating both cat-
egorical and continuous variables (De’ath and Fabricius
2000). The procedure hierarchically partitions values of the
independent variables through successive splits based on the
amount of variance explained in values of the dependent
variable. We used cost-complexity pruning to eliminate
nonsignificant nodes. CART was not used in our study to
establish definitive threshold values for the predictor vari-
ables. Rather, CART provided a way to rank the relative
predictive strength of the independent variables of interest.
Results
Categorical Comparisons between Mature and
Old-Growth Forests
Old-growth sites in our data set had significantly higher
mean levels of aboveground tree biomass in comparison
with mature stands. This held true for live, standing dead,
and total (live ⫹standing dead) tree biomass as well as for
DCWD. We found old-growth biomass to be significantly
greater at both subregional and regional scales (Table 2). As
an average for the region, old-growth forests contained 266
and 41 Mg/ha of live and standing dead aboveground tree
biomass, respectively. These values were compared with
regional means of 195 (live) and 25 Mg/ha (standing dead)
for mature stands. Thus, the mature (80- to 150-year-old)
stands contained 73% of the live biomass and 61% of the
standing dead tree biomass found on average in old-growth
forests. The biomass contrasts were consistent with other
measures of structural complexity. For instance, basal areas
(live and standing dead), large tree densities (live and stand-
ing dead), and quadratic means diameters were all signifi-
cantly greater in old-growth compared with mature forest
sites. Stem densities were not significantly different be-
tween age classes.
Although the biomass means were statistically different,
there was considerable variation across the region. For
example, mature live tree biomass ranged from 90 Mg/ha at
a site in New York to as high as 391 Mg/ha in a productive,
well-developed stand approaching 150 years of age in
Maine. Old-growth live tree biomass, in comparison, ranged
from 104 Mg/ha at a recently disturbed site to 569 Mg/ha,
a value almost twice the regional average for old-growth;
both sites were in Maine. The highest average live tree
aboveground biomass levels for old-growth were reached in
Maine (283 Mg/ha), with New Hampshire (278 Mg/ha)
intermediate and New York (248 Mg/ha) the lowest, al-
though these subregional means were not statistically dif-
ferent. Live biomass in mature forests was proportionately
lowest, relative to that in old-growth, in New York (67%)
compared with New Hampshire (70%) and Maine (80%);
again these contrasts between subregions were not statisti-
cally significant.
Both DCWD volume and biomass were significantly
greater in old-growth relative to that in mature forests for
New York and the region; DCWD biomass was not signif-
icantly different between age classes for Maine although
volumes were (Table 3). Regional mean volumes were 83
m
3
/ha in mature stands (or 32% lower) compared with 122
m
3
/ha in old-growth stands. The mature sites in New York
had a larger percentage (55%) of DCWD volume in less-
decayed classes (1–2), whereas old-growth sites had a larger
proportion (61%) in well-decayed classes (3–5). Biomass
contained in the downed log pool represented 11.1 and
10.5% of the total aboveground biomass (DCWD ⫹live
and standing dead trees), not including understory vegeta-
tion and fine litter, for mature and old-growth stands, re-
spectively. Thus, we did not find a significant difference
between age classes in DCWD biomass as a proportion of
total aboveground biomass. Old-growth stands had signifi-
cantly higher mean DCWD biomass values (36 Mg/ha) in
relation to mature stands (27 Mg/ha). Like the other struc-
tural variables, DCWD biomass varied widely throughout
the region and within age classes (Table 2). Levels ranged
from 6 to 71 Mg/ha for mature stands and from 11 to 65
Mg/ha for old-growth, similar ranges although the means
differed. At the subregional scale, New York had higher
DCWD biomass levels than Maine for both mature and
old-growth sites; these differences were statistically
significant.
Forest Science 57(6) 2011 495

Biomass Development in Relation to Dominant
Tree Age
Aboveground tree biomass showed positive relationships
with dominant tree age for our sites in the Adirondack
region of New York (Figure 1). There were significant
relationships for live (r
2
⫽0.52, P⬍0.001), standing dead
(r
2
⫽0.36, P⬍0.001), and total (r
2
⫽0.63, P⬍0.001)
aboveground tree biomass. In each case we fitted logarith-
mic curves to the data because these explained the greatest
amount of variation, although linear trends exhibited similar
correlation coefficients. The logarithmic curves exhibited
moderately strong asymptotic relationships, showing evi-
dence of leveling off at values close to 280, 230, and 50
Mg/ha for total, live, and standing dead aboveground tree
biomass, respectively, in stands with dominant tree age
⬎400 years. However, the trend lines did not reach clear
asymptotes over the range of dominant tree ages evaluated,
suggesting the potential for continued net positive biomass
accumulations into greater ages. There was no evidence of
peaks or declines in biomass before the maximum dominant
tree age achieved in the data set. This conclusion was based
on fitting polynomial curves capable of detecting such re-
lationships. The regression results were limited by the rel-
ative lack of data points for age values between 150 and
250, although the 205-year-old site had aboveground live
biomass very close to the trend line (Figure 1). Although the
predicted biomass trends for this gap in the age range are
similar to basal area trends previously reported for the
Adirondacks (Ziegler 2000), we cannot rule out the possi-
bility that the asymptote is reached earlier than predicted by
our regression equations.
The strong positive relationship between biomass and
dominant tree age reflects, in part, the increasing density of
large trees (⬎50 cm dbh) as stands develop (Figure 2).
Large tree densities were strongly related to dominant tree
ages based on regression results. A logarithmic curve ex-
plained the greatest variation for live large trees (r
2
⫽0.64,
P⬍0.001), whereas a linear relationship had the best fit for
large standing dead trees (r
2
⫽0.41, P⬍0.001). Large tree
density is, in turn, strongly correlated (r
2
⫽0.48, P⬍
0.001) with aboveground biomass, as demonstrated by a
regression of live large density against live tree biomass
across all 94 sites in our data set (Figure 2, bottom panel).
Neither total (all sizes) live tree nor total standing dead
densities were significantly correlated with age across the
age range examined, with linear regressions exhibiting
slopes near zero.
That total aboveground biomass is increasing across the
range of dominant tree ages in our data set is demonstrated
also by the positive relationship between age and DCWD
biomass (Figure 3). Although DCWD biomass was highly
variable among sites, a logarithmic curve explained 24% of
the variation in the relationship with dominant tree age.
DCWD biomass approached an asymptote of approximately
55 Mg/ha in stands 400 years of age or older, but like
standing tree biomass, did not level off over the range of
ages assessed. Thus, continued DCWD biomass accumula-
tions are possible into older dominant tree ages.
Figure 1. Aboveground biomass in relation to dominant canopy tree age in the Adirondacks, New York
(nⴝ29). Shown are logarithmic models for live [yⴝ65.39ln(x)ⴚ158.19], standing dead [yⴝ17.389ln(x)ⴚ
61.633], and total [live ⴙstanding dead, yⴝ82.779ln(x)ⴚ219.83] trees.
496 Forest Science 57(6) 2011

Relative Influence of Age Class, Ecoregion,
and Species Composition
Forest age has a controlling and dominant effect on
biomass development relative to ecoregional variability and
the conifer component in mixed hardwood-conifer stands.
This conclusion was clearly supported by CART results, in
which age class emerged as the primary (top tier) predictor
of aboveground live tree biomass (Figure 4). However, both
ecoregion and percent conifer were included in the CART
(and thus were statistically significant) as secondary (lower
tiered) predictors. For old-growth sites, percent conifer ex-
plained variation in aboveground biomass, with sites com-
prising between 5 and 22% conifer basal area having the
greatest aboveground biomass (mean ⫽301 Mg/ha). Sites
falling to less than 5% were intermediate (mean ⫽288
Mg/ha) and sites greater than 22% were lowest (mean ⫽
Figure 2. Top panel: Density of large trees (>50 cm dbh) in relation to stand age (nⴝ29). A logarithmic
model [yⴝ29.166ln(x)ⴚ127.03] explained the most variation for live trees, whereas standing dead trees
exhibited a linear relationship (yⴝ0.0295xⴚ0.584). Bottom panel: Relationship between live large tree
density and aboveground live tree biomass (nⴝ79), exhibiting a linear model fit (yⴝ2.6814xⴙ143.09).
Forest Science 57(6) 2011 497

223 Mg/ha) in biomass within the old-growth age class.
Together these results signaled a range in which a minor
conifer component is sometimes associated with elevated
biomass in old mixed stands, provided this does not exceed
approximately one-quarter of total live tree basal area. Vari-
ance (or deviance) in our data set explained by percent
conifer was approximately one-third of that explained by
age class alone on the basis of the CART output.
Biomass development was also associated with ecolog-
ical variability attributable to differences among ecoregions
(or sections). Ecoregion was selected in the CART as a
secondary explanatory variable for mature sites. Within the
Figure 3. Relationship between DCWD biomass and stand age (nⴝ29). A logarithmic curve explained
the greatest variation in this relationship [yⴝ18.463ln(x)ⴚ57.495].
Figure 4. CART model, showing statistically significant (
␣
ⴝ0.05) independent variables selected, split values, and
partitioned mean values (bottom) of the dependent variable (aboveground live tree biomass, nⴝ94). The figure ranks
variables by predictive strength (top to bottom). Length of each vertical line is proportional to the amount of deviance
explained. Minimum observations required for each split ⴝ15; minimum deviance ⴝ0.01.
498 Forest Science 57(6) 2011

mature age class, aboveground biomass was lowest
(mean ⫽166 Mg/ha) at sites in the Adirondack Highlands
section (M212D) of New York, and two sections (212A and
212C) in the Laurentian Hills and Lowlands Province of
southeastern to northeastern Maine. Mature forest biomass
was higher (mean ⫽222 Mg/ha) in the White Mountains
section (M212A) of New Hampshire and western Maine,
and one section (212B) of the Laurentian Province in
Maine. Ecoregion explained approximately one-fourth of
the variation explained by age class alone.
Error Related to Subsampling Intensity
An important difference among the data sets was the
intensity of subsampling (i.e., use of multiple plots) within
sites. In one data set, subsampling consisted of 6 –10 plots
per site, whereas for the other data sets there were only 1 or
2 plots. Because total gap area in late-successional/old-
growth northern hardwood forests can range from 5 to 10%
(Runkle 1982, Ziegler 2002), sampling methods that fail to
capture fine-scaled spatial variability (Curzon and Keeton
2010) are prone to error in terms of estimating forest struc-
ture attributes. For instance, excluding gaps from the sample
may cause an overestimation of basal area and aboveground
biomass in both mature and old-growth stands, although the
error will be greater for the latter because of their higher gap
fractions (Dahir and Lorimer 1996).
Despite these limitations, single plot per site data from
old forests are commonly used for regional and even global
estimates of biomass dynamics and carbon stocks (see, for
example, sources reviewed in Luyssaert et al. 2008 and
Keith et al. 2009). Nevertheless, it is important to recognize
that the data derived from sites sampled more intensively
are probably more robust. However, we found no statisti-
cally significant difference in basal area estimations for the
Adirondacks in a comparison of the sites sampled inten-
sively versus those sampled with only two plots. This com-
parison held for both old-growth (P⫽0.35) and mature
(P⫽0.27) forest sites. Likewise, there were no statistical
differences when we compared mature (P⫽0.16) and
old-growth (P⫽0.47) basal areas among all four data sets
based on analysis of variance results. These results suggest
that the random plot placement used in all four data sets
may have minimized this source of error.
Discussion
Evaluation of Competing Hypotheses
Our results support the hypothesis that biomass has the
potential to increase very late into stand development,
showing only slight declines as dominant trees pass 300
years of age, and continued additions to 400 years and
beyond. Our data showed no evidence of peaks in early
old-growth (e.g., approximately 200 years of age), declines
subsequently (i.e., a “transition” phase), or steady-state dy-
namics as predicted by Bormann and Likens (1979). Cor-
relations between total aboveground biomass and dominant
tree age were related not just to increases in the standing
dead tree component but also to substantial biomass accrual
in live trees. Large tree densities (live and standing dead)
were strongly correlated with both age and aboveground
biomass, whereas total live and standing dead stem densities
showed no relationship with age across the range repre-
sented in our data set, a phenomena observed previously
(Keeton et al. 2007). These results suggest that large trees
make a proportionately greater contribution to total biomass
as stands undergo late-successional development (Brown et
al. 1997) and where disturbance dynamics allow persistence
of some canopy dominants to very old ages.
Continued biomass development leads to substantially
greater carbon storage levels in both live and standing dead
tree pools (standing and downed), in comparison to levels
achieved during maturation stages of development (Franklin
et al. 2002). This inference was strongly supported by the
categorical comparisons of age classes, which resulted in
consistent findings for all of the subregions examined. Col-
lectively the results suggest the potential for biomass addi-
tions (positive NEP) and related high-magnitude carbon
storage over longer time frames than predicted previously
for the northern hardwood region (Figure 5, top panel).
Maximum attainable biomass levels and accrual rates are
likely to vary by site, species composition, and ecoregion, as
suggested by the CART results. For instance, total above-
ground biomasses reached at our oldest sites (approximately
250 –350 Mg/ha in stands ⬎350 years old) are still only
equal to or slightly below peak levels initially predicted for
Hubbard Brook by the JABOWA model (350 Mg/ha,
reached at approximately 200 years, declining to a long-
term average of 300 Mg/ha after 400 years) (Whittaker et al.
1974).
Our findings did not directly support the alternate hy-
pothesis offered by Fahey et al. (2005) regarding biomass
stabilization or decline earlier than predicted by Bormann
and Likens (1979). Likewise, they were not consistent with
Martin and Bailey (1999) who, working at the Bowl Re-
search Natural Area in New Hampshire, found that a logged
stand recovered biomass comparable to average levels (208
Mg/Ha) after 100 years in a mosaic of (variable structure)
old-growth areas. However, there are several possible ex-
planations for these discrepancies. We propose three, each
of which warrants further investigation and none of which is
mutually exclusive.
Intraregional Variability
The first is that discrepancies among empirical studies
may reflect regional- and site-level variability in late-suc-
cessional biomass development processes. By this explana-
tion, the Hubbard Brook findings may represent a lower
point within this spectrum, whereas the highest values re-
ported in this and other articles (e.g., Woods and Cogbill
1994, Ziegler 2000) may define the upper limits. Evidence
supporting this line of thinking include the documented
effects of anthropogenic stressors on forest productivity,
expressed heterogeneously throughout the northern hard-
wood region (e.g., Aber et al. 2001, Schaberg et al. 2001),
the region’s pronounced site-specific edaphic variability
(Seymour 1995), and variable stand dynamics attributable
to different natural disturbance histories (Seymour et al.
Forest Science 57(6) 2011 499

2002). Land-use history is another major source of variabil-
ity affecting stand development rates and trajectories in
secondary forests (Foster et al. 1998, McLachlan et al.
2000).
The CART results showed that although forest age was
the dominant factor, other sources of variability are impor-
tant determinants of biomass development potential. Like-
wise, our data set contained a wide range of aboveground
biomass values, with substantial overlap between the lower
values obtained in the old-growth age class and the higher
values presented in the mature age class. This variability
clearly reflected the range of conditions encountered in all
stands, including extent of recent wind throw and ice dam-
age (Manion and Griffin 2001, Curzon and Keeton 2010),
site quality (Keeton et al. 2007), and mortality related to
beech bark disease (Gavin and Peart 1993, McGee et al.
1999). It also may reflect varying degrees of historical
harvesting effects in mature stands. For instance, some of
our unmanaged mature stands fell within the range of highly
disturbed old-growth stands.
Past versus Present
A second explanation is that biomass development may
have occurred differently in the past than it does in the
present. It is important to recognize that our study is a
space-for-time substitution, and there are limitations to the
inferences that can be drawn from chronosequence research.
Our data do not track the individual pathways along which
each stand has developed over past centuries or recent
decades, and, therefore, they do not provide a basis for
evaluating current biomass dynamics. It is thus possible that
sustained positive biomass development may have been
more likely in the past, whereas anthropogenic stressors
Figure 5. Conceptual models of stand-scale aboveground biomass development pathways for northern
hardwood-conifer forests. Top panel: Pathway A depicts the widely referenced Bormann and Likens (1979)
model, based on empirical data for young stands and theoretical projections for late-successional stands.
Pathway B shows observed live tree biomass trends in a secondary stand at Hubbard Brook Experimental
Forest, with earlier than predicted stabilization and possible decline (Fahey et al. 2005). Pathway C is based
on empirical data presented in this and other studies (e.g., Ziegler 2000), suggesting continued long-term
biomass accumulation potential in late-successional forests. Collectively these pathways may represent a
range of variability for the northern hardwood region. Bottom panel: Hypothetical aboveground biomass
dynamics in response to partial disturbances, such as microbursts, moderate-severity hurricanes, torna-
does, and ice storms. Note that live tree biomass declines following each disturbance, but it is hypothesized
to recover asymptotically due to maintenance of multiaged to uneven-aged structures. If the time between
partial disturbances is long enough, steady-state dynamics driven by gap dynamics (indicated in the panel
by the dip before the second disturbance) are theoretically possible, although these were not evident in our
data set.
500 Forest Science 57(6) 2011

may have reduced this potential in the present as proposed
by Fahey et al. (2005).
Primary versus Secondary Forest Dynamics
A third possibility is that biomass dynamics may be
fundamentally different in primary forests compared with
that in the secondary forests now recovering across the
northeastern United States and observed at Hubbard Brook.
This difference, in turn, may relate to the natural distur-
bances affecting primary forest landscapes over past centu-
ries. The literature for the northeastern United States has
emphasized either high-frequency, low-intensity distur-
bances, such as fine-scaled gap-forming events or very
low-frequency (e.g., return intervals exceeding 1,000
years), high-intensity disturbances, such as high-severity
hurricanes, and may have underreported intermediate inten-
sity disturbances (Seymour et al. 2002). However, there is a
growing recognition of intermediate-intensity events oper-
ating across centennial-scale frequencies, such as micro-
bursts, tornadoes, insect outbreaks, and ice storms (North
and Keeton 2008). They are also caused by low- to moder-
ate-severity hurricanes in interior and northwestern portions
of the region (Boose et al. 2001). Intermediate-intensity
disturbances cause partial mortality with high, although
spatially variable, densities of residual live trees (Millward
and Kraft 2004, Woods 2004, Hanson and Lorimer 2007).
They have been shown to produce nonequilibrium dynamics
in Adirondack old-growth forests (Ziegler 2002, Curzon
and Keeton 2010).
Nonequilibrium dynamics attributable to partial distur-
bances would produce biomass responses very different
from those for the Bormann and Likens (1979) model for
secondary forests. The latter predicted an early peak in
biomass primarily due to an initial period of even-aged
development. The predicted point of maximal biomass de-
velopment was a function of efficient growing space allo-
cation. Canopy disturbances, mortality in the dominant co-
hort, and changing species composition were predicted to
result in subsequent declines and steady-state dynamics in
stands composed of a “shifting mosaic of patches” (Whit-
taker et al. 1974). However, in a forest landscape subject to
spatially heterogeneous, variable frequency but periodic
partial disturbances, complete reinitiation of stand develop-
ment would seldom occur (Schulte and Mladenoff 2005,
D’Amato and Orwig 2008), and thus there would rarely, if
ever, be an early stage of true even-aged structural devel-
opment at the stand scale. Instead, stand-scale biomass
would drop after each disturbance, only to recover subse-
quently while maintaining multiaged structure (Woods
2004, Hanson and Lorimer 2007) (Figure 5, bottom panel).
This type of dynamic would produce the same logarithmic
relationships with dominant tree age we found across our
chronosequence. Bormann and Likens (1979) actually sug-
gested that such a dynamic might be possible in uneven-
aged systems characterized by partial disturbances. Thirty
years later there is emerging empirical evidence that may
support this prediction. We propose this as a hypothesis
worthy of further investigation.
Error in Biomass Estimation
A potential source of error in our study was the use of
diameter-based allometric equations to estimate large trees
biomass. The equations of Jenkins et al. (2003) incorporate
a correction factor for large trees based on Freedman (1984)
but do not account for rotten wood specifically. Lacking an
estimate of heartrot frequency, we may have overpredicted
biomass for large trees in some cases. Error also may have
been introduced by the assumed relationship between tree
diameter and height as predicted by the equations of Jenkins
et al. (2003). However, the assumption of a positive rela-
tionship for dominant and codominant trees was supported
by the significantly greater canopy heights at the old-growth
compared with those at mature sites in New York as re-
ported by Keeton et al. (2007). Having higher mean stand
diameters, large tree densities, and canopy heights, the
predicted biomass differences are probably robust.
Standing dead tree biomass estimates were not corrected
for height or fragmentation because of incomplete data
across the full regional data set. This is an important limi-
tation given that snapped stems can be common in late-suc-
cessional northern hardwood-conifer forests (Curzon and
Keeton 2010). However, an analysis of tree height data in
the Keeton data set showed significantly higher mean snag
heights at the old-growth sites, consistent with the inferred
biomass trends. In addition, there were no differences in
snag density or basal area distributions by decay stage
(which incorporate evidence of stem breakage and deterio-
ration) in relation to age class or dominant tree age. Thus,
these sources of error are not likely to have altered the
fundamental age-related contrasts reported in this article,
although we are likely to have overestimated standing dead
tree biomass, possibly by as much as 22%.
Implications for Carbon Dynamics and Storage
in Northeastern Temperate Forests
Our results add to a growing body of literature showing
that old-growth temperate forests have greater carbon se-
questration and storage potential than previously realized
(Keeton et al. 2010). A key finding of recent research,
including gas exchange studies, has been that equilibrium
NEP conditions, as previously predicted and modeled in
regional and global carbon budgets (Houghton 2005), are
seldom reached at stand scales in temperate and boreal
forests (Pregitzer and Euskirchen 2004, Luyssaert et al.
2008). Instead, net positive carbon uptake frequently con-
tinues very late into stand development (Carey et al. 2001,
Keith et al. 2009). For example, high levels of net annual
carbon storage (3.0 Mg/ha) have been measured by the eddy
covariance method in a ⬃200-year-old old-growth eastern
hemlock stand in New England (Hadley and Schedlbauer
2002). Hadley and Schedlbauer accounted for known
sources of error in the eddy covariance method (see Field
and Kaduk 2004), such as incorrect instrument response and
footprint size of the flux system but could not rule out error
due to advective CO
2
flux. There are conflicting findings,
such as the observed declines at Hubbard Brook, attributed
to reduced growth in live trees, leading Siccama et al.
Forest Science 57(6) 2011 501

(2007) to conclude that W6 is no longer a carbon sink.
Long-term, net positive NEP has been attributed both to
increases in the dead wood and soil carbon pools (Harmon
and Marks 2002) as well as to efficient three-dimensional
allocation of growing space in uneven-aged, old-growth
forests (Luyssaert et al. 2008). The latter is a recognition
that tree mortality in the upper canopy of uneven-aged
stands is compensated for by fast colonization of available
growing space by co- and subdominant trees, a function of
the vertically complex canopies and wide range of tree
sizes/ages characteristic of old-growth forests (Franklin and
Van Pelt 2004). However, research has also shown subor-
dinate trees to have lower light use efficiencies, which may
contribute to lower NPP in some forests with heterogeneous
canopies (Binkley et al. 2010).
In this context our results suggest that there is a signif-
icant potential to increase total carbon storage in the North-
east’s northern hardwood-conifer forests. Young to mature
secondary forests in the northeastern United States today
have aboveground biomass (live and dead) levels of ⬃107
Mg/ha on average (Turner et al. 1995, Birdsey and Lewis
2003). Thus, assuming a maximum potential aboveground
biomass range for old-growth of approximately 250 – 450
Mg/ha, a range consistent with upper thresholds in our data
set and the lower threshold observed at Hubbard Brook, our
results suggest a potential to increase in situ forest carbon
storage by a factor of 2.3– 4.2, depending on site-specific
variability. This would sequester an additional 72–172
Mg/ha of carbon. A reconstruction of pre-European settle-
ment forest carbon storage in Wisconsin produced estimates
within similar ranges for both old-growth carbon storage
and current recovery potentials (Rhemtulla et al. 2009).
Although secondary forests may have the potential to
more than double their aboveground carbon storage in some
cases, there are many factors that will influence future
trajectories of biomass development. Type and intensity of
forest management approach will play a major role (Keeton
2006; Harmon et al. 2009). With development of carbon
markets, this may include both reserve-based approaches,
leading to development of late-successional/old-growth
conditions, as well as those designed to optimize net carbon
storage in actively managed forests (Harmon and Marks
2002, Seidl et al. 2007, Nunery and Keeton 2010). Carbon
management probably will entail a combination of more
intensive approaches intended to achieve net emissions re-
ductions though product and energy substitution (Eriksson
et al. 2007) and less intensive approaches, such as extended
rotations and structural retention, favoring maintenance of
high levels of in situ forest carbon storage (Ray et al.
2009b). Thus, carbon storage recovery is likely to be het-
erogeneous at landscape scales as a function of management
alone.
However, there other important sources of uncertainty
that must be considered. Global change, including climate
system disruption and spread of exotic organisms, ranks
foremost among these. Climate change may influence both
rates and pathways (e.g., successional dynamics) of future
biomass development (Aber et al. 2001, Iverson et al. 2008).
Whether these result in negative or positive effects on
late-successional forest carbon storage potential will depend
on many factors, including atmospheric CO
2
fertilization
effects, intensity of warming and precipitation changes,
extent of species range shifts, and interactions with other
stressors, such as disturbances, disease, airborne pollutants,
and land use (Ollinger et al. 2002, Beckage et al. 2008).
Thus, although old-growth reference stands suggest an in-
herent capacity within the system, future carbon storage
dynamics are likely to differ from historic benchmarks as
environmental boundary conditions change (Seidl et al.
2008, Ollinger et al. 2008).
Literature Cited
ABER, J., R.P. NEILSON,S.MCNULTY, J.M. LENIHAN,D.BACHE-
LET,AND R.J. DRAPEK. 2001. Forest processes and global
environmental change: predicting the effects of individual and
multiple stressors. BioScience 51:735–751.
ABRAMS, M.D., AND D.A. ORWIG. 1996. A 300-year history of
disturbance and canopy recruitment for co-occurring white pine
and hemlock on the Allegheny Plateau, USA. J. Ecol. 84:
353–363.
ANGERS, V.A., C. MESSIER,M.BEAUDET,AND A. LEDUC. 2005.
Comparing composition and structure in old-growth and har-
vested (selection and diameter-limit cuts) northern hardwood
stands in Quebec. For. Ecol. Manag. 217:275–293.
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 Ver-
mont. Proc. Natl. Acad. Sci. USA 105:4197– 4202.
BINKLEY, D., J. LUIZ STAPE, W.L. BAUERLE,AND M.G. RYAN.
2010. Explaining growth of individual trees: Light interception
and efficiency of light use by Eucalyptus at four sites in Brazil.
For. Ecol. Manag. 259:1704 –1713.
BIRDSEY, R.A., AND G.M. LEWIS. 2003. Carbon in U.S. forests and
wood products, 1987–1997: State-by-state estimates. US For.
Serv. Gen. Tech. Rep. NE-310.
BOOSE, E.R., K.E. CHAMBERLIN,AND D.R. FOSTER. 2001. Land-
scape and regional impacts of hurricanes in New England.
Ecol. Monogr. 71:27– 48.
BORMANN, F.H. AND G.E. LIKENS. 1979. Pattern and process in a
forested ecosystem. Springer-Verlag, New York. 253 p.
BOTKIN, D.B., J.F. JANAK,AND J.R. WALLIS. 1972. Some ecolog-
ical consequences of a computer model of forest growth. J.
Ecol. 60:849 – 872.
BROWN, S., P. SCHROEDER,AND R. BIRDSEY. 1997. Aboveground
biomass distribution of US eastern hardwood forests and the
use of large trees as an indicator of forest development. For.
Ecol. Manag. 96:37– 47.
BROWN, S.L., P. SCHROEDER,AND J.S. KERN. 1999. Spatial distri-
bution of biomass in forests of the eastern USA. For. Ecol.
Manag. 123:81–90.
CAREY, E.V., A. SALA,R.KEANE,AND R.M. CALLAWAY. 2001.
Are old forests underestimated as global carbon sinks? Glob.
Change Biol. 7:339 –344.
CROW, T.R., D.S. BUCKLEY, E.A. NAUERTZ,AND J.C. ZASADA.
2002. Effects of management on the composition and structure
of northern hardwood forests in upper Michigan. For. Sci.
48:129 –145.
CURZON, M.T., AND W.S. KEETON. 2010. Spatial characteristics of
canopy disturbances in riparian old-growth hemlock-northern
hardwood forests, Adirondack Mountains, New York, USA.
Can. J. For. Res. 40:67– 80.
D’AMATO, A.W., AND D.A. ORWIG. 2008. Stand and landscape-
level disturbance dynamics in old-growth forests in western
Massachusetts. Ecol. Monogr. 78:507–522.
502 Forest Science 57(6) 2011

D’AMATO, A.W., D.A. ORWIG,AND D.R. FOSTER. 2006. New
estimates of Massachusetts old-growth forests: Useful data for
regional conservation and forest reserve planning. Northeast.
Natural. 13:495–506.
DAHIR, S.E., AND C.G. LORIMER. 1996. Variation in canopy gap
formation among developmental stages of northern hardwood
stands. Can. J. For. Res. 26:1875–1892.
DAVIS, M.B. 1996. Extent and location. P. 18 –34 in Eastern
old-growth forests: Prospects for rediscovery and recovery
(Davis, M.B., ed.). Island Press, Washington, DC.
DE’ATH, G., AND K.E. FABRICIUS. 2000. Classification and regres-
sion trees: A powerful yet simple technique for ecological data
analysis. Ecology 81:3178 –3192.
DIXON, G.E. 2002. Essential FVS: A user’s guide to the Forest
Vegetation Simulator. US Forest Service, Forest Management
Service Center, Fort Collins, CO.
DUNWIDDIE, P., D. FOSTER,D.LEOPOLD,AND R.T. LEVERETT.
1996. Old-growth forests of southern New England, New York,
and Pennsylvania. P. 126 –143 in Eastern old-growth forests:
Prospects for rediscovery and recovery (Davis, M.B., ed.).
Island Press, Washington, DC.
ERIKSSON, E.A., R. GILLESPIE,L.GUSTAVSSON,O.LANGVALL,M.
OLSSON,R.SATHRE,AND J. STENDAH. 2007. Integrated carbon
analysis of forest management practices and wood substitution.
Can. J. For. Res. 37:671– 681.
FAHEY, T.J., T.G. SICCAMA, C.T. DRISCOLL, G.E. LIKENS,J.
CAMPBELL, C.E. JOHNSON, J.J. BATTLES, J.D. ABER, J.J. COLE,
M.C. FISK, P.M. GROFFMAN, S.P. HAMBURG, R.T. HOLMES,
P.A. SCHWARZ,AND R.D. YANAI. 2005. The biogeochemistry
of carbon at Hubbard Brook. Biogeochemistry 75:109 –176.
FIELD, C.B., AND J. KADUK. 2004. The carbon balance of an
old-growth forest: Building across approaches. Ecosystems
7:525–533.
FOSTER, D.R. 1988. Disturbance history, community organization
and vegetation dynamics of the old-growth Pisgah forest,
south-western New Hampshire, U.S.A. J. Ecol. 76:105–134.
FOSTER, D.R., G. MOSKIN,AND B. SLATER. 1998. Land-use history
as long-term broad scale disturbance: Regional forest dynamics
in central New England. Ecosystems 1:96 –119.
FRANKLIN, J.F., T.A. SPIES,R.VAN PELT, A.B. CAREY, D.A.
THORNBURGH, D.R. BERG, D.B. LINDENMAYER, M.E. HARMON,
W.S. KEETON, D.C. SHAW,K.BIBLE,AND J.Q. CHEN. 2002.
Disturbances and structural development of natural forest eco-
systems with silvicultural implications, using Douglas-fir for-
ests as an example. For. Ecol. Manag. 155:399 – 423.
FRANKLIN, J.F., AND R. VAN PELT. 2004. Spatial aspects of struc-
tural complexity in old-growth forests. J. For. 102:22–28.
FREEDMAN, B. 1984. The relationship between the aboveground
dry weight and diameter for a wide size range of erect land
plants. Can. J. Bot. 62:2370 –2374.
GAVIN, D.G., AND D.R. PEART. 1993. Effects of beech bark disease
on the growth of American beech (Fagus grandifolia). Can. J.
For. Res. 23:1566 –1575.
GOODALE, C.L., AND J.D. ABER. 2001. The long-term effects of
land-use history on nitrogen cycling in northern hardwood
forests. Ecol. Applic. 11:253–267.
GOODBURN, J.M., AND C.G. LORIMER. 1999. Population structure
in old-growth and managed northern hardwoods: An examina-
tion of the balanced diameter distribution concept. For. Ecol.
Manag. 118:11–29.
HADLEY, J.L., AND J.L. SCHEDLBAUER. 2002. Carbon exchange of
an old-growth eastern hemlock (Tsuga canadensis) forest in
central New England. Tree Physiol. 22:1079 –1092.
HALE, C.M., J. PASTOR,AND K.A. RUSTERHOLTZ. 1999. Compar-
ison of structural and compositional characteristics in old-
growth and mature, managed hardwood forests of Minnesota,
U.S.A. Can. J. For. Res. 29:1479 –1489.
HANSON, J., AND C.G. LORIMER. 2007. Forest structure and light
regimes following moderate wind storms: Implications for
multi-cohort management. Ecol. Applic. 17:1325–1340.
HARMON, M.E., W.K. FERRELL,AND J.F. FRANKLIN. 1990. Effects
on carbon storage of conversion of old-growth forests to young
forests. Science 247:699 –702.
HARMON, M.E., AND B. MARKS. 2002. Effects of silvicultural
practices on carbon stores in Douglas-fir-western hemlock for-
ests in the Pacific Northwest, USA: Results from a simulation
model. Can. J. For. Res. 32:863– 877.
HARMON, M.E., A. MORENO,AND J.B. DOMINGO. 2009. Effects of
partial harvest on carbon stores in Douglas-fir/western hemlock
forests: A simulation study. Ecosystems 12:777–791.
HARMON, M.E., C.W. WOODALL,B.FASTH,AND J. SEXTON. 2008.
Woody detritus density and density reduction factors for tree
species in the United States: A synthesis. US For. Serv. Gen.
Tech. Rep. NRS-29.
HOCKER, H.W., AND D.J. EARLY. 1983. Biomass and leaf area
equations for northern forest species. Res. Rep. 102. New
Hampshire Agricultural Experiment Station, Durham, NH.
HOUGHTON, R.A. 2005. Aboveground forest biomass and the
global carbon balance. Glob. Change Biol. 11:945–958.
HUDIBURG, T., B. LAW, D.P. TURNER,J.CAMPBELL,D.DONATO,
AND M. DUANE. 2009. Carbon dynamics of Oregon and North-
ern California forests and potential land-based carbon storage.
Ecol. Applic. 19:163–180.
IVERSON, L., A. PRASAD,AND S. MATTHEWS. 2008. Modeling
potential climate change impacts on the trees of the northeast-
ern United States. Mitig. Adapt. Strat. Glob. Change 13:
487–516.
JENKINS, J.C., D.C. CHOMNACKY, L.S. HEATH,AND R.A. BIRDSEY.
2003. National scale biomass estimators for United States tree
species. For. Sci. 49:12–35.
KEETON, W.S. 2006. Managing for late-successional/old-growth
characteristics in northern hardwood-conifer forests. For. Ecol.
Manag. 235:129 –142.
KEETON, W.S., M. CHERNYAVSKYY,G.GRATZER,M.MAIN-
KNORN,M.SHPYLCHAK,AND Y. BIHUN. 2010. Structural char-
acteristics and aboveground biomass of old-growth spruce-fir
stands in the eastern Carpathian Mountains, Ukraine. Plant
Biosyst. 144:1–12.
KEETON, W.S., AND J.F. FRANKLIN. 2004. Fire-related landform
associations of remnant old-growth trees in the southern Wash-
ington Cascade Range. Can. J. For. Res. 34:2371–2381.
KEETON, W.S., AND J.F. FRANKLIN. 2005. Do remnant old-growth
trees accelerate rates of succession in mature Douglas-fir for-
ests? Ecol. Monogr. 75:103–118.
KEETON, W.S., C.E. KRAFT,AND D.R. WARREN. 2007. Mature and
old-growth riparian forests: Structure, dynamics, and effects on
Adirondack stream habitats. Ecol. Applic. 17:852– 868.
KEITH, H., B.G. MACKEY,AND D.B. LINDENMAYER. 2009. Re-
evaluation of forest biomass carbon stocks and lessons from the
world’s most carbon-dense forests. Proc. Natl. Acad. Sci. USA
106:11635–11640.
LEAK, W.B. 1985. Relationships of tree age to diameter in old-
growth northern hardwoods and spruce-fir. US For. Serv. Res.
Note NE-329.
LORIMER, C.G., AND A.S. WHITE. 2003. Scale and frequency of
natural disturbances in the northeastern U.S.: Implications for
early-successional forest habitats and regional age distribu-
tions. For. Ecol. Manag. 185:41– 64.
Forest Science 57(6) 2011 503

LUYSSAERT, S., E.D. SCHULZE,A.BORNER,A.KNOHL,D.HES-
SENMOLLER, B.E. LAW,P.CIAIS,AND J. GRACE. 2008. Old-
growth forests as global carbon sinks. Nature 455:213–215.
MAINE CRITICAL AREAS PROGRAM. 1985. A summary of old-
growth stands in Maine. Maine Critical Areas Program, Maine
State Planning Office, Augusta, ME. 22 p.
MANION, P.D., AND D.H. GRIFFIN. 2001. Large landscape scale
analysis of tree death in the Adirondack Park, New York. For.
Sci. 47:542–549.
MARTIN, C.W., AND A.S. BAILEY. 1999. Twenty years of change in
a northern hardwood forest. For. Ecol. Manag. 123:253–260.
MCGEE, G.G., D.J. LEOPOLD,AND R.D. NYLAND. 1999. Structural
characteristics of old-growth, maturing, and partially cut north-
ern hardwood forests. Ecol. Applic. 9:1316 –1329.
MCLACHLAN, J.S., D.R. FOSTER,AND F. MENALLED. 2000. An-
thropogenic ties to late-successional structure and composition
in four New England hemlock stands. Ecology 81:717–733.
MCMARTIN, B. 1994. The great forest of the Adirondacks. North
Country Books, Utica, NY.
MCNAB, W.H., D.T. CLELAND, J.A. FREEOUF, J.E. KEYS, G.J.
NOWACKI, G.J., AND C.A. CARPENTER (COMPS.). 2007. De-
scription of ecological subregions: Sections of the contermi-
nous United States. US For. Serv. Gen. Tech. Rep. WO-76B.
MILLWARD, A.A., AND C.E. KRAFT. 2004. Physical influences of
landscape on a large-extent ecological disturbance: The north-
eastern North American ice storm of 1998. Landsc. Ecol.
19:99 –111.
NORTH, M.P., AND W.S. KEETON. 2008. Emulating natural distur-
bance regimes: An emerging approach for sustainable forest
management. P. 341–372 in Patterns and processes in forest
landscapes—Multiple use and sustainable management (La-
fortezza, R., J. Chen, G. Sanesi, and T.R. Crow, eds.). Springer,
Amsterdam, The Netherlands.
NUNERY, J.S., AND W.S. KEETON. 2010. Forest carbon storage in
the northeastern United States: Net effects of harvesting fre-
quency, post-harvest retention, and wood products. For. Ecol.
Manag. 259:1363–1375.
ODUM, E.P. 1969. The strategy of ecosystem development. Science
164:262–270.
OLLINGER, S.V., J.D. ABER, P.B. REICH,AND R.J. FREUDER. 2002.
Interactive effects of nitrogen deposition, tropospheric ozone,
elevated CO
2
, and land use history on the carbon dynamics of
northern hardwood forests. Glob. Change Biol. 8:545–562.
OLLINGER, S.V., C.L. GOODALE,K.HAYHOE,AND J.P. JENKINS.
2008. Potential effects of climate change and rising CO
2
on
ecosystem processes in northeastern U.S. forests. Mitig. Adapt.
Strat. Glob. Change 13:467– 485.
PREGITZER, K.S., AND E.S. EUSKIRCHEN. 2004. Carbon cycling and
storage in world forests: Biome patterns related to forest age.
Glob. Change Biol. 10:2052–2077.
RAY, D.G., M.R. SAUNDERS,AND R.S. SEYMOUR. 2009a. Recent
changes to the Northeast Variant of the Forest Vegetation
Simulator and some basic strategies for improving model out-
puts. North. J. Appl. For. 26:31–34.
RAY, D.G., R.S. SEYMOUR, N.S. SCOTT,AND W.S. KEETON.
2009b. Mitigating climate change with managed forests: Bal-
ancing expectations, opportunity, and risk. J. For. 107(1):
50 –51.
RHEMTULLA, J.M., D.J. MLADENOFF,AND M.K. CLAYTON. 2009.
Historical forest baselines reveal potential for continued carbon
sequestration. Proc. Natl. Acad. Sci. USA 106:6082– 6087.
RUDDELL, S., R. SAMPSON,M.SMITH,R.GIFFEN,J.CATHCART,J.
HAGAN,D.SOSLAND,J.GODBEE,J.HEISSENBUTTEL,S.
LOVETT,J.HELMS,W.PRICE,AND R. SIMPSON. 2007. The role
for sustainably managed forests in climate change mitigation.
J. For. 105:314 –319.
RUNKLE, J.R. 1982. Patterns of disturbance in some old-growth
mesic forests of eastern North America. Ecology 62:
1041–1051.
RUNKLE, J.R. 1991. Gap dynamics of old-growth eastern forests:
Management implications. Nat. Areas J. 2(1):19 –25.
SCHABERG, P.G., D.H. DEHAYES,AND G.J. HAWLEY. 2001. An-
thropogenic calcium depletion: A unique threat to forest eco-
system health? Ecosyst. Health 7:214 –228.
SCHULTE, L.A., AND D.J. MLADENOFF. 2005. Severe wind and fire
regimes in northern forests: Historical variability at the regional
scale. Ecology 86:431– 445.
SEIDL, R., W. RAMMER,D.JAGER, W.S. CURRIE,AND M.J. LEXER.
2007. Assessing trade-offs between carbon sequestration and
timber production within a framework of multi-purpose for-
estry in Austria. For. Ecol. Manag. 248:64 –79.
SEIDL, R., W. RAMMER,P.LASCH, F.W. BADECK,AND M.J.
LEXER. 2008. Does conversion of even-aged, secondary conif-
erous forests affect carbon sequestration? A simulation study
under changing environmental conditions. Silva Fenn. 42:
369 –386.
SEYMOUR, R.S. 1995. The northeastern region. P. 31– 80 in Re-
gional silviculture of the United States, 3rd ed. (Barrett, J., ed.).
John Wiley & Sons, New York.
SEYMOUR, R.S., A.S. WHITE,AND P.G. DEMAYNADIER. 2002.
Natural disturbance regimes in northeastern North America:
Evaluating silvicultural systems using natural scales and fre-
quencies. For. Ecol. Manag. 155:357–367.
SHIVERS, B.D., AND B.E. BORDERS. 1996. Sampling techniques for
forest inventory. John Wiley & Sons, New York. 356 p.
SICCAMA, T.G., T.J. FAHEY, C.E. JOHNSON, T.W. SHERRY, E.G.
DENNY, E.B. GIRDLER, G.E. LIKENS,AND P.A. SCHWARZ. 2007.
Population and biomass dynamics of trees in a northern hard-
wood forest at Hubbard Brook. Can. J. For. Res. 37:737–749.
SICCAMA, T.G., S.P. HAMBURG, M.A. ARTHUR, R.D. YANAI, F.H.
BORMANN,AND G.E. LIKENS. 1994. Corrections to allometric
equations and plant tissue chemistry for Hubbard Brook Ex-
perimental Forest. Ecology 75:246 –248.
SMITH, J.E., L.S. HEATH, K.E. SKOG,AND R.A. BIRDSEY. 2006.
Methods for calculating forest ecosystem and harvested carbon
with standard estimates for forest types of the United States.US
For. Serv. Gen. Tech. Rep. NE-343.
SMITH, M.L., AND C. CARPENTER. 1996. Application of the USDA
Forest Service National Hierarchical Framework of Ecological
Units at the sub-regional level: The New England-New York
example. Environ. Monitor. Assess. 39:187–198.
STOVALL, J., W.S. KEETON,AND C.E. KRAFT. 2009. Late-succes-
sional riparian forest structure results in heterogeneous peri-
phyton distributions in low-order streams. Can. J. For. Res.
29:2343–2354.
STOY, P.C., G.G. KATUL, M.B.S. SIQUEIRA,J.JUANG,J.NOVICK,
H.R. MCCARTHY, A.C. OISHI,AND R. OREN. 2008. Role of
vegetation in determining carbon sequestration along ecologi-
cal succession in the southeastern United States. Glob. Change
Biol. 14:1409 –1427.
THORNLEY, J.H.M., AND M.G.R. CANNELL. 2000. Managing for-
ests for wood yield and carbon storage: A theoretical study.
Tree Physiol. 20:477– 484.
TURNER, D.P., G.J. KOERPER, M.E. HARMON,AND J.J. LEE. 1995.
A carbon budget for forests of the conterminous United States.
Ecol. Applic. 5:421– 436.
TWERY, M.J., P.D. KNOPP, S.A. THOMASMA, H.M. RAUSCHER,
D.E. NUTE, W.D. POTTER,F.MAIER,J.WANG,M.DASS,H.
UCHIYAMA,A.GLENDE,AND R.E. HOFFMAN. 2005. NED-2: A
504 Forest Science 57(6) 2011

decision support system for integrated forest ecosystem man-
agement. Comput. Electron. Agric. 49:24 – 43.
TYRRELL, L.E., AND T.R. CROW. 1994. Structural characteristics of
old-growth hemlock-hardwood forests in relation to stand age.
Ecology. 75:370 –386.
TYRRELL, L.E., G.J. NOWACKI, T.R. CROW, D.S. BUCKLEY, E.A.
NAUERTZ, J.N. NIESE, J.L. ROLLINGER,AND J.C. ZASADA.
1998. Information about old-growth for selected forest type
groups in the eastern United States. US For. Serv. Gen. Tech.
Rep. NC-197.
VAN WAGNER, C.E. 1968. The line-intercept method in forest fuel
sampling. For. Sci. 14:20 –26.
WARREN, D.R., C.E. KRAFT, W.S. KEETON, J.S. NUNERY,AND
G.E. LIKENS. 2009. Dynamics of wood recruitment in streams
of the northeastern U.S. For. Ecol. Manag. 258:804 – 813.
WARREN, W.G., AND P.F. OLSEN. 1964. A line intersect technique
for assessing logging waste. For. Sci. 13:267–276.
WHITMAN, A.A., AND J.M. HAGAN. 2007. A late-successional
index for temperate and boreal forest. For. Ecol. Manag.
246:144 –154.
WHITTAKER, R.H., F.H. BORMANN, G.E. LIKENS,AND T.G. SIC-
CAMA. 1974. The Hubbard Brook ecosystem study: Forest
biomass and production. Ecol. Monogr. 44:233–252.
WOODS, K.D. 2004. Intermediate disturbance in a late-succes-
sional hemlock-northern hardwood forest. Ecology 92:
464 – 476.
WOODS, K.D., AND C.V. COGBILL. 1994. Upland old-growth for-
ests of Adirondack Park, New York, USA. Natur. Areas J.
14:241–257.
ZIEGLER, S.S. 2000. A comparison of structural characteristics
between old-growth and postfire second-growth hemlock-hard-
wood forests in Adirondack park, New York, U.S.A. Glob.
Ecol. Biogeog. 9:373–389.
ZIEGLER, S.S. 2002. Disturbance regimes of hemlock-dominated
old-growth forests in northern New York, U.S.A. Can. J. For.
Res. 32:2106–2115.
Forest Science 57(6) 2011 505