Content uploaded by R. Scot Duncan
Author content
All content in this area was uploaded by R. Scot Duncan on Feb 12, 2018
Content may be subject to copyright.
Consequences of plantation harvest during tropical forest
restoration in Uganda
R. Scot Duncan
a,*
, Colin A. Chapman
a,b
a
Department of Zoology, P.O. Box 118525, University of Florida, Gainesville, FL 32611, USA
b
Wildlife Conservation Society, 185th Street and Southern Boulevard, Bronx, NY 10460, USA
Received 1 June 2001; accepted 26 December 2001
Abstract
Timber plantations have recently received considerable attention as a forest restoration strategy for heavily degraded lands in
the humid tropics. Plantations can facilitate secondary forest regrowth by providing an understory environment more favorable
for native plant recruitment than unmanaged degraded habitats. To better understand secondary forest development and to
evaluate plantation use as a restoration strategy, we studied natural forest succession after plantation harvest in Kibale National
Park, Uganda. We asked four questions concerning using plantations as a restoration tool. How does timber harvest affect forest
succession? How does initial on-site recruit availability (e.g., seeds, seedlings) after logging in¯uence successional pathways?
How easily can forest succession be enhanced through intervention? How does using exotic timber plantations to restore forests
compare with ®re exclusion as a restoration strategy? Timber harvesting killed or severely damaged many native stems,
hindering subsequent forest regrowth. Despite this setback, native stem densities 4±6 years after logging equaled or surpassed
native stem densities in unlogged plantations, suggesting timber removal accelerated forest succession. Successional habitats
with high and low initial densities of on-site recruits ®rst diverged in forest structure and composition, but then converged for
many of these variables within 6 years of logging. Intervening to accelerate forest succession met with mixed results. Removing
non-tree vegetation did not enhance tree establishment, growth, or survival after 2 years. However, leaving standing, dead timber
trees as perches for seed-dispersing birds seemed to increase seedling establishment relative to control areas. Mortality and
growth of seedlings planted into successional habitats 1±2 and 5±6 years after logging were similar, and predicting individual
species responses based on successional status was unsuccessful. We compared succession in unlogged and logged (5±6 years
after logging) plantations to a similar aged site where ®res were excluded but no plantation species established. Our results
suggest excluding ®re is a better strategy for promoting forest succession than establishing then not harvesting plantations. Fire
exclusion versus establishing then harvesting timber are comparable restoration strategies differentially enhancing tree sapling
recruitment and growth, respectively. While forest regeneration was successful where ®re was excluded long-term ®re exclusion
may be dif®cult, and several non-ecological challenges to using plantations exist (e.g., the con¯ict of managing for biodiversity
versus timber production). In summary, our research suggests managers should carefully weigh the risks of using plantations or
®re exclusion against other forest restoration strategies.
#2002 Elsevier Science B.V. All rights reserved.
Keywords: Arrested succession; Exotic tree plantations; Reforestation; Regeneration; Restoration ecology; Secondary forests; Tropical forest
succession
Forest Ecology and Management 173 (2003) 235±250
*
Corresponding author. Fax: 352-392-3704.
E-mail address: duncan@zoo.u¯.edu (R.S. Duncan).
0378-1127/02/$ ± see front matter #2002 Elsevier Science B.V. All rights reserved.
PII: S 0378-1127(02)00009-9
1. Introduction
Forest regrowth is often slow or absent on heavily
degraded lands in the tropics (Lamb, 1998). However,
restored forests on these lands can be useful for con-
servation, timber production, carbon-dioxide seques-
tration and can provide other ecosystem services
(Brown and Lugo, 1990, 1994). Establishing timber
plantations on degraded lands can facilitate forest
succession by providing an understory environment
favorable for native plant recruitment (Chapman and
Chapman, 1996; Lugo, 1997; Parrotta et al., 1997). An
additional bene®t of plantations is that timber sales can
help pay for the restoration. While plantations have
attracted the attention of many tropical restoration
ecologists, there has been little investigation of how
best to use plantations in restoration (though see Ashton
et al., 1998), how plantation harvest affects subsequent
forest regrowth, or how using plantations compares to
alternative restoration strategies. Furthermore, our
understanding of forest succession is poor (Chazdon,
1994), thus limiting our ability to manage secondary
forests after plantation harvest.
We studied forest regrowth after the harvesting of
exotic timber plantations in Kibale National Park,
Uganda. These plantations were planted in the
1950±1960s on previously forested lands that had
been used by agriculturalists, but had been abandoned
and were dominated by grasses (especially Pennise-
tum purpureum) (Osmaston, 1959). Active ®re exclu-
sion was important to protecting young plantations,
but became less necessary as plantations matured. As
plantations matured, native trees established in their
understory and were not removed by managers (Chap-
man and Chapman, 1996; Fimbel and Fimbel, 1996).
After logging, surviving native stems, the soil seed
bank, and seed dispersal contributed to secondary
forest growth. In contrast, similar areas not planted
with exotics are still ®re-maintained grassland, con-
®rming plantations can facilitate forest regrowth on
degraded lands (Duncan and Duncan, 2000; Zanne
and Chapman, 2001).
Two successional trajectories emerge from planta-
tion logging at Kibale. As a result of differential native
stem recruitment before harvesting, logged pine plan-
tations (Pinus caribeae and P. patula) have greater
initial on-site recruit densities (e.g., more native
stems) than do logged cypress (Cupressus lusitanica)
plantations harvest (Fimbel and Fimbel, 1996). Logged
pine plantations are dominated by early- and mid-
successional trees, while logged cypress plantations
are dominated by early- and mid-successional shrubs
and trees. The juxtaposition of disturbed habitats with
varying initial recruit availability is common among
degraded lands in the tropics (Uhl et al., 1982, 1988a;
Adedeji, 1984; De Rouw, 1994; Garciamontiel and
Scatena, 1994; Fernandes and Sanford, 1995; Gillespie
et al., 2000). Typically, more heavily degraded sites
start succession with lower recruit availability than do
less degraded sites (Brown and Lugo, 1990, 1994).
Understanding how forest succession proceeds given
different levels of initial recruit availability has not been
well studied. However, such knowledge would assist
managers choosing among disturbance types for invest-
ing restoration efforts. Or, this knowledge could be
useful for managers predicting when successional for-
ests will provide extractive resources or adequate habi-
tat for species of interest.
We investigated four questions relevant to mana-
ging plantations for restoring forests and managing
successional forests. First, how does initial on-site
recruit availability in¯uence subsequent successional
trajectories? We compared the ®rst 6 years of succes-
sion between logged cypress and pine plantations.
Second, how does timber harvesting affect ensuing
forest successions? We investigated how long a forest
takes to regain the structure and composition present
in the plantation understory before harvest, the
damage in¯icted on native stems during harvest, the
effects of felling versus felling and timber removal,
and the effect of variation in disturbance intensity
within and among logged plantations. Third, how easily
can forest succession be enhanced through interven-
tion? In particular, we examined whether removal of all
non-tree vegetation enhanced tree recruitment and
growth, the success of planting tree and shrub seedlings
into successional habitats of two ages, and leaving dead
remnant trees in successional habitats to enhance seed-
ling recruitment by attracting seed-dispersing birds.
Fourth, how does using exotic timber plantations to
restore forests compare with ®re exclusion as a restora-
tion strategy? Recurring ®re is a severe impediment to
forest succession on many degraded lands, especially
grasslands (Uhl et al., 1988b; Uhl and Kauffman,
1990). We contrasted forest succession among logged
and unlogged plantations and a formergrassland where
236 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
®re was excluded since the time of plantation esta-
blishment. Finally, we evaluate pros and cons of using
timber plantations to restore forests in light of our
®ndings.
2. Methods
2.1. Study site
Located on the foothills of the Ruwenzori Moun-
tains, Kibale National Park, Uganda, is dominated by
moist-evergreen forest. Average elevation is 1500 m,
and average rainfall is 1543 mm yr
1
(1903±1999),
though it has recently increased to 1765 mm yr
1
(1991±1999). Softwood plantations in the park are
surrounded by natural forest (plantation interiors are
<250 m from natural forest) and have been described
by Chapman and Chapman (1996), Fimbel and Fimbel
(1996), and Zanne and Chapman (2001). Plantation
logging began in 1993 when Kibale became a national
park, and current management is allowing natural
forests to grow in their place. From 1998 to 2000
we studied forest succession among plantations
logged between 1993 and 2000.
Mature cypress plantations have lower mean basal
areas than do pine plantations (44 m ha
1
versus
71 m ha
1
, respectively), shorter canopies than pine
plantations (23 m versus 33 m, respectively), but simi-
lar mean DBHs relative to pine plantations (36 cm for
both; Fimbel and Fimbel, 1996). Timber is felled with
chainsaws, and this results in many native stems being
killed or damaged. Felled timber is rolled or winched
to nearby portable sawmills, pitsawing stations, or
roadsides. During this movement, more native stems
are killed or damaged. When harvest is completed, few
stems >1 m tall remain. However, many native seed-
lings survive, and root sprouts and coppice from native
species are common. There is no regeneration of
plantation species. Because unlogged pine plantations
have more native stems than unlogged cypress planta-
tions, hereafter, we refer to them as ``high density''
and ``low density'' plantations, respectively.
2.2. In¯uence of initial recruit availability
To assess how initial recruit availability in¯uences
succession, we quanti®ed how vegetation differed
between the only remaining unlogged low density
plantation and an adjacent unlogged high density
plantation (thus attempting to control for site history
and landscape position). Within each plantation, 15
plots were positioned by randomly choosing distances
along a 100 m transect, then randomly choosing a
distance of 1±60 m perpendicular to the transect. Plots
were non-overlapping and within 100 m of natural
forest edge. The openness of plantation understories
allowed use of 25 m
2
circular plots, which were faster
to survey than rectangular plots used in logged planta-
tions. In these and all plots surveyed in logged planta-
tions, we measured density, species richness, and
height of tree stems 1.0 m tall. These stems were
termed ``saplings'', though a few early-successional
trees were mature. Smaller ht 0:25mand inter-
mediate stems ht 0:260:99 mwere described in
4m
2
subplots. We termed these smaller stems ``seed-
lings'' and intermediate stems as ``large seedlings''
though stems of both groups likely were no longer
dependent on seed reserves or may have been sprouts.
Plant identi®cation was based on manuals by Eggeling
and Dale (1952), Polhill (1952), Hamilton (1991) and
Katende et al. (1995).
We established long-term plots in a low density and
high density plantation within a year after logging (9
and 4 months, respectively). These were the only
similarly aged recently logged high and low density
plantations available. Thirty plots 5m5min each
logged plantation were placed randomly (non-over-
lapping) along parallel transects (10 m apart) covering
the logged area. All plots were within 100 m of natural
forest edge. Plots were marked, surveyed, and
surveyed again after 2 years; one plot in the logged
high density plantation was destroyed by human
activity before the second survey. During both sur-
veys, vegetation was quanti®ed as done in unlogged
plantations. Some plots in the logged plantations were
heavily disturbed areas (pit-sawing stations, portable
sawmill locations and roads) where soil and vegetation
were more disturbed than in moderately disturbed
plots where the main activity was timber removal.
Heavily disturbed plots were excluded from the pre-
sent analyses, but are considered below. Many stems
(23%) classi®ed as seedlings in the ®rst year after
logging were actually root sprouts or coppice from
stems buried during logging (Duncan, unpublished
data).
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 237
Finally, we surveyed the oldest successional habi-
tats in logged plantations at Kibale (4±6 years after
logging). Vegetation plots were surveyed in two low
density plantations, one logged 4 years before (n30
plots) and another logged 5 years before (n28
plots). Thirty plots were surveyed in a high density
plantation logged 5±6 years before. In these planta-
tions, plot positioning and sampling followed that of
long-term plots. We used 2 m 10 m plots with 4 m
2
subplots for smaller stem to ease sampling of the dense
vegetation. Plots from the two logged low density
plantations were combined for comparisons to plots in
the logged high density plantation.
We used non-parametric analyses for these and
most comparisons because data were often highly
zero-skewed. Consequently, we present medians to
describe data, but also provide means to aid compar-
isons with other studies where means are reported.
2.3. Harvesting effects on forest succession
2.3.1. Native vegetation before and after timber
harvest
From the surveys described above, we compared
native vegetation present before logging to that at <1
and 4±6 years after logging in moderately disturbed
plots. High and low density plantations were analyzed
separately to control for differences in initial recruit
availability.
2.3.2. Stem damage from logging
The process of felling and removing timber can
severely damage native vegetation. During plot sam-
pling in recently logged plantations, we classi®ed
surviving native stems into four categories: (1) unda-
maged, (2) bent stems pinned to or near the ground by
felled trees, (3) broken stems broken near their base and
(4) cut stems cut at their base. Some stems fell in
multiple damage categories. We pooled all plots (heav-
ily and moderately disturbed, high and low density
plantations) and compared numbers of damaged stems
to undamaged stems.
2.4. Damage from felling
When plantations are managed to promote natural
forest regrowth, managers may want to develop har-
vest methods reducing damage to native stems and
encouraging regeneration. Thus, it is important to
know how damage during felling compares to that
during timber removal. For reasons unknown to us,
one section of high density plantation (0.4 ha) was
felled and the timber left in place. We sampled this
area 3 years after timber was felled (felled-only plots).
Plot positioning and sampling followed that in logged
plantations. Plots n20were 2 m 10 m, with
4m
2
subplots for sampling stems <1 m tall. We com-
pared this habitat to the high density plantation logged
5±6 years previously (only moderately disturbed
plots), the most similarly aged logged high density
plantation.
2.4.1. Within-site disturbance intensity
Damage to native stems during logging varies spa-
tially, and depends on intensity of harvest activity
(e.g., portable sawmill locations versus adjacent
logged areas). We pooled plots from the two recently
logged plantations and compared stems from plots on
moderately n49and heavily disturbed n10
sites.
We also compared effects on forest succession of
two timber processing methods used in the same
plantation. In one part of the high density plantation
logged 6 years previously, felled timber had been
winched or wheeled with a gurney to the roadside,
then transported to a sawmill. In the other part of the
plantation, timber was cut into logs and rolled to
nearby pitsawing stations (each station 100 m
2
).
At pitsawing stations, logs were rolled onto scaffold-
ing above a pit. Then, two men, one on the log and the
other in the pit, used a large handsaw to cut the log into
boards. Where pitsawing was used to process timber,
stations cover 20±30% of the area. Left at each station
was a thick layer of sawdust, compacted soil, and
timber scraps potentially hindering forest regrowth. In
addition, scaffolding was often constructed from
native tree saplings cut from the plantation or adjacent
forest. On the other hand, while pitsawyers felled all
timber, only larger stems were processed, potentially
leaving patches where native stems were less damaged
than where all stems were extracted for millsawing.
We compared how these two timber processing meth-
ods affected forest regrowth by comparing vegetation
plots that had been randomly placed into either pit-
sawn or millsawn parts of the plantation (n16 and
10 plots, respectively).
238 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
2.5. Facilitating forest succession
2.5.1. Vegetation removal experiment
Resource competition among plants in early-suc-
cessional habitats can restrict recruitment, growth, and
survival of tree species (Holl, 1998a). To evaluate this,
we removed all non-tree vegetation from 5 m 5m
randomly placed plots in the recently logged low
density plantation (5 months after logging; n22
plots) and high density plantation (10 months after
logging; n20 plots). Approximately, every 4
months (mean S:D:4:11:5 months), plots
were cleared of all aboveground non-tree vegetation.
Larger stems ht00:5mwere cut at the base, while
smaller stems were pulled from the ground. Plots were
surveyed when ®rst cleared and 2 years later. During
surveys, all tree saplings were identi®ed, counted and
their heights measured; stems <1 m tall were surveyed
in a 2 m 2 m subplot in the plot center. Vegetation
removal plots were compared to the vegetation plots of
the same successional age in the same plantation (con-
trol plots; no heavily disturbed plots included). Height,
density, and species richness of trees were compared
between removal and control plots for initial and ®nal
surveys. Seedling density and species richness were
compared between treatments for ®nal surveys.
2.5.2. Seedling planting experiment
Planting seedlings into successional habitats has
been suggested to be an effective strategy for over-
coming dispersal limitation and improving species
composition or structure of developing forests (Tucker
and Murphy, 1997; Parrotta and Knowles, 1999). We
contrasted success of nursery-raised seedlings trans-
planted into two plantations logged 7±20 and 63±75
months earlier (hereafter, ``younger'' and ``older'' site,
respectively). Initially, the younger site was open with
little vegetation above 0.25 m, but by the end of the
experiment, it had a dense, short canopy (1.5 m),
composed of herbaceous growth, shrubs and trees.
Thus, light availability shifted from direct to indirect
during the experiment. The older site's canopy was
3±7 m tall, composed of tree saplings and shrubs;
usually only indirect sunlight and sun ¯ecks reached
the understory.
Seedlings of local tree and shrub species were
selected based on seed availability and nursery germi-
nation success. Seeds of each species were planted in
nursery beds in full sun or below a short (1 m) thatch
roof. Planted seeds were watered every few days (unless
it had rained) until seedlings were transplanted, usually
2±4 weeks after germination. Within a species, all
transplanted seedlings had the same number of coty-
ledons and leaves, and were roughly the same height.
After planting, seedlings were watered once (200 ml)
to reduce transplanting mortality. When possible, seed-
lings germinating in full sun were planted into the
younger site (65% of the time), and seedlings germinat-
ing in shade were planted into the older site (94% of the
time). When few seedlings germinated from one of the
light treatments, seedlings were taken from the other
light treatment for both successional habitats (e.g.,
shade-germinated seedlings planted into the younger
site). The length of time that transplanted seedling
species were monitored varied because timing of ger-
mination varied. Altogether, 33 species were planted
into the nursery, only 17 species produced enough
seedlings for experimentation.
Seedlings were planted into the two successional
habitats during the same or subsequent days. In the
younger site, seedlings were planted along ten 50 m
transects placed parallel to one another and separated
by 10 m; seedlings stations were at 1 m intervals.
Conspeci®cs were planted with 15 m minimum dis-
tance between them on transects. A similar design was
used in the old successional habitat, but we used
variable length parallel transects, each separated by
>5 m (shorter distance due to area constraints).
The proportion of seedlings surviving to the end of
the experiment was calculated for each species, and
the proportions from all species were compared
between treatments. Within each species, proportions
of surviving seedlings were also compared between
treatments. For each seedling surviving to the end of
the experiment, relative growth rate (RGR) was cal-
culated by subtracting the natural log of initial seed-
ling height from the natural log of ®nal seedling
height, and dividing this difference by the number
of days growth was monitored. In this experiment,
RGR is a product of intrinsic and extrinsic (e.g.,
herbivory) factors. To look for overall growth differ-
ences between treatments, a mean RGR was calculated
for each species in each treatment, and means were
compared between younger and older sites.Within each
species, RGRs of individual seedlings were compared
between treatments. Survival proportions and RGRs of
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 239
early- versus mid- and late-successional species were
also compared (the latter two categories were lumped
due to low numbers of late-successional species). These
designations were based on Eggeling and Dale (1952),
Polhill (1952), Hamilton (1991), Katende et al. (1995),
and our own observations.
2.5.3. Recruitment below snags
In the recently logged low density plantation, many
dead cypress trees remained standing (cause of death
unknown). To determine whether these snags attracted
seed dispersers and facilitated seedling recruitment, we
surveyed seedlings below snags and in adjacent open
areas at 17±19 months after logging. Twenty snags
ht 10 mwere randomly chosen from 50 candi-
dates, and a 2 m
2
plot was placed below the tallest part
of each. A control plot was placed in a random direction
10 m from the snag in an area without snags or emer-
gent trees. In both plots, all tree and shrub stems were
identi®ed, counted and their heights were measured.
2.6. Alternatives to plantationsЮre exclusion
Within one of the high density plantations is a
section (1.7 ha) where pine seedlings were either
not planted, or died soon after planting for unknown
reasons (possibly drought or ®re according to B.
Kisembo (Uganda Forestry Department) who helped
to establish and maintain the plantations, pers.
comm.). We believe seedling death was not due to
edaphic factors, since this plot was surrounded by
mature high density plantation and mature forest. This
site and the adjacent plantations were ®re-maintained
grasslands when the pines were planted in the 1950s
and 1960s. As the surrounding plantations matured,
the maintenance of ®re-breaks and reduction of
grasses (probably due to shading) in plantation unders-
tories reduced or stopped ®re occurrence, and a well-
developed successional forest now occupies the site.
This represents a case study of succession following
®re exclusion on degraded grassland (hereafter, ``®re-
excluded site'').
Twenty-®ve 10 m 2 m plots were placed and
surveyed in the ®re-excluded site following the meth-
ods used in surveys of logged plantations. The
obvious control for assessing how effective ®re exclu-
sion is as a restoration tool are Kibale grass lands never
planted with timber species. While not resurveyed for
this study, these areas are still ®re-dominated grass-
lands where tree establishment and growth is minimal
(Duncan and Duncan, 2000; Zanne and Chapman,
2001).
Given that ®re exclusion appears to promote forest
succession, we wanted to compare it to the use of
plantations as a forest restoration tool. Because estab-
lishing plantations and not harvesting them is a pos-
sible restoration tool, we compared successional forest
in the ®re-excluded site to that in the unlogged high
density plantation. Because establishing plantations
and later harvesting them is another restoration strat-
egy, we compared the ®re-excluded site to the high
density plantation surveyed 5±6 years after logging.
Like many large-scale studies, the spatial replication
of this design is limited. However, given that all three
sites share a similar landscape position and site history
prior to plantation establishment, this was a rare
opportunity to compare ®re exclusion and plantation
establishment as restoration strategies.
3. Results
3.1. In¯uence of initial recruit availability
Tree seedling density and species richness were
similar between the unlogged high and low density
plantations, but the high density plantation had sig-
ni®cantly greater tree sapling density (16more) and
species richness (7more) than did the low density
plantation (Table 1). In the high density plantation,
tree saplings were signi®cantly taller (3more) than
in the low density plantation.
Within a year of logging, tree seedling density and
species richness were similar between the high and
low density plantations (Table 1). The high density
plantation still had signi®cantly more tree saplings
(16more) and sapling species (8more). Tree
saplings were still signi®cantly taller (2more) in
the high density than in the low density plantation.
Four to six years after logging, tree seedling density
and species richness were still similar between planta-
tion types (Table 1). Tree sapling density and species
richness in the low density plantation were now not
different from those in the high density plantation.
Tree sapling heights in low density plantation were
similar to those in the high density plantation.
240 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
3.2. Harvesting effects on forest succession
3.2.1. Native vegetation before and after timber
harvest
In the high density plantation, trees were signi®-
cantly or marginally more (2more) dense, species-
rich and taller before harvest than <1 year after harvest
(Table 2). However, tree seedlings were signi®cantly
more dense and species-rich after logging than they
were before. The high density plantation harvested 5±
6 years previously had signi®cantly more tree seed-
lings (6more) and tree seedling species (2more)
than did the unlogged high density plantation
(Table 2). Tree saplings were signi®cantly more dense
(13more) in the unlogged than in the logged planta-
tion, though tree sapling species richness and height
were similar between plantations.
In low density plantations, stem variables were
similar before and within 1 year of logging, except
tree seedlings were signi®cantly more (3more)
Table 1
Median (top number) and mean (bottom number S:D:) values for trees in unlogged and logged high and low density plantations
a
Stem variable Unlogged 6 Months 4±6 Years
High density PLow density High density PLow density High density PLow density
Seedling density
(stems m
2
)
0.3 0.3 0.8 0.8 1.5 1.0
0.4 (0.5) 0.3 (0.4) 1.3 (1.3) 1.7 (2.8) 2.4 (3.5) 2.4 (3.1)
Seedling species
richness (species m
2
)
0.3 0.3 0.5 0.5 0.5 0.5
0.3 (0.3) 0.3 (0.3) 0.6 (0.6) 0.5 (0.5) 0.6 (0.5) 0.6 (0.4)
Sapling density (stems m
2
) 0.6
***
<0.1 0.3
***
<0.1 0.5 0.5
0.8 (0.4) 0.1 (0.2) 0.5 (0.4) 0.1 (0.2) 0.6 (0.4) 0.7 (0.6)
Sapling species richness
(species m
2
)
0.3
***
<0.1 0.2
***
<0.1 0.3 0.3
0.3 (0.1) 0.1 (0.1) 0.2 (0.1) 0.1 (0.1) 0.3 (0.1) 0.3 (0.2)
Sapling height (m) 2.9
***
1.1 1.5
***
0 5.5 5.0
2.9 (0.4) 1.3 (1.5) 1.7 (0.6) 0.7 (1.1) 6.1 (3.3) 5.6 (4.0)
a
Results of Mann±Whitney contrasts of stem variables between the plantation types within a successional age are also presented: (
)
P0.01, (
)P0.05, (
)P0.10.
***
P0:001.
Table 2
Results (Pvalues) of Mann±Whitney comparisons of stem variables between the plantation types within a successional age are also presented
when signi®cant: P0.001
a
Stem variable Unlogged versus 6 months after logging Unlogged versus 6 years after logging
High density Low density High density Low density
Seedling density (stems m
2
) 0.014 0.042 0.007 <0.001
L>UL L >UL L >UL L >UL
Seedling species richness (species m
2
) 0.051 0.112 0.041 0.008
L>ULL>UL L >UL
Sapling density (stems m
2
) 0.010 0.689 0.047 <0.001
UL >LUL>LL>UL
Sapling species richness (species m
2
) <0.001 0.450 0.744 <0.001
UL >LL>UL
Sapling height (m) <0.001 0.390 0.001 <0.001
UL >LL>UL L >UL
a
Directions of signi®cant or marginally insigni®cant (in parentheses) differences are indicated below Pvalues (see Table 1 for tree variable
values).
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 241
dense after than before logging (Table 2). Low density
plantations logged 4±6 years before had signi®cantly
greater tree seedling density (4greater) and species
richness (2greater), and tree sapling density (13
greater), species richness (8greater), and height (4
greater), than did the unlogged low density plantation
(Table 2).
3.2.2. Stem damage from logging
During surveys, it was impossible to determine
whether stems had survived logging or established
after logging. Signi®cantly more stems found during
initial surveys were undamaged than damaged for all
size classes (80% and 20%, respectively; Table 3). Tree
saplings generally had more damage than did smaller
size classes, probably because many of the smaller
stems recruited after logging. Nearly one-third of tree
saplings found had been damaged during logging. One
pioneer tree species, Trema orientalis, only established
after logging. When these stems were excluded, nearly
half of all tree saplings showed damage (Table 3).
3.2.3. Damage from felling
Tree seedling density and species richness were
similar between the felled-only site and the high density
site logged 5±6 years previously (Table 4). There were
non-signi®cant trends for greater tree sapling density
and species richness in the felled-only than logged site,
but there was a trend for taller tree saplings in the logged
than felled-only site. These results suggests that most
damage to native stems occurs during timber felling,
rather than timber extraction.
3.2.4. Within-site disturbance intensity
Signi®cantly more tree seedlings (2more) and
tree seedling species (2more) were found in mod-
erately disturbed than in heavily disturbed sites <1
year after logging (high density and low density
plantations; Table 5). For all other stem variables,
no differences existed between plot types.
When we compared regeneration between millsawn
and pitsawn areas in the 6-year-old logged low density
plantation, signi®cantly more tree saplings (3more)
and marginally more sapling species (1.5more)
were found in millsawn plots. Tree seedling density
and species richness, and tree sapling height were
similar between the two areas.
3.3. Facilitating forest succession
3.3.1. Vegetation removal experiment
Initial height, density, and species richness of tree
saplings were similar between plots where all non-tree
Table 3
Percentages of trees with and without damage within 1 year of logging and extraction of timber species
a
Stem size class Bent Broken Cut % total damaged stems % undamaged stems Total stems P
Seedling 3.2 3.5 0.3 7.0 93.0 (87.1) 345 <0.001
Large seedling 3.7 12.0 3.3 19.8 80.2 (74.6) 242 <0.001
Sapling 11.7 7.3 11.3 32.3 67.7 (57.6) 452 0.002
a
Three damage categories, total percent damaged, percent undamaged stems, and total stem number are presented. Also results of
Wilcoxon signed rank tests comparing for each category, the total damaged and undamaged stems in plots are also presented. Percentages of
undamaged stems when T. orientalis is excluded are presented in parentheses.
Table 4
Median (top number) and mean (bottom number S:D:) for stem
variables from felled-only plots
a
Stem variable Median and
mean (S.D.),
felled-only
P
Seedling density (stems m
2
) 1.1 0.669
1.8 (2.6)
Seedling species richness
(species m
2
)
0.5 0.362
0.5 (0.2)
Sapling density (stems m
2
) 0.7 0.081
0.7 (0.3) F>L
Sapling species richness
(species m
2
)
0.3 0.054
0.3 (0.1) F>L
Sapling height (m) 3.8 0.068
4.3 (1.4) L>F
a
Stem variables from felled-only plots (F) were compared to
those from a logged plantations where timber was felled and
extracted (L). Directions of signi®cant or marginally insigni®cant
(in parentheses) differences are indicated below Pvalues (Mann±
Whitney tests; see Table 1 for tree stem values for unlogged and
logged plantations).
242 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
vegetation were removed and control plots in both the
logged low and high density plantations (Mann±Whit-
ney test, P>0:10 for all contrasts; Table 6). After 2
years, height, density, and species richness of tree
sapling and seedling variables in experimental plots
were still similar to those in control plots of either
successional habitat (Table 6).
3.3.2. Seedling planting experiment
The average proportion of seedlings surviving to the
end of the experiment was not signi®cantly different
between younger and older habitats (t-test paired by
species, t0:8;P0:458, proportions were arcsine-
square root transformed). Within either the older or
younger habitat, there was no difference in proportion
Table 5
Median (top number), mean (bottom number S:D:) and Mann±Whitney comparisons of stem variables from plots in heavily and moderately
disturbed areas in high density and low density plantations <1 year after logging, and from millsawn and pitsawn sites within one low density
plantation 6 years after logging
Stem variable Heavily
disturbed plots
Moderately
disturbed plots
PMillsawn
plots
Pitsawn
plots
P
Seedling density (stems m
2
) 0 0.8 0.042 2.3 0.9 0.232
0.9 (2.3) 1.5 (2.2) 3.7 (3.7) 1.7 (2.3)
Seedling species
richness (species m
2
)
0 0.50 0.037 0.8 0.5 0.771
0.25 (0.41) 0.57 (0.53) 0.7 (0.4) 0.6 (0.4)
Sapling density (stems m
2
) 0.2 0.2 0.984 0.6 0.2 0.014
0.2 (0.2) 0.3 (0.4) 0.7 (0.5) 0.3 (0.3)
Sapling species
richness (species m
2
)
0.1 0.1 0.652 0.3 0.2 0.062
0.1 (0.1) 0.1 (0.1) 0.3 (0.1) 0.2 (0.2)
Sapling height (m) 1.4 1.3 0.489 3.4 8.5 0.280
1.6 (1.2) 1.3 (1.0) 4.2 (2.2) 7.9 (6.0)
Table 6
Median (top number) and mean (bottom number S:D:) values for trees in plots cleared of all non-tree vegetation versus unmanipulated
controls
a
Plantation
and year
Tree size
class
Height (m) Density (stems m
2
) Species richness (species m
2
)
Control Cleared Control Cleared Control Cleared
Logged low density plantation
Year 1 Saplings 0 0.5 <0.1 <0.1 <0.1 <0.1
0.7 (1.1) 0.8 (1.3) 0.1 (0.2) 0.1 (0.1) 0.1 (0.1) 0.1 (0.1)
Year 3 Seedlings ± ± 0 0 0 0
0.2 (0.2) 0.2 (0.5) 0.1 (0.2) 0.1 (0.2)
Saplings 1.6 1.7 0.4 0.3 0.2 0.1
1.7 (1.1) 1.6 (1.0) 0.5 (0.5) 0.4 (0.3) 0.2 (0.1) 0.2 (0.1)
Logged high density plantation
Year 1 Saplings 1.5 1.5 0.3 0.4 0.2 0.2
1.7 (0.6) 1.5 (0.4) 0.5 (0.4) 0.8 (1.2) 0.2 (0.1) 0.2 (0.1)
Year 3 Seedlings ± ± 0 0.1 0 0.1
0.2 (0.4) 0.4 (0.7) 0.2 (0.3) 0.2 (0.3)
Saplings 3.6 3.2 0.8 0.7 0.3 0.2
3.7 (1.2) 3.3 (1.2) 0.9 (0.6) 1.0 (1.0) 0.3 (0.1) 0.2 (0.1)
a
Treatments were compared with Mann±Whitney tests, but no differences (P>0:10 for all contrasts) were found.
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 243
of seedlings surviving between early-successional
versus mid- and late-successional species (Mann±
Whitney test, P0:118 and 0.590, respectively;
Table 7). When proportions surviving were compared
within species, two early-successional species (Mae-
sopsis eminii and Polyscias fulva) had signi®cantly
more seedlings surviving (10and 5more, respec-
tively) in the older habitat, and two others (Securinega
virosa and T. orientalis) had signi®cantly more seed-
lings surviving (11more for S. virosa;noT. orien-
talis survived in the older habitat) in the younger
habitat (Table 7). Two other early-successional species
(Erythrococca trichogyne and Maesa lanceolata)
showed trends for greater survival (2and 7greater,
respectively) in the younger habitat, while a third
species (Prunus africana) showed a trend for greater
survival (1.1greater) in the older habitat. Two late-
successional species (Uvariopsis congensis Monodora
myristica) had signi®cantly greater survival (4
greater) in the older than in the younger habitat, while
another (Monodora myristica) showed a trend for
greater survival (2greater) in the older habitat.
With means from all species pooled, mean RGRs
were similar between younger and older successional
habitats (t-test paired by species, t1:0;P
0:351). In the younger and older successional habitats,
there was a trend for early-successional species to
grow faster (1.9and 1.7, respectively) than mid-
and late-successional species (Mann±Whitney test,
P0:079 and 0.064, respectively). RGRs within
species differed signi®cantly between the two succes-
sional habitats for several species (Table 7). For four
early-successional species (Erythrina excelsa,P. afri-
cana,P. fulva,E. trichogyne), RGRs for seedlings in
the younger successional habitat were signi®cantly
greater (1.4±1.6greater) than were those in the older
successional habitat. For one late-successional species
(M. myristica), RGRs were signi®cantly greater in the
Table 7
RGRs and proportions surviving for seedlings transplanted into younger and older successional habitats (averaging 14 and 69 months after
logging, respectively)
a
Seedling species Guild Number planted Days
observed
Mean
height (cm)
RGR per 100 days Proportion surviving
Young Old Young Old PYoung Old P
Trees
A. grandibracteata E 30 30 171 3.9 0.34 0.29 0.325 0.37 0.47 0.549
Aphania senegalensis L 30 30 132 6.8 0.01 0.09 0.325 0.60 0.70 0.631
Bridelia micrantha E 15 12 358 4.1 0.41 0.40 0.624 0.27 0.25 0.933
Celtis durandii M30 23
b
374 3.6 0.37 0.27 0.302 0.20 0.30 0.440
Cordia abyssinica M 30 30 41 3.0 1.53 1.18 0.588 0.20 0.37 0.225
Diospyros abyssinica
c
M 15 15 359 5.3 0.07 0.06 0.837 0.60 0.67 0.819
E. excelsa E 30 30 172 5.8 1.13 0.68 <0.001 0.67 0.70 0.876
Fagaropsis angolensis
c
M 9 8 179 2.5 0.54 0.42 1.000 0.22 0.38 0.560
M. lanceolata
c
E 30 30 443 0.5 0.87 0.83 0.617 0.20 0.03 0.059
M. eminii E 30 30 384 6.1 0.24 0.35 0.384 0.03 0.30 0.011
M. myristica
c
L 30 30 309 11.0 0.37 0.02 <0.001 0.33 0.77 0.024
P. fulva E 30 30 384 3.0 0.58 0.37 0.021 0.10 0.50 0.005
P. africana
c
E 30 15 440 5.3 0.46 0.31 0.002 0.63 0.67 0.095
T. orientalis E 30 30 440 1.3 0.40 ± ± 0.27 0.00 ±
U. congensis
c
L 30 15 261 7.3 0.04 0.07 0.230 0.13 0.53 0.006
Shrubs
E. trichogyne
c
E 30 30 447 1.7 0.60 0.43 0.005 0.73 0.40 0.086
S. virosa E 30 30 434 0.8 0.64 0.82 0.752 0.33 0.03 0.007
Mean (all species) 313 4.2 0.46 0.41 0.35 0.42
S.D. (all species) 129 2.7 0.46 0.32 0.22 0.25
a
Provided for each species is successional guild (E: early-, M: mid- and L: late-successional), sample number, number of days seedlings w
ere monitored, and mean initial height. Pvalues for RGR and proportion surviving are from Mann±Whitney and chi-square tests, respectively.
b
18 of these seedlings germinated in shade plots.
c
Species that germinated only from sun plots.
244 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
older successional habitat (M. myristica had negative
growth in the young habitat). Four species showed
trends or signi®cant differences between treatments
for both RGR and survival. Oddly, two of these had
better growth but lower survival in the sun (P. afri-
cana,P. fulva).
3.3.3. Recruitment below snags
Density and species richness of animal-dispersed
tree seedlings and shrub saplings were signi®cantly
greater (1.5±3greater) below snags than control
plots (Table 8). However, animal-dispersed shrub
seedling and tree sapling density and species richness
were similar between snag and control plots. Densities
and species richnesses of non-animal-dispersed trees
and shrubs were similar between snag and control
plots (Mann±Whitney tests, P>0:10 for all contrasts)
suggesting these areas were similar for other factors
potentially affecting recruitment.
3.4. Alternatives to plantationsЮre exclusion
The successional forest on the ®re-excluded site had
a tall (15±25 m), closed canopy of Albizia grand-
ibracteata trees, with an open understory composed
mainly of early- and mid-successional tree saplings
Table 8
Median (top number) and mean (bottom number S:D:) density and species richness of seedlings and saplings below snags and in adjacent
control areas within a low density plantation 1.5 years after logging
a
Stem variable Density (stems m
2
) Species richness (species m
2
)
Snag plots PControl plots Snag plots PControl plots
Tree seedlings 0.8
*
0.3 0.5
*
0.3
1.8 (2.4) 0.4 (0.5) 0.8 (0.9) 0.3 (0.4)
Shrub seedlings 0 0 0 0
0.1 (0.3) 0.2 (0.3) 0.1 (0.2) 0.2 (0.3)
Tree saplings 0 0.5 0 0.5
0.7 (1.1) 0.5 (0.5) 0.5 (0.6) 0.5 (0.4)
Shrub saplings 1.5
*
0.5 0.5
*
0.5
1.3 (0.9) 0.8 (0.8) 0.7 (0.4) 0.5 (0.4)
a
Results of Wilcoxon Signed Rank comparisons of stem values are presented when signi®cant: P0:05.
Table 9
Median and mean stem variables from the ®re-excluded site, and results of Mann±Whitney contrasts of stem variables (Pvalues) between this
site and logged and unlogged high plantations
a
Stem variable Contrasts with fire-excluded site (FES)
Median and mean (S.D.) Unlogged high density (UHD) Logged high density (LHD)
Seedling density (stems m
2
) 0.50 0.031 0.104
0.93 (0.91) FES >UHD LHD >FES
Seedling species richness (species m
2
) 0.50 0.079 0.383
0.47 (0.34) FES >UHD
Sapling density (stems m
2
) 1.00 0.240 0.003
0.90 (0.38) FES >LHD
Sapling species richness (species m
2
) 0.35 0.021 0.028
0.35 (0.15) FES >UHD FES >LHD
Sapling height (m) 4.25 0.005 0.012
3.93 (1.18) FES >UHD LHD >FES
a
The directions of signi®cant differences are indicated below Pvalues. Marginally insigni®cant trends are indicated within parentheses
(see Table 1 for stem values for unlogged and logged plantations).
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 245
and few shrubs. When compared to the unlogged high
density site, the ®re-excluded site had signi®cantly
greater tree seedling density, tree sapling species
richness, and tree sapling height (2greater for
all variables); there was a trend for greater tree seed-
ling species richness in the ®re-excluded site (Table 9).
Tree sapling density was similar between the unlogged
and ®re-excluded site.
When the ®re-excluded site was compared with the
high density plantation at 5±6 years after logging, the
former had greater tree sapling density (2greater)
and species richness (1.5greater). In contrast, the
logged plantation had taller tree saplings (1.3taller).
There was a trend for greater tree seedling density in
the logged plantation, but tree seedling species rich-
ness was similar between sites.
4. Discussion
Our results provide insight into how exotic tree
plantations can initiate forest regrowth in arrested
successional habitats, a restoration strategy recently
receiving much interest from researchers and man-
agers. We relate our ®ndings to four considerations for
using plantations as a restoration tool; in¯uence of
initial recruit availability; effects of harvest on succes-
sion; facilitation of succession and alternatives to
using plantations.
4.1. In¯uence of initial recruit availability
Differences in tree variables in unlogged high and
low density plantations led to differences in postharvest
succession between plantation types. Fimbel and Fim-
bel (1996) also found these disparities between
unlogged high and low density plantations in Kibale.
They concluded these differences were not soil-related,
since pH and quantities of P, K, Ca, Mn and B were
similar in soils from either plantation type. Concentra-
tions of Mg, Cu and Zn were signi®cantly lower in high
density than low density plantations, a trend that should
support more, not less, recruitment in low density than
high density plantations. In addition, recruitment dif-
ferences are probably not related to landscape position
since planting sites for low density and high density
plantations were similar (Osmaston, 1959; Fimbel and
Fimbel, 1996). Fimbel andFimbel (1996) hypothesized
recruitment differences arose from faster maturation
rates of pines than cypress, more understory light in
high density than low density plantations, and greater
seed dispersal into high density than low density
plantations. Studies elsewhere in the tropics have also
found differences in native stem recruitment among
plantation types (Parrotta, 1995, 1999; though see
Geldenhuys, 1997; Keenan et al., 1997; Powers et al.,
1997; Otsamo, 2000), suggesting succession will vary
among logged plantations in other regions.
We expected forest regrowth in logged low density
plantations to be slower than in logged high density
plantations. While true during the ®rst few years of
succession, by 4±6 years tree sapling density and
species richness were similar between the succes-
sional habitats. This rapid convergence suggests man-
agers may not always need to invest resources into
accelerating succession on sites with low initial recruit
densities. For example, if managers at Kibale want
forest structure in logged low density plantations to
match that in logged high density plantations, they
could intervene in the logged low density plantations
(e.g., seedling planting), or wait 5±10 years for unas-
sisted convergence with succession in the logged high
density plantations. When management goals are
more speci®c (e.g., obtaining a certain density for
one species), waiting for convergence may not be
suf®cient. The dif®culty will be predicting when
convergence will happen naturally, or when interven-
tion is needed.
Other studies of succession in the tropics have not
found evidence for convergence. In Nigeria, Adedeji
(1984) found no convergence in stem recruitment after
2 years of succession where slash was either burned or
removed after forest clearance. Uhl et al. (1982,
1988a) found no convergence during the ®rst 8 years
of succession following light, moderate, or heavy
pasture use in Amazonia. Rates of successional con-
vergence probably depend on various factors, includ-
ing variation in site degradation and availability of
seed sources and seed dispersers. Managers should be
cautious about assuming whether convergence will
occur among different successional habitats.
4.2. Harvesting effects on forest succession
Logging in the low density plantations had no effect
on most stem variables within the ®rst year; the
246 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
exception was an increase in tree seedling density in the
logged plantation. In contrast, logging in high density
plantations led to declines for sapling variables and
increases for seedling variables within the ®rst year.
This increase in seedling density in both plantation
types after logging was likely due to increased resource
availability (especially light) for seedlings in the
understory. Perhaps saplings declined in the high den-
sity and not in the low density plantation because
unlogged high density plantations have twice the den-
sity of timber stems and 16 times the density of native
tree saplings than do unlogged low density plantations.
In the low density plantations, timber extraction
seems to accelerate forest succession. Tree seedlings
and saplings are more abundant and species-rich 4±6
years after logging than in unlogged plantations. In the
high density plantations, timber extraction leads to
increased seedling recruitment and taller saplings, but
sapling density is still less than in unlogged high
density plantations 5±6 years after logging. Once tree
seedlings survive and recruit into larger size classes,
sapling densities in logged high density plantations
may match or exceed densities in unlogged high
density plantations. Thus, despite heavy damage to
native stems during logging, timber extraction accel-
erates natural forest succession in plantations. Thus,
when managing for natural forest regrowth, it may be
better to harvest timber than leaving plantations intact.
This supports the ®ndings of others that removal of
tropical plantation species can promote native seed-
ling growth (Ashton et al., 1998; Otsamo, 1998).
Many questions remain about using plantations for
forest restoration. For example, how do we determine
when to harvest during plantation maturity to max-
imize native tree recruitment? Because native stems
become bigger and, initially, more dense as planta-
tions age (Geldenhuys, 1997; Keenan et al., 1997;
Parrotta, 1999), harvesting when plantations are older
may be better. On the other hand, if logging destroys
much of the native plant community, then logging
when plantations are younger may reduce native stem
loss. When plantations are harvested, managers could
plan felling and extraction to limit native stem
damage. For example, timber could be directionally
felled to avoid patches of native vegetation or native
species of particular interest. Or, selective logging
could reduce damage to advanced regeneration, while
still accelerating forest succession. Any extraction
techniques reducing damage to native stems will also
bene®t succession. For example, in one small planta-
tion area at Kibale, timber was winched by cable to the
road, thus sparing many native trees.
Variation in harvesting activities within logged
forests can produce spatial variation in forest regrowth
(Pinard et al., 2000). Our data suggest seedling recruit-
ment is negatively related to disturbance intensity.
However, tree saplings were similar in heavily and
moderately disturbed areas, perhaps due to their over-
all low densities early in succession. When we com-
pared millsawn and pitsawn areas within a single
plantation, we found greater tree sapling densities
in millsawn than pitsawn plots. One explanation for
this is there is less damage to native stems in areas
where timber is processed off-site (millsawn) than
areas where timber is pitsawn on-site. However, tree
seedling variables were similar between treatments,
suggesting recruitment processes are similar between
millsawn and pitsawn areas later in succession.
4.3. Facilitating forest succession
We quanti®ed whether vegetation removal, direct
plantings, and leaving snags helped facilitate forest
succession in logged plantations. These interventions
were within the ®rst years of succession, when suc-
cession is potentially more tractable than it is later
when it becomes more dif®cult to alter composition or
structure of vegetation.
Previous studies have indicated that vegetation
removal can have either positive or negative effects
on forest succession. Successional habitats often have
high densities of non-tree vegetation that may limit
tree recruitment and growth (Tilman, 1990; Putz and
Canham, 1992; Berkowitz et al., 1995; Sun and Dick-
inson, 1996; Holl, 1998a). On the other hand, harsh
abiotic conditions of recently disturbed habitats can be
unfavorable for tree establishment (Aide and Cavelier,
1994; Brown and Lugo, 1994), and shade provided
by non-tree vegetation may help establishing trees
(Nepstad et al., 1991; Vieira et al., 1994; Zahawi and
Augspurger, 1999). Our removal of all non-tree vege-
tation did not effect recruitment, survival, or growth of
trees at the community level. This suggests that any
facilitative and inhibitive effects of non-tree vegeta-
tion on trees were weak, or that they offset each other.
We conclude that removal of non-tree vegetation at
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 247
our site is not a bene®cial management strategy in
early-successional habitats, although it may be useful
for particular tree species. Chapman et al. (in press)
found that tree density declined in plots where most
non-tree vegetation was removed in a 4±6-year-old
forest on a logged low density plantation at Kibale. In
contrast, others have found that thinning non-tree
vegetation in successional habitats enhances tree
recruitment and growth (Holl, 1998a; Otsamo,
1998; Guariguata, 1999). Until more is known, man-
agers should be cautious about assuming that remov-
ing non-tree vegetation will bene®t forest regrowth in
early-successional habitats.
Less than half of seedlings we planted into succes-
sional habitats at Kibale survived, and successional
habitat age was important for few species. Direct
seedling planting into degraded habitats tends to be
a successful strategy, although it is expensive, time
consuming, and labor intensive (Tucker and Murphy,
1997; Parrotta and Knowles, 1999). Direct planting
overcomes recruitment limitations due to limited seed
dispersal, high seed predation, and poor germination.
However, seedlings are vulnerable to other hazards in
disturbed habitats including herbivory, water-stress,
and shading from adjacent plants. Our limited success
could be improved if plantings of individual species
were timed to optimize survival and growth.
We found greater density of animal-dispersed tree
seedlings and shrub saplings below snags than in
adjacent areas without snags. However, animal-dis-
persed tree saplings and shrub seedlings were similar
between snag and control plots. One explanation for
this pattern is that optimal conditions for recruitment
were early during succession for shrub species and
later for tree species. Non-animal-dispersed species
were similar between snag and control plots suggest-
ing that above patterns were not due to differential
recruitment conditions. Seed rain is usually greater
below emergent perches than in adjacent open areas of
other disturbed tropical systems (Holl, 1998b; Duncan
and Chapman, 1999; Toh et al., 1999). Dispersal to
emergent perches may be especially important where
other vegetative resources for succession are absent.
However, enhanced seed dispersal must be coupled
with favorable conditions for seedling recruitment
(Holl, 1998b), and the effectiveness of snags or other
perches may be limited since they often cover only
small portions of degraded areas.
4.4. Alternatives to plantationsЮre exclusion
We compared forest succession in the ®re-excluded
site to that in an unlogged and logged (5±6 years after
logging) high density plantation, each of which repre-
sents a strategy for managing plantations to promote
natural forest regrowth. We found that tree sapling
density in the ®re-excluded site was similar to that in
the unlogged high density plantation, and that sapling
densities in both these treatments were greater than in
the logged plantation. Tree sapling species richness
followed a similar pattern. Thus, in terms of native tree
sapling recruitment, ®re exclusion was at least as good
as plantations as a restoration strategy. Tree saplings
were taller in the logged plantation than the ®re-
excluded site (though only by 1.3 m). Saplings were
taller in the ®re-excluded site than the unlogged
plantation. This suggest that plantation establishment
and logging may bene®t forest succession more than
®re exclusion for tree growth. However, the ®re-
excluded site had a taller canopy than the logged site
due to the presence of tall pioneer trees not always
found in plots. Seedling recruitment was greater in the
logged plantation than the ®re-excluded site, and both
treatments had greater seedling recruitment than the
unlogged plantation. Without seedlings survivorship
data, it is dif®cult to predict whether the higher
seedling density in the logged plantation would enable
it to surpass the ®re-excluded site for sapling densities.
Whether ®re exclusion or plantation establishment
with subsequent logging is better for promoting forest
succession may depend on which variables are empha-
sized (e.g., sapling growth versus recruitment). Since
from this analysis neither strategy is clearly better,
decisions on whether to exclude ®re or plant timber in
degraded ®re-prone habitats may depend on other
considerations. For example, the costs of ®re exclu-
sion versus plantations establishment and maintenance
may differ, ®re exclusion will not provide the ®nancial
bene®ts possible through timber production, and
degraded areas may need many years of ®re exclusion
until they become less ¯ammable.
4.5. Plantations as a restoration tool
If plantations are to be used for forest restoration,
we need to know more about how to use them to
maximize native tree recruitment and growth. Our
248 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250
®ndings suggest harvesting timber promotes forest
succession more than leaving timber intact, and ®re
exclusion seems to be just as or more effective than
either of these strategies, although one must consider
the costs of excluding ®re and the timescale of the
program. Approaches using a combination of strate-
gies may be particularly useful (Lamb, 1998). For
example, large degraded grasslands could be protected
by ®re-breaks. Smaller areas within the grassland
could be planted with timber species to generate
income, an extractable resource, and an alternative
restoration strategy if ®re exclusion fails. Potentially
invasive species should be avoided. Using suitable
native species would avoid this threat, and probably
increase plantation visitation by native animals
(Lamb, 1998).
Non-ecological challenges exist to using planta-
tions as a restoration strategy (Lamb, 1998). Predict-
ing pro®ts from the sale of timber involves estimating
the market value of timber several decades in advance.
This may be dif®cult since market prices depend on
¯uctuating local, regional, and global timber supplies
(Leslie, 1999). In addition, there can be a constant
temptation to manage plantations for pro®t rather than
for conservation priorities (Lamb, 1998). For example,
because an understory of native species will slow
timber growth, there will be temptation to clear colo-
nizing native vegetation (Lamb, 1998). Or, managers
may decide to replant harvested plantations with more
timber species rather than allowing natural forest
succession to proceed. For these reasons, managers
and institutions backing them need to carefully con-
sider how using plantations to restore biodiversity
compares to alternative strategies.
Acknowledgements
We thank Kaoru Kitajima, Carmine Lanciani, Doug
Levey, and Francis ``Jack'' Putz for helpful comments
on this paper. We would also like to thank ®eld
assistants Swaibu Katusabe Amooti and Francis
Katuramu Amooti for their dedication and hard work.
Funding for this research was provided by the Ford
Foundation, Wildlife Conservation Society, National
Geographic Society, The Explorer's Club, Lindbergh
Foundation, University of Florida, and National
Science Foundation (NSF Graduate Fellowship and
grant numbers SBR-9617664, SBR-990899). Permis-
sion to conduct this research was given by the Of®ce of
the President, Uganda, National Council for Science
and Technology, Uganda, Uganda Wildlife Authority
and Ugandan Forest Department.
References
Adedeji, F.O., 1984. Nutrient cycles and successional changes
following shifting cultivation practice in moist semi-deciduous
forests in Nigeria. For. Ecol. Manage. 9, 87±99.
Aide, T.M., Cavelier, J., 1994. Barriers to lowland tropical forest
restoration in the Sierra Nevada de Santa Marta Colombia.
Rest. Ecol. 2, 219±229.
Ashton, P.M.S., Gamage, S., Gunatilleke, I., Gunatilleke, C.V.S.,
1998. Using Caribbean pine to establish a mixed plantation:
testing effects of pine canopy removal on plantings of rain
forest tree species. For. Ecol. Manage. 106, 211±222.
Berkowitz, A.R., Canham, C.D., Kelly, V.R., 1995. Competition vs.
facilitation of tree seedling growth and survival in early
successional communities. Ecology 76, 1156±1168.
Brown, S., Lugo, A.E., 1990. Tropical secondary forests. J.
Trop. Ecol. 6, 1±32.
Brown, S., Lugo, A.E., 1994. Rehabilitation of tropical lands: a key
to sustaining development. Rest. Ecol. 2, 97±111.
Chapman, C.A., Chapman, L.J., 1996. Exotic tree plantations and
the regeneration of natural forests in Kibale National Park.
Uganda. Biol. Conserv. 76, 253±257.
Chapman, C.A., Chapman, L.J., Zanne, A.E., Burgess, M., in press.
Does weeding promote regeneration of an indigenous tree
community in felled pine plantations in Uganda? Rest. Ecol.
Chazdon, R.L., 1994. The primary importance of secondary forests
in the tropics. Tropinet 5, 1.
De Rouw, A., 1994. Effect of ®re on soil, rice, weeds and forest
regrowth in a rain-forest zone (Cote-Divoire). Catena 22, 133±
152.
Duncan, R.S., Chapman, C.A., 1999. Seed dispersal and potential
forest succession in abandoned agriculture in tropical Africa.
Ecol. Appl. 9, 998±1008.
Duncan, R.S., Duncan, V.E., 2000. Forest succession and distance
from forest edge in an Afro-tropical grassland. Biotropica 32,
33±41.
Eggeling, W.J., Dale, I.R., 1952. The indigenous trees of the
Uganda Protectorate. Government of Uganda Printer, Entebbe,
Uganda.
Fernandes, D.N., Sanford, R.L.J., 1995. Effects of recent land-use
practices on soil nutrients and succession under tropical wet
forest in Costa Rica. Conserv. Biol. 9, 915±922.
Fimbel, R.A., Fimbel, C.C., 1996. The role of exotic conifer
plantations in rehabilitating degraded forest lands: a case study
from the Kibale forest in Uganda. For. Ecol. Manage. 81, 215±
226.
Garciamontiel, D.C., Scatena, F.N., 1994. The effect of human
activity on the structure and composition of a tropical forest in
Puerto Rico. For. Ecol. Manage. 63, 57±78.
R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250 249
Geldenhuys, C.J., 1997. Native forest regeneration in pine and
eucalypt plantations in northern province, South Africa. For.
Ecol. Manage. 99, 101±115.
Gillespie, T.W., Grijalva, A., Farris, C.N., 2000. Diversity,
composition, and structure of tropical dry forests in central
America. Plant Ecol. 147, 37±47.
Guariguata, M.R., 1999. Early response of selected tree species to
liberation thinning in a young secondary forest in northeastern
Costa Rica. For. Ecol. Manage. 124, 255±261.
Hamilton, A., 1991. A Field Guide to Ugandan Forest Trees.
Makerere University Press, Kampala, Uganda.
Holl, K.D., 1998a. Do bird perching structures elevate seed rain
and seedling establishment in abandoned tropical pasture? Rest.
Ecol. 6, 253±261.
Holl, K.D., 1998b. Effects of above- and below-ground competition
of shrubs and grass on Calophyllum brasiliense (Camb.)
seedling growth in abandoned tropical pasture. For. Ecol.
Manage. 109, 187±195.
Katende, A.B., Birnie, A., Tengnas, B., 1995. Useful trees and
shrubs for Uganda: identi®cation, propagation, and manage-
ment for agricultural and pastoral communities. Regional Soil
Consultants Unit, Nairobi, Kenya.
Keenan, R., Lamb, D., Woldring, O., Irvine, T., Jensen, R., 1997.
Restoration of plant biodiversity beneath tropical tree planta-
tions in northern Australia. For. Ecol. Manage. 99, 117±131.
Lamb, D., 1998. Large-scale ecological restoration of degraded
tropical forest lands: the potential role of timber plantations.
Rest. Ecol. 6, 271±279.
Leslie, A.J., 1999. For whom the bell tolls. Trop. For. Update 9,
13±15.
Lugo, A.E., 1997. The apparent paradox of reestablishing species
richness on degraded lands with tree monocultures. For. Ecol.
Manage. 99, 9±19.
Nepstad, D.C., Uhl, C., Serrao, E.A.S., 1991. Recuperation of a
degraded Amazonian landscape: forest recovery and agricultur-
al restoration. Ambio 20, 248±255.
Osmaston,H.A., 1959. Workingplan for the Kibale and Itwara forests:
period 1959±1965. Government of Uganda Printer, Entebbe.
Otsamo, R., 1998. Removal of Acacia mangium overstory
increased growth of underplanted Anisoptera marginata
(Dipterocarpaceae) on an Imperata cylindrica grassland site
in south Kalimantan, Indonesia. New For. 16, 71±80.
Otsamo, R., 2000. Secondary forest regeneration under fast-
growing forest plantations on degraded Imperata cylindrica
grasslands. New For. 19, 69±93.
Parrotta, J.A., 1995. In¯uence of overstory composition on
understory colonization by native species in plantations on a
degraded tropical site. J. Veg. Sci. 6, 627±636.
Parrotta, J.A., Turnbull, J.W., Jones, N., 1997. Introduction
catalyzing native forest regeneration on degraded tropical
lands. For. Ecol. Manage. 99, 1±7.
Parrotta, J.A., 1999. Productivity, nutrient cycling, and succession
in single- and mixed-species plantations of Casuarina equise-
tifolia,Eucalyptus robusta and Leucaena leucocephala in
Puerto Rico. For. Ecol. Manage. 124, 45±77.
Parrotta, J.A., Knowles, O.H., 1999. Restoration of tropical moist
forests on bauxite-mined lands in the Brazilian Amazon. Rest.
Ecol. 7, 103±116.
Pinard, M.A., Barker, M.G., Tay, J., 2000. Soil disturbance and
post-logging forest recovery on bulldozer paths in Sabah,
Malaysia. For. Ecol. Manage. 130, 213±225.
Polhill, R.M., 1952. Flora of Tropical East Africa. A.A. Balkema,
Rotterdam.
Powers, J.S., Haggar, J.P., Fisher, R.F., 1997. The effect of
overstory composition on understory woody regeneration and
species richness in 7-year-old plantations in Costa Rica. For.
Ecol. Manage. 99, 43±54.
Putz, F.E., Canham, C.D., 1992. Mechanisms of arrested succes-
sion in shrublands: root and shoot competition between shrubs
and tree seedlings. For. Ecol. Manage. 49, 267±275.
Sun, D., Dickinson, G.R., 1996. The competition effect of
Brachiaria decumbens on the early growth of direct-seeded
trees of Alphitonia petriei in tropical north Australia.
Biotropica 28, 272±276.
Tilman, D., 1990. Constraints and tradeoffsÐtoward a predictive
theory of competition and succession. Oikos 58, 3±15.
Toh, I., Gillespie, M., Lamb, D., 1999. The role of isolated trees in
facilitating tree seedling recruitment at a degraded sub-tropical
rainforest site. Rest. Ecol. 7, 288±297.
Tucker, N.I.J., Murphy, T.M., 1997. The effects of ecological
rehabilitation on vegetation recruitment: some observations
from the wet tropics of north Queensland. For. Ecol. Manage.
99, 133±152.
Uhl, C., Kauffman, J.B., 1990. Deforestation, ®re susceptibility,
and potential tree responses to ®re in the eastern Amazon.
Ecology 71, 437±449.
Uhl, C., Jordon, C., Clark, K., Clark, H., Herrera, R., 1982.
Ecosystem recovery in Amazon caatinga forest after cutting,
cutting and burning, and bulldozer clearing treatments. Oikos
38, 313±320.
Uhl, C., Buschbacher, R., Serrao, E.A.S., 1988a. Abandoned
pastures in eastern Amazonia. I. Patterns of plant succession. J.
Ecol. 76, 663±681.
Uhl, C., Kauffman, J.B., Cummings, D.L., 1988b. Fire in the
Venezuelan Amazon. 2. Environmental conditions necessary
for forest ®res in the evergreen rainforest of Venezuela. Oikos
53, 176±184.
Vieira, I.C.G., Uhl, C., Nepstad, D., 1994. The role of the shrub
Cordia multispicata Cham. as a succession facilitator in an
abandoned pasture, Paragominas, Amazonia. Vegetatio 115,
91±99.
Zahawi, R.A., Augspurger, C.K., 1999. Early plant succession in
abandoned pastures in Ecuador. Biotropica 31, 540±552.
Zanne, A.E., Chapman, C.A., 2001. Expediting forest regeneration
in tropical grasslands: distance and isolation from seed sources
in plantations. Ecol. Appl. 11, 1610±1621.
250 R.S. Duncan, C.A. Chapman / Forest Ecology and Management 173 (2003) 235±250