Ecological Applications, 28(2), 2018, pp.

373–384
© 2017 by the Ecological Society of America

Early ecological outcomes of natural regeneration and tree

plantations for restoring agricultural landscapes


RICARDO G. CESAR ,1,4 VANESSA S. MORENO,1 GABRIEL D. COLETTA,2 ROBIN L. CHAZDON,1,3
SILVIO F. B. FERRAZ,1 DANILO R. A. DE ALMEIDA,1 AND PEDRO H. S. BRANCALION1
1
Department of Forest Sciences, “Luiz de Queiroz” College of Agriculture, University of S~ao Paulo,
Padua Dias Avenida 11, Piracicaba, SP 13400-970 Brazil
2
Institute of Biology, University of Campinas - UNICAMP, Cidade Universitaria Zeferino Vaz - Bar~ao Geraldo,
Campinas, SP 13083-970 Brazil
3
Department of Ecology and Evolutionary Biology, University of Connecticut, Storrs, Connecticut 06269-3043 USA

Abstract. Mixed tree plantings and natural regeneration are the main restoration
approaches for recovering tropical forests worldwide. Despite substantial differences in imple-
mentation costs between these methods, little is known regarding how they differ in terms of
ecological outcomes, which is key information for guiding decision making and cost-effective
restoration planning. Here, we compared the early ecological outcomes of natural regeneration
and tree plantations for restoring the Brazilian Atlantic Forest in agricultural landscapes. We
assessed and compared vegetation structure and composition in young (7–20 yr old) mixed tree
plantings (PL), second-growth tropical forests established on former pastures (SGp), on former
Eucalyptus spp. plantations (SGe), and in old-growth reference forests (Ref). We sampled trees
with diameter at breast height (DBH) 1–5 cm (saplings) and trees at DBH > 5 cm (trees) in a
total of 32 20 9 45 m plots established in these landscapes. Overall, the ecological outcomes
of natural regeneration and restoration plantations were markedly different. SGe forests
showed higher abundance of large (DBH > 20 cm) nonnative species, of which 98% were
resprouting Eucalyptus trees, than SGp and PL, and higher total aboveground biomass; how-
ever, aboveground biomass of native species was higher in PL than in SGe. PL forests had
lower abundance of native saplings and lianas than both naturally established second-growth
forests, and lower proportion of animal dispersed saplings than SGe, probably due to higher
isolation from native forest remnants. Rarefied species richness of trees was lower in SGp,
intermediate in SGe and Ref and higher in PL, whereas rarefied species richness of saplings
was higher in SG than in Ref. Species composition differed considerably among regeneration
types. Although these forests are inevitably bound to specific landscape contexts and may pre-
sent varying outcomes as they develop through longer time frames, the ecological particulari-
ties of forests established through different restoration approaches indicate that naturally
established forests may not show similar outcomes to mixed tree plantings. The results of this
study underscore the importance that restoration decisions need to be based on more robust
expectations of outcomes that allow for a better analysis of the cost-effectiveness of different
restoration approaches before scaling-up forest restoration in the tropics.
Key words: active restoration; ecological monitoring; forest biomass; forest succession; passive restora-
tion; restoration ecology; second-growth forests; tropical forest restoration.

INTRODUCTION 2011). The scale of restoration initiatives also affects the

degree of intervention. Lower levels of intervention are
Selection of the best restoration approaches is largely
adopted when ecosystem recovery is planned for larger
defined by the potential for autogenic regeneration of
spatial and temporal scales in landscapes that offer
the target site and by management objectives, both of
appropriate conditions for natural regeneration (Chaz-
which determine the level of intervention required to fos-
don and Uriarte 2016), while more intensive and costly
ter ecosystem recovery (McDonald et al. 2016). Because
interventions have been employed in restoration projects
of the spatial variation of these two driving factors, a
requiring faster results but at smaller spatial scales, as in
wide gradient from passive to active restoration is
the case of mandatory projects to comply with environ-
observed, ranging from land abandonment and site
mental laws (Brancalion et al. 2016). For tropical for-
protection to highly costly interventions to reconstruct
ests, a number of science-based guidelines have been
ecological communities in degraded sites (Holl and Aide
developed to guide the selection of the most appropriate
Manuscript received 24 March 2017; revised 7 October 2017;
restoration approach by practitioners, with the goal of
accepted 19 October 2017. Corresponding Editor: Yude Pan. increasing restoration effectiveness and cost reduction
4
E-mail: ricardogoce@yahoo.com.br (Holl and Aide 2011, Stanturf et al. 2014).

373

Ecological Applications
374 CESAR ET AL.
Natural regeneration without human assistance is the
most widely used restoration approach for restoring
tropical forests (Chazdon and Guariguata 2016), while
mixed-species tree planting is preferred for active
restoration (Holl and Aide 2011, Rodrigues et al. 2011).
These contrasting approaches can potentially affect the
structure and composition of restored forests, with
implications for expected outcomes and rates of recov-
ery. During natural regeneration, for instance, the
species pool is limited by the interaction of natural local
dispersal and colonization processes, vegetation cover in
the surrounding landscape, and micro-site conditions for
tree species establishment (Holl et al. 2000, Hooper
et al. 2005). Human activities, such as seed collection
and seedling production in nurseries, overcome dispersal
limitations of species that will be introduced in mixed
tree plantings in restoration sites, allowing the species
composition of the restored forest to vary independently
of the natural potential of tree species to recolonize the
site (Brancalion et al. 2012).
The contribution of landscape dynamics also differs
between these restoration approaches. Whereas natural
regeneration is favored in closer proximity and connec-
tivity to forest remnants (Barnes and Chapman 2014,
Arroyo-Rodriguez et al. 2015, Sloan et al. 2015), mixed
tree plantings are commonly established in sites further
from remnant forests or in landscapes with reduced
forest cover with low levels of seed rain (Rodrigues et al.
2009, Holl and Aide 2011).
Economics and ecology are integrated into decision
making frameworks to select restoration approaches in
different socio-ecological contexts (Holl 2017). Thus,
understanding the ecological outcomes of natural regen-
eration and mixed tree planting in contexts in which both
are needed and viable is a first step towards the selection
of restoration approaches with higher cost effectiveness,
an emerging challenge for up-scaling forest restoration
efforts globally (Birch et al. 2010). Although natural
regeneration has considerably lower costs (but see Zahawi
et al. [2014]), few studies have compared the ecological
outcomes between this approach and tree planting while
controlling for other factors (such as forest age, soil type,
and prior land use), thus preventing more accurate cost-
effectiveness analyses (but see Gilman et al. 2016).
The few comparative publications to date have shown
marked differences between natural regeneration and
mixed-species restoration plantations. Overall, young (5–
10 yr old) mixed tree plantations show higher tree species
richness and lower canopy openness than naturally estab-
lished forests in Australian wet tropics uplands (Shoo
et al. 2016), Andean Ecuador (Wilson and Rhemtulla
2016), humid forests of Costa Rica (Holl et al. 2013, Holl
and Zahawi 2014), and seasonal tropical forests in Brazil
(Brancalion et al. 2016). In contrast, Gilman et al. (2016)
observed that species richness and composition of natu-
rally recruiting individuals in the understory of mixed tree
plantings vs. spontaneously regenerating forests were sim-
ilar in wet forests in Costa Rica, but mixed tree plantings
Vol. 28, No. 2
had higher aboveground biomass due to the planted
trees. These differences may emerge from the compar-
ison of passive and active restoration in conditions
where the potential for autogenic restoration was natu-
rally low, thus limiting the full potential of natural
regeneration to recover forest structure and diversity,
which is known to take several decades or more (Martin
et al. 2013, Chazdon 2014).
The ecological outcomes in single species tree plant-
ings are influenced by the composition, management
treatments, harvesting cycles, and context (Lamb 1998).
As an alternative to mixed tree plantings, single species
plantations consistently house less native plant biodiver-
sity than second-growth forests in their understory, but
such plantations accumulate biomass at faster rates
(Barlow et al. 2007, Bonner et al. 2013). Evidently, the
results of these comparisons are context dependent.
This study compares early outcomes of natural regen-
eration and mixed tree planting in scenarios where both
restoration approaches succeeded in reestablishing a for-
est community. We compared structure, species richness
and composition of saplings and trees in forests under-
going restoration established through mixed-species tree
plantings with natural regeneration in pastures and
Eucalyptus spp. plantations. We also compare these sec-
ond-growth forests with old-growth “reference” forests
near the study region. We expected that second-growth
forests naturally established on pastures and abandoned
Eucalyptus plantations will show, when compared to
mixed tree plantings:
(1) Higher abundance of tree saplings and climbers,
because tree plantings are usually carried out in less
resilient areas, more isolated from seed sources (Holl
and Aide 2011); and because mixed tree plantings
may require more time to support understory colo-
nization by seeds produced by planted trees.
(2) Lower abundance of bigger trees and reduced
aboveground biomass (AGB) of native species, since
management interventions (fertilization, weeding,
control of leaf-cutter ants, isolation from fires and
cattle grazing, and regular spacing) will favor tree
growth in mixed tree plantings, in contrast with self-
organizing, second-growth forests.
(3) Lower species richness, due to the reduced species pool
in agricultural landscapes and the high diversity of spe-
cies introduced in mixed tree plantings in the region.
(4) Higher proportion of abiotic-dispersed trees and
more pioneer species, because practitioners favor
biotic-dispersed species (P. H. S. Brancalion, per-
sonal communication) and a balanced proportion of
fast- and slow-growing species in mixed tree plant-
ings in our study region (Rodrigues et al. 2009).
(5) A more similar composition to old-growth forests,
since composition of second-growth forests will be
essentially determined by the same regional species
pool, while plantations introduce species that are
outside the regional species pool or are locally rare.
March 2018 NATURAL REGENERATION VS. TREE PLANTINGS
METHODS
Study site
We studied second-growth forests established without
human assistance and mixed tree species plantings in the
Corumbata
ı river basin of S~ao Paulo State, southeast
Brazil, in seasonal semi-deciduous tropical forests of the
Atlantic Forest biome. Old-growth, reference forests
were sampled as close as possible from the Corumbata
ı
river basin, because no single conserved forest was found
within it (23.8  4.9 km [mean  one standard devia-
tion] from the basin, minimum 17.2 km, maximum
28.5 km). The climate of the region is classified as Cwa
according to the K€ oppen classification (Alvares et al.
2013), with dry winters and wet summers, mean annual
precipitation of 1,367 mm and mean temperatures of
20.5°C (minimum and maximum monthly averages of
15.6° and 29.5°C, respectively). Altitude varies from 470
to 1,060 m above sea level. The basin occupies an area
of 1,700 km2 and the main soil types are Acrisols (44%)
and Ferralsols (22%). The basin is an ecotone between
the Atlantic Forest (42% of the basin) and Cerrado
(tropical savanna, 58%). The seasonal semi-deciduous
forests characterize the “interior” biogeographical zone
of the Atlantic Forest, the second-most threatened of
this biome, with only 7% native cover remaining
(Ribeiro et al. 2009). The recent increase of native forest
cover in some regions in this biome (Baptista and Rudel
2006, Rezende et al. 2015) provide a unique opportunity
to study the dynamics of tropical forest regeneration in
highly deforested agricultural landscapes.
Most of the deforestation occurred in the 19th century
for coffee plantations, which subsided in the early 20th
century when coffee was gradually replaced by pastures
and intensive agriculture in flatter terrains, mostly sugar-
cane plantations. During the 1970s, the region developed
industrially, resulting in the migration of the rural popu-
lation to urban centers and land abandonment followed
by forest regeneration in some areas unsuitable for mech-
anized agriculture (Dean 1977). This contributed to a
doubling of native forest cover in the six landscapes ana-
lyzed in this study (described in Landscape selection for
sampling forests undergoing restoration) from 8% to 16%
between 1962 and 2008 (Ferraz et al. 2014). Currently,
the main land uses in the Corumbata
ı watershed are pas-
tures and sugarcane fields, occupying 43.7% and 29.4%
of the basin, respectively. Native forest remnants, active
Eucalyptus plantations for firewood and timber, and
other land uses (buildings, roads, water bodies, other
agricultural plantations, etc.) occupy 12.4%, 7.3%, and
7.2% of the entire basin, respectively. Native forest cover
in this region is represented by a mosaic of disturbed
remnants of different sizes and levels of human-mediated
disturbances, mostly resulting from cattle grazing and
recurrent fires coming from sugarcane fields, which were
burned historically before manual harvesting. More
recently, burning of sugarcane fields was legally prohib-
ited and mechanized harvesting has predominated, which
375
favors forest regeneration in slopes that can no longer be
used for mechanized production (Molin et al. 2017).
Landscape selection for sampling forests undergoing
restoration
To locate second-growth forest in the 1,700 km2 Corum
bata
ı River watershed, we selected six landscapes follow-
ing the diversity variability analysis proposed by Pasher
et al. (2013). First, we divided the basin into 1 9 1,
2 9 2, 3 9 3, 4 9 4, and 5 9 5 km square grid cells. For
each grid size, we calculated the Shannon landscape diver-
sity index (McGarigal and Marks 1994) based on a 30-m
resolution land-use map from 2002. Finally, we plotted
the mean landscape diversity index of each grid size
against the cell size. This method identified the 4 9 4 km
cell (16 km2) as the smallest cell grid that shows no varia-
tion when compared to landscape diversity index of larger
sizes, thus representing the landscape diversity of the study
area.
Using the 4-km square grid cell, we submitted the study
region to a moving window analysis and calculated, for
each pixel, the proportion of sugarcane, pasture, and
native forest cover of a sampling window centered on it.
Finally, we selected six of the 4 9 4 km landscapes that
had, in 2008 (latest image available), at least 70% of agri-
cultural matrix and at least 10% native forest cover. To dis-
tribute the six selected landscapes more evenly across the
study region, three landscapes were chosen randomly in
the southern part of the basin and three in the northern
part of the basin. For more details, see Ferraz et al. (2014).
Experimental design
In the six chosen landscapes described previously, we
located second-growth forests that established without
human assistance on planted cattle pastures (hereafter
SGp) and in Eucalyptus spp. plantations abandoned after
harvesting (hereafter SGe), the two most common prior
land uses of second-growth forests in this region. When
we refer to both of these second-growth forests together,
we use the abbreviation SG. We identified SG forests that
were present in the 2008 satellite imagery but not in the
1995 imagery; since data gathering occurred in 2015, we
estimated that sampled SG forests were 7–20 yr old.
Before the establishment of SGp, pastures had low pro-
ductivity (i.e. fewer than two animals per hectare, with-
out regular fertilization or other inputs), harbored very
few isolated native trees, and were covered by nonnative
fodder grasses, mainly Urochloa spp. In SGe, Eucalyptus
spp., a nonnative species, was hand planted and har-
vested using a chainsaw (which probably occurred 6–8 yr
after planting). Some of the harvested eucalyptus
resprouted, resulting in a mixed community of resprout-
ing Eucalyptus spp. and naturally regenerating native
trees. Except for the resprouting Eucalyptus spp. in SGe,
no seedlings were planted in SG forests and all individu-
als sampled originated from natural regeneration.

Ecological Applications
376 CESAR ET AL. Vol. 28, No. 2

High-diversity mixed species tree plantations (here-
after PL) of the same age range (7–20 yr) were identified
in the same study region (basin) but not in the same 16-
km2 landscapes where we sampled SG forests, due to the
lack of mixed tree plantings for restoration plantings in
these small landscapes. These tree plantings were typi-
cally established by sugarcane mills to comply with
Brazilian environmental law (see details in Rodrigues
et al. [2011]). Mixed tree plantings were established in
sites used for decades for intensive sugarcane cultivation,
where low potential for autogenic recovery limited the
use of passive restoration approaches. Over 50 species of
native tree seedlings were planted in a 3 9 2 m spacing
(1,666 seedlings/ha; Rodrigues et al. 2011). Mixed tree
plantings were protected from fire, fenced to prevent cat-
tle invasion, and planted seedlings were fertilized and
weeded for 2–3 yr after planting.
Given that natural regeneration is more likely to occur
near existing forest remnants (Rodrigues et al. 2011,
Sloan et al. 2015) and on slopes (Molin et al. 2017),
mixed tree plantings for forest restoration usually occur
in more isolated areas with low resilience (Rodrigues
et al. 2011); accordingly, the landscape context of SG
and PL forests differed (Fig. 1). Among the 18 SG
forests sampled, only one was not adjacent to a native
forest remnant. On the other hand, the average distance
of the seven PL forests sampled from the closest native
forest remnant was 1,275 m (minimum 10 m, maximum
3,300 m); all PL forests sampled were <100 m from
rivers or water reservoirs and all, except one, was sur-
rounded by large sugarcane plantations. The forest stand
area where each plot was installed was 21  4 ha for
SGp, 50  10 ha for SGe, and 48  36 ha for PL.
Finally, we identified old-growth conserved forests
(Ref) of the same vegetation type in other agricultural
landscapes near the study region. We selected four forests
with no history of large disturbances in the last 100 yr,
protected from human and cattle encroachment and
belonging to relatively large forest areas for the regional
context (50–250 ha), where small remnants predominate.
We installed a 20 9 45 m (900 m2) plot to gather vege-
tation data in seven SGp forests, 11 SGe, and 10 PL
(hereafter collectively referred as “regeneration types”),
and in four Ref sites, totaling 32 plots. To avoid pseudo-
replication, we did not place plots in forests with the
same previous land use located in the same continuous
forest remnant, or in the same continuous planting area
for PL. In each 900-m2 plot, we measured the diameter
at breast height (DBH) and identified to the highest taxo-
nomic level possible all living rooted trees and shrubs
DBH ≥ 5 cm (hereafter trees). We carried out superficial
excavations to confirm if stems of the same species next
to one another belonged to the same individual tree and
we measured all stems DBH > 5 cm of the same tree.
Additionally, we installed a 4 9 30 (120-m2) subplot at
the center of each plot to count and identify all trees and
shrubs with DBH 1–5 cm (hereafter saplings). All sam-
pled individuals were classified according to successional

A)

B) C)

FIG. 1. (A) Location of the selected landscapes in the Corumbata
ı basin, S~ao Paulo State, Brazil. (B and C) High-resolution
satellite images showing overall landscape context of mixed tree plantings for forest restoration (B) and naturally established
second-growth forests (C) sampled in this study. Each regeneration type is highlighted in red in the bottom figure. Note that
second-growth forests are next to existing forest remnants while tree plantings are located along riparian zones within sugarcane
plantations, more isolated from forest fragments.
March 2018 NATURAL REGENERATION VS. TREE PLANTINGS
group (pioneer or non-pioneer), dispersal syndrome
(animal-dispersed, abiotic dispersal, or not classified) and
species origin (native or nonnative to the study region,
Swaine and Whitmore 1988, Crees and Turvey 2015). We
evaluated whether individuals sampled in PL were intro-
duced during planting or established naturally by observ-
ing the distribution of individuals in the planting lines.
Additionally, we estimated liana abundance by walking
two 45-m transects in each plot and counting the number
of touches of lianas with diameter >1 cm in a 2-m high
pole, a method adapted from Vidal et al. (1997). In two
PL plots, we were not able to sample lianas and saplings.
Data analyses
To analyze forest structure, we divided all sampled
trees into three diameter classes based on their largest
stem: 1–5, 5–20, and >20 cm; the first class was sampled
in the 120-m2 subplot, while the others were sampled in
the larger 900-m2 plot. We analyzed the abundance of
trees in each diameter class per plot separately for native
and nonnative species. We estimated AGB of each stem
based on the equation developed by Chave et al. (2014)
that requires wood density. Data on wood density of the
sampled trees were obtained from several references, but
mainly from Chave et al. (2009) and Zanne et al. (2009).
When wood density data was not available for a given
species, we would use the following wood density values,
in this order: (1) average of the species of the same genus
sampled in this study, (2) average of species of the same
genus in Zanne et al. (2009), or (3) average of the species
of the same family sampled in this study. For species
identified only to the genus or family level, we followed
the steps mentioned previously. For unidentified mor-
phospecies, we considered wood density as the average
density of all species sampled in the study site. We used a
specific equation developed locally to estimate AGB of
Eucalyptus spp. trees (Campos et al. 1992).
Aboveground biomass of trees, abundance of native
and nonnative species for each diameter class, species
density of native species, liana abundance, and propor-
tion of pioneers and animal-dispersed native trees and
saplings were compared among regeneration types using
ANOVA and means were compared by the Tukey multi-
ple comparison procedure (a = 0.05) when data were
normally distributed. We used a general mixed model
considering the logarithmical distribution when data
were non-parametrically distributed. In both cases,
means between regeneration types were compared by the
Tukey multiple comparison procedure (a = 0.05).
Rarefied species richness, indicator species and species
composition analyses were carried out using the R package
vegan (Oksanen et al. 2016). We included Ref in these
analyses. Rarefied species richness accumulation curves
were developed for trees and spontaneously regenerating
saplings of each regeneration type using the function rar-
efy. Given the low abundance of spontaneously regenerat-
ing saplings in PL, we were unable to develop a rarefaction
377
curve for this regeneration type. We used the function
indval to identify indicator tree species in each regenera-
tion type with indicator value > 0.5 and P < 0.05. To com-
pare species composition of saplings and trees among
regeneration types, we calculated the Chao–Jaccard dis-
similarity index (Chao et al. 2004) between each plot and
created a graph using nonmetric multidimensional analysis
to visualize similarity among plots of different regenera-
tion types using the mds function. We considered only
native species to calculate rarefied species richness, and
both native and nonnative species for indicator species and
nonmetric multidimensional analyses.
Trees that could not be identified due to the absence or
difficulty to sample vegetative or reproductive material
were not considered for calculating species density, diver-
sity, rarefied richness, and composition, while unidenti-
fied morphospecies for which vegetative material was
collected were included in these analyses. Given that non-
native tree species and pioneer native species are well
described and easily identifiable in the field, we classified
morphospecies that were not identified and trees that we
could not collect vegetative material, due to the absence
of vegetative material or difficulty to reach the canopy, as
native non-pioneer species in order to analyze abundance,
AGB, and proportion of pioneers. Given the low number
of repetitions and the distance from other regeneration
types, Ref forests were not included in the statistical
comparison of structure, biomass, species density, liana
abundance and proportion of biotic-dispersed trees and
pioneers among regeneration types, but the data obtained
in these forests are included in the results for reference
and discussion purposes. Datasets and R scripts for data
analyses can be found in the supplementary file Data S1.
RESULTS
Forest structure
We sampled a total of 1,025 saplings and 3,167 trees, of
which 68 (6.6%) and 410 (12.9%) belonged to nonnative
species, respectively. We could not identify to the species
level 143 (13.9%) and 198 (6.0%) of all saplings and trees,
respectively. Overall, abundance of saplings of native spe-
cies differed among regeneration types, being higher in
SG forests (Appendix S1). Most of the saplings (87.9%)
in PL were planted, with low abundance of spontaneously
regenerating individuals. Forests of different types had
similar abundance of trees DBH 5–20 cm, but SG showed
lower abundance of larger native trees (DBH > 20 cm)
when compared to PL (Fig. 2). Liana abundance differed
among regeneration types, being lower in PL, whereas
SGp and SGe did not differ (Appendix S1).
The abundance of saplings of nonnative species was sim-
ilar among regeneration types; no Eucalyptus saplings were
found in any of the sites (Fig. 2; Appendix S1). However,
SGe had the highest abundance of nonnative tree species
among the largest trees (DBH > 20 cm), mostly large
Eucalyptus spp. individuals. In SGe, resprouted Eucalyptus

Ecological Applications
378 CESAR ET AL. Vol. 28, No. 2

FIG. 2. Tree community structure in 7–20 yr-old second-growth tropical forests established over pastures (SGp) and Eucalyptus
spp. plantations (SGe), mixed-species tree plantings (PL) and reference forests (Ref) in agricultural landscapes of southeastern
Brazil. Abundance of native and nonnative species is shown in the left and right panels, respectively. Community was divided in the
three diameter at breast height (DBH) classes listed in the vertical axis of each graph at the left. Most nonnative trees in SGe are
resprouting Eucalyptus spp. Abundance of native species DBH 1–5 cm was compared using ANOVA (a = 0.05), while the abun-
dance of native species of other DBH classes and all nonnative classes were non-parametrical and a generalized mixed linear model
considering lognormal distribution of data was used (a = 0.05). In both cases, means were compared by the Tukey multiple compar-
ison procedure (a = 0.05). Ref forests were not included in statistical analyses. Boxplots with the same letter above them do not
differ statistically in each DBH class. N = 4 for Ref forests and N = 7, 11, and 10 for SGp, SGe, and PL forests, respectively. Note
that graphs have different scales. Middle line of box plots: median; lower and upper box edges: first and third quartiles, respectively;
lower and upper whiskers: minimum and maximum values, respectively; circles: outliers.

spp. represented 0%, 4.1%  2.2% and 64.0%  6.8% of
trees with DBH 1–5 cm (saplings), 5–20 cm, and >20 cm,
respectively.
Aboveground biomass (AGB) differed among regener-
ation types (Appendix S1). SGe showed similar biomass
of native species to SGp, but less than PL, while total
AGB (native + nonnative species) was higher in SGe
than in the other forests (Fig. 3; Appendix S1). Virtually
all AGB stocked in PL (97.3%) was from planted trees.
Species richness and composition
For trees, we sampled a total of 56 families, 161
genera, and 241 species (Appendix S2). For saplings, we
have sampled a total of 39 families, 100 genera, and 176
species (Appendix S3). Similar values of species density
of native trees were found among forests. Species density
of saplings was higher in SG than in PL when we
included for planted trees in PL (Appendix S1). When
considering only naturally regenerating saplings, PL had
much lower species density than SG forests, with only
two non-planted species found in one plot.
Rarefied species richness of trees was lowest in SGp,
intermediate in SGe and Ref, and highest in PL (Fig. 4).
All the 12 spontaneously regenerating saplings in PL
were represented by only seven species. Additionally, all
the 13 nonnative (to Corumbata
ı Basin) species found in
PL were introduced during planting. Rarefied species
richness of saplings was similar between SG and lowest
in Ref. The few saplings in PL prevent reliable inferences
about rarefied species richness in these forests and
are mostly represented by planted seedlings, with only
March 2018 NATURAL REGENERATION VS. TREE PLANTINGS 379

services and conserving biodiversity in the highly modi-

fied agricultural landscapes included in this study. As
expected, the ecological outcomes in SG and PL were
markedly different in structure and composition, and dif-
fered from reference forests. These differences highlight
that, at least in this region, forests established through
spontaneous regeneration and mixed tree plantings are

FIG. 3. Tree community aboveground biomass in 7–20 yr-

old second-growth forests established over pastures (SGp) and
Eucalyptus spp. plantations (SGe), mixed tree species plantings
(PL), and old-growth reference forests (Ref) in agricultural
landscapes of southeastern Brazil. Gray and white bars repre-
sent aboveground biomass of native and nonnative species,
respectively. Error bars represent one standard error from total
aboveground biomass. Regeneration types were compared using
ANOVA (a = 0.05) and means were compared by the Tukey
multiple comparison procedure (a = 0.05). Bars with the same
letter above them do not differ statistically. Upper- and lower-
case letters refer to total and native species aboveground bio-
mass, respectively. N = 4 for Ref forests and N = 7, 11, and 10
for SGp, SGe, and PL forests, respectively.

twelve individuals of this size class belonging to seven

species found spontaneously in these forests (Fig. 4).
Species composition varied widely among plots, with
each regeneration type composed by a particular species
pool. Composition of SG and PL differed among them-
selves and from Ref (Fig. 5). We identified indicator spe-
cies of trees in each regeneration type (Table 1). Overall,
indicator species of SGp are shade intolerant and wind
dispersed, in SGe, they are shade tolerant and animal
dispersed, in PL they have a balanced distribution
regarding shade tolerance, but abiotic-dispersed species
predominate, and finally, indicator species in Ref forests
are predominantly shade tolerant, with both biotic and
abiotic-dispersed species (Table 1). The proportion of
trees with biotic dispersal was similar among regenera-
tion types, while the proportion of biotic-dispersed sap- FIG. 4. Native species accumulation curve of trees DBH >
lings was higher in SGe than in PL (Appendices S1, S4). 5 cm and DBH 1–5 cm in 7–20 yr-old second-growth (SG) for-
Proportion of pioneer trees was higher in PL than SGe ests established naturally over pastures and Eucalyptus spp.
plantations, mixed tree species plantings, and reference forests
and similar among saplings in all regeneration types
in agricultural landscapes of southeastern Brazil. Species accu-
(Appendix S1). mulation curves for trees DBH 1–5 cm represent only naturally
established trees, and only five regenerating individuals belong-
ing to two species were found in mixed tree plantings in this size
DISCUSSION class. Dotted lines represent one standard error from the mean
number of species. Trees DBH 1–5 cm in mixed plantings are
Both SG and PL forests promoted a rapid recovery of represented mostly by planted individuals, with only 12 trees
forest structure and housed many native tree species, thus belonging to seven species found spontaneously regenerating in
showing a potential contribution for supplying ecosystem this regeneration type.

Ecological Applications
380 CESAR ET AL.
FIG. 5. Two-dimensional nonmetric scaling plot of the
Chao-Jaccard dissimilarity index for 7–20 yr-old forests estab-
lished naturally over pastures (diamonds) and Eucalyptus spp.
plantations (triangles), mixed tree species planting (squares),
and reference forests (circles) in agricultural landscapes of
southeastern Brazil. Data is show for trees DBH > 5 cm and
DBH 1–5 cm in the top and bottom graphs, respectively. We
computed only non-planted trees DBH 1–5 cm for PL.
complementary and fulfill different roles and conditions
in the process of restoring agricultural landscapes.
Surprisingly, resprouting Eucalyptus spp. did not
affect native species density and increased total above-
ground biomass without hindering biomass accumula-
tion by native species when compared to second-growth
forests without Eucalyptus spp. (SGp). Although Euca-
lyptus spp. trees in our plots were reproductively mature
(as observed by the several fallen fruits in our plots), we
have not found a single sapling of this species in SGe
(Appendix S2). Over time, this nonnative species will
gradually disappear from these forests.
Plantings were established in more isolated areas
after intensive agriculture, where natural regeneration is
Vol. 28, No. 2
unlikely to occur for decades and practitioners could not
rely on spontaneous regeneration. In the early succes-
sional stages sampled in our study, PL forests are provid-
ing habitat structure and housing dozens of native
species in agricultural landscapes where it is highly unli-
kely that forest could spontaneously establish in the
short- to mid-term. In spite of the restricted colonization
of native trees and lianas in young PL, mixed tree plant-
ings have shown positive successional development over
time in other forests in the same biome, with a gradual
increase in forest structure and diversity (Suganuma and
Durigan 2015, Bertacchi et al. 2016, Garcia et al. 2016).
These studies highlight that the limited recolonization of
mixed tree plantings understory by native species, as
found in PL in our study, may not compromise the long
term ecological sustainability of these plantings.
As expected for the early successional stages of SG
and PL, these forests differ in most attributes of struc-
ture and composition from Ref. While tree composition
in SG and PL are distinct from Ref, saplings in SG for-
ests tend to be more similar to Ref (Fig. 5). However,
the persistence of planted native tree species in restora-
tion sites is not yet fully understood (Suganuma and
Durigan 2015).
The role of Eucalyptus in natural regeneration
There is much controversy regarding the use of nonna-
tive species to foment the establishment of native tropi-
cal forests (Bremer and Farley 2010). However, recent
studies are reporting the positive effect of the initial use
of nonnative species to create a forest structure and
accelerate successional processes in degraded areas
(Brockerhoff et al. 2013, Catterall 2016). We observed
that resprouting Eucalyptus spp. increased rarefied spe-
cies richness in SGe, when compared to SG forests with-
out Eucalyptus (SGp), probably due to two distinct
processes. First, the establishment of an open, undis-
turbed, shaded understory may have allowed the contin-
uous accumulation of seeds in the soil and promoted
seedling and sapling recruitment. Compared to pastures,
commercial Eucalyptus spp. plantations have less inten-
sive management, as some pastures are annually burned
in our study region. After harvesting of canopy trees,
natural regeneration could then advance faster and with
higher diversity. Second, large Eucalyptus spp. individu-
als may mimic the ecological role of native emergent
trees, establishing a more complex vertical forest struc-
ture that attracts seed dispersers and creates favorable
micro-site conditions for spontaneously colonizing seed-
lings (Johnstone et al. 2016). Lower sapling species rich-
ness in Ref may be due to more intense ecological filters
in the shaded understory of old growth forests, as well as
the reduced number of plots sampled in these forests
(n = 4, Fig. 4).
Populations of Eucalyptus spp. were mostly composed
by few large individuals that increased total biomass of
SGe when compared to SGp, while biomass of native
March 2018 NATURAL REGENERATION VS. TREE PLANTINGS 381

TABLE 1. Indicator species for forests that established naturally in abandoned pastures or Eucalyptus spp. plantations or by mixed
tree species plantings.

Regeneration type Shade tolerance Dispersal syndrome IndVal P

Abandoned pastures (SGp)
Moquiniastrum polymorphum (Less.) G. Sancho I A 0.83 0.001
Psidium guajava L.† I A 0.72 0.007
Machaerium hirtum (Vell.) Stellfeld I A 0.61 0.006
Luehea candicans Mart. & Zucc. I A 0.56 0.014
Machaerium nyctitans (Vell.) Benth. I A 0.54 0.021
Platypodium elegans Vogel I A 0.51 0.017
Abandoned Eucalyptus spp. plantations (SGe)
Eucalyptus spp. I A 0.93 0.001
Casearia sylvestris Sw. T B 0.54 0.020
Mixed tree species plantings (PL)
Schinus terebinthifolius Raddi I B 0.70 0.002
Citharexylum myrianthum Cham. I B 0.69 0.016
Peltophorum dubium (Spreng.) Taub. T A 0.63 0.004
Ceiba speciosa (A.St.-Hil.) Ravenna T A 0.56 0.004
Senegalia polyphylla (DC.) Britton & Rose I A 0.51 0.025
Clitoria fairchildiana R.A.Howard I A 0.50 0.010
Genipa americana L. T B 0.50 0.008
Inga vera Willd. I B 0.50 0.009
Reference forest (Ref)
Roupala montana Aubl. I A 0.72 0.002
Amaioua intermedia Mart. ex Schult. & Schult.f. T B 0.50 0.013
Calyptranthes clusiifolia O.Berg T B 0.50 0.019
Chrysophyllum gonocarpum (Mart. & Eichler ex Miq.) Engl. T B 0.50 0.008
Eugenia dodonaeifolia Cambess. T B 0.50 0.014
Galipea jasminiflora (A.St.-Hil.) Engl. T A 0.50 0.014
Ixora brevifolia Benth. T B 0.50 0.013
Maytenus gonoclada Mart. T B 0.50 0.007
Metrodorea nigra A.St.-Hil. T A 0.50 0.015
Savia dictyocarpa M€ ull.Arg. T A 0.50 0.012
Xylopia brasiliensis Spreng. T B 0.50 0.014
Notes: IndVal, indicator value of the species for that regeneration type. Shade tolerance: I, intolerant to shading; T, tolerant to
shading. Dispersal syndrome: A, abiotic dispersal; B, biotic dispersal. †Nonnative species.

species remained similar between these forests (Fig. 3).
Indicating another benefit of a nonnative species in
accelerating carbon sequestration in second-growth for-
ests, without hindering other native species, at least in
the early stages observed (Bonner et al. 2013).
Naturally established forests and mixed species
plantations
The effect of restoration approaches could not be
decoupled from that of landscape context in our study
(PL are usually far more distant from remnant forests),
so the observed ecological outcomes result from both
the restoration approach and the landscape context in
which it is implemented. Although the influence of land-
scape and methods factors can be separated experimen-
tally, the reality of restoration projects will always reflect
this combination of factors (at least in our study region),
since tree plantings are the preferred method of restora-
tion when natural regeneration is not a viable approach.
Thus, our findings reflect the typical condition in which
different restoration approaches are employed, and the
inherent features of these approaches should be consid-
ered when comparing their cost-effectiveness.
The lack of regenerating trees in the understory of PL
may compromise their future sustainability, as a conse-
quence of the collapse of forest structure following the
senescence of pioneer trees. At the same time, the
absence of lianas in PL demonstrates that other plant
growth forms may have problems to recolonize restora-
tion sites and lead to a less biologically complex restora-
tion community, with limited potential to conserve both
plants and animals (Crouzeilles et al. 2016a, Garcia
et al. 2016, Mayfield 2016).
Although further time may be necessary to assess if
recruitment will increase as these forests age, the limited
spontaneous regeneration of trees and lianas in PL may
be caused by a synergy of local and landscape factors.
Local factors may include reduced seed bank, the lack of
seed production from planted trees, and reduced habitat
heterogeneity promoted by fast-growing planted species
(Holl et al. 2000), factors that may change as PL develop.

Ecological Applications
382 CESAR ET AL.
Landscape factors may include restrictions in seed
sources and dispersers, due to large-scale defaunation and
deforestation in human-modified landscapes of the Atlan-
tic Forest (Sansevero et al. 2011, Crouzeilles et al. 2016b),
which are intensified by the higher level of isolation in the
landscape of restoration plantations. If connectivity
increases in the landscape, these patterns may change.
Despite the benefits of silvicultural treatments like fer-
tilization and weeding for enhancing tree growth applied
during PL establishment (Campoe et al. 2010), these for-
ests showed similar biomass to SGp, the latter regenerat-
ing without human intervention. Our results contrasts
with other works that observed higher AGB in mixed
tree plantings (Holl and Zahawi 2014, Shoo et al. 2016)
and highlights the potential of naturally established sec-
ond-growth forests to develop structurally and sequester
carbon from the atmosphere in relatively short periods
(Chazdon et al. 2016).
SG and PL differed greatly in species taxonomic and
functional composition. As expected, PL had higher spe-
cies richness than SG, due to the introduction of over 50
tree species per planting in the study region (Rodrigues
et al. 2011), but species density at the plot scale was sim-
ilar among regeneration types for both saplings and
trees. Abundance of nonnative trees was low (12.9% of
all individuals sampled), and is not a major source of
divergence among SG or PL forests.
SG and PL also showed striking divergences in species
composition. As expected, the human-oriented selection
of tree species for PL lead to higher levels of floristic dis-
similarities between SG and PL than between naturally
established forests with different previous land uses
(SGp and SGe) for both trees and saplings (Atkinson
and Mar
ın-Spiotta 2015). Sapling composition in PL is
mostly composed by planted individuals and is more dis-
tinct from Ref than SG forests. However, there is overall
high variability among all forests sampled.
Implications for conservation
As climate mitigation policies become increasingly inte-
grated in international environmental agendas, carbon
sequestration through natural regeneration and mixed tree
plantings becomes a promising strategy to remove carbon
from the atmosphere (Locatelli et al. 2015, Chazdon et al.
2016). At least in agricultural landscapes, higher invest-
ments for tree planting in PL may not compensate the
additional gain in AGB considering the success of SG
restoration. If we consider only the costs for establishing
the studied second-growth forests ($0, as lands were sim-
ply abandoned) and the implementation and maintenance
costs of high-density mixed species restoration plantations
in the region (~US$8,000/ha over a three-year period), we
believe that the cost-effectiveness of forest restoration for
carbon stocking is far more favorable in natural regenera-
tion (Chazdon et al. 2016). Former Eucalyptus plantations
can be repurposed for promoting natural regeneration at
virtually no cost, contributing to both carbon stocking
Vol. 28, No. 2
and biodiversity conservation. Nevertheless, larger mixed
tree plantings may also provide valuable ecosystem ser-
vices in isolated landscapes, such as in situ biodiversity
conservation of planted tree species and habitat refugia in
highly deforested landscapes.
Given the low abundance of spontaneously regenerat-
ing individuals and the dissimilarity in species composi-
tion with Ref, enrichment plantings may be needed to
safeguard the persistence of PL forests with limited
understory regeneration (Cole et al. 2011, Bertacchi et al.
2016), as well as to reintroduce climbers (Bourlegat et al.
2013). Complementary to introducing new species, the
removal of nonnative trees from the canopy may promote
the development of natural regeneration in PL and SGe
(Swinfield et al. 2016). Both approaches (SG and PL) can
be strategically used to complement each other in agricul-
tural landscapes. There is a considerable deficit for
restoration in degraded and/or isolated areas that will
probably not support natural regeneration (Soares-filho
et al. 2014), implementing mixed tree plantings in these
areas will be important for carbon storage but may play a
limited role for biodiversity conservation in the long term
unless landscape connectivity is improved. Finally, in lar-
ger scales, differences in composition between SG and PL
may increase beta-diversity in agricultural landscapes
(Rother et al., unpublished manuscript). In this context,
mixed tree plantings provide an opportunity to conserve
endangered or dispersal-limited species.
Finally, decisions have to be made on scientifically
based expectations of ecological outcomes to allow a
better analysis of the cost-effectiveness of restoration
approaches. Simply comparing the costs of natural
regeneration and mixed tree plantings is not sufficient to
support decision making, since the outcomes resulting
from these approaches are not the same. Our findings
reinforce that both restoration approaches have lessons
to provide to each other, since the adoption of silvicul-
tural treatments to assist spontaneously regenerating
seedlings may greatly contribute to the development of
SG, while benchmarking the community assembly of SG
by PL may increase their capacity to overcome the biotic
and abiotic filters operating in degraded sites, reducing
the need for human assistance.
ACKNOWLEDGEMENTS
Funding for this research was provided by the FAPESP grant
2014/14503-7. S. F. B. Ferraz received funding from FAPESP
projects 2011/19767-4 and 2013/22679-5. P. H. S. Brancalion
thanks the National Council for Scientific and Technological
Development of Brazil (CNPq) for a productivity grant (No.
304817/2015-5). R. C
esar was supported by a fellowship from
the Coordination for the Improvement of Higher Education
Personnel of Brazil (CAPES) for research grant (No.
88881.064976/2014-01). This work was also supported by the
PARTNERS Research Coordination Network grant no.
DEB1313788 from the U.S. NSF Coupled Natural and Human
Systems Program. The authors also would like to thank the sev-
eral volunteers during field work and the landowners that
allowed forest sampling in their properties.
March 2018 NATURAL REGENERATION VS. TREE PLANTINGS
LITERATURE CITED
Alvares, C. A., J. L. Stape, P. C. Sentelhas, J. L. De Moraes
Goncßalves, and G. Sparovek. 2013. Koppen’s climate classifi-
cation map for Brazil. Meteorologische Zeitschrift 22:711–728.
Arroyo-Rodriguez, V., F. P. Melo, M. Martinez-Ramos,
F. Bongers, R. L. Chazdon, J. A. Meave, N. Norden, B. A.
Santos, I. R. Leal, and M. Tabarelli. 2015. Multiple succes-
sional pathways in human-modified tropical landscapes: new
insights from forest succession, forest fragmentation and
landscape ecology research. Biological Reviews of the Cam-
bridge Philosophical Society 92:326–340.
Atkinson, E. E., and E. Mar
ın-Spiotta. 2015. Land use legacy
effects on structure and composition of subtropical dry
forests in St. Croix, U.S. Virgin Islands. Forest Ecology and
Management 335:270–280.
Baptista, S. R., and T. K. Rudel. 2006. A re-emerging Atlantic
forest? Urbanization, industrialization and the forest transi-
tion in Santa Catarina, southern Brazil. Environmental Con-
servation 33:195.
Barlow, J., et al. 2007. Quantifying the biodiversity value of trop-
ical primary, secondary, and plantation forests. Proceedings of
the National Academy of Sciences USA 104:18555–18560.
Barnes, A. D., and H. M. Chapman. 2014. Dispersal traits
determine passive restoration trajectory of a Nigerian mon-
tane forest. Acta Oecologica 56:32–40.
Bertacchi, M. I. F., N. T. Amazonas, P. H. S. Brancalion, G. E.
Brondani, A. C. S. Oliveira, M. A. R. Pascoa, and R. R.
ROdrigues. 2016. Establishment of tree seedlings in the
understory of restoration plantations: natural regeneration
and enrichment plantings. Restoration Ecology 24:100–108.
Birch, J. C., A. C. Newton, C. A. Aquino, E. Cantarello,
C. Echeverria, T. Kitzberger, I. Schiappacasse, and N. T.
Garavito. 2010. Cost-effectiveness of dryland forest restoration
evaluated by spatial analysis of ecosystem services. Proceedings
of the National Academy of Sciences USA 107:21925–21930.
Bonner, M. T. L., S. Schmidt, and L. P. Shoo. 2013. A meta-
analytical global comparison of aboveground biomass accu-
mulation between tropical secondary forests and monoculture
plantations. Forest Ecology and Management 291:73–86.
Bourlegat, J. M. G., S. Gandolfi, P. H. S. Brancalion, and C. T.
S. Dias. 2013. Enriquecimento de floresta em restauracß~ao por
meio de semeadura direta de lianas. Hoehnea 40:465–472.
Brancalion, P. H. S., D. Schwizer, U. Gaudare, J. R. Mangueira,
F. Lamonato, F. T. Farah, A. G. Nave, and R. R. Rodrigues.
2016. Balancing economic costs and ecological outcomes of
passive and active restoration in agricultural landscapes: the
case of Brazil. Biotropica 48:856–867.
Brancalion, P. H. S., R. A. G. Viani, J. Aronson, R. R. Rodrigues,
and A. G. Nave. 2012. Improving planting stocks for the Brazil-
ian Atlantic Forest restoration through community-based seed
harvesting strategies. Restoration Ecology 20:704–711.
Bremer, L. L., and K. A. Farley. 2010. Does plantation forestry
restore biodiversity or create green deserts? A synthesis of the
effects of land-use transitions on plant species richness. Biodi-
versity and Conservation 19:3893–3915.
Brockerhoff, E. G., H. Jactel, J. A. Parrotta, and S. F. B. Ferraz.
2013. Role of eucalypt and other planted forests in biodiver-
sity conservation and the provision of biodiversity-related
ecosystem services. Forest Ecology and Management 301:
43–50.
Campoe, O. C., J. L. Stape, and J. C. T. Mendes. 2010. Can inten-
sive management accelerate the restoration of Brazil’s Atlantic
forests? Forest Ecology and Management 259:1808–1814.
Campos, J. C. C., J. A. Silva, and B. R. Vital. 1992. Volume e
biomassa do tronco e da copa de eucalipto de grande porte.


Revista Arvore 16:319–336.
383
Catterall, C. P. 2016. Roles of non-native species in large-scale
regeneration of moist tropical forests on anthropogenic grass-
land. Biotropica 48:809–824.
Chao, A., R. L. Chazdon, R. K. Colwell, and T.-J. Shen. 2004.
A new statistical approach for assessing similarity of species
composition with incidence and abundance data. Ecology
Letters 8:148–159.
Chave, J., D. A. Coomes, S. Jansen, S. L. Lewis, N. G. Swenson,
and A. E. Zanne. 2009. Towards a worldwide wood eco-
nomics spectrum. Ecology Letters 12:351–366.
Chave, J., et al. 2014. Improved allometric models to estimate
the aboveground biomass of tropical trees. Global Change
Biology 20:3177–3190.
Chazdon, R. L. 2014. Second growth: the promise of tropical
forest regeneration in the age of deforestation. The University
of Chicago Press, Chicago, Illinois, USA.
Chazdon, R. L., and M. R. Guariguata. 2016. Natural regener-
ation as a tool for large-scale forest restoration in the tropics:
prospects and challenges. Biotropica 48:716–730.
Chazdon, R. L., and M. Uriarte. 2016. Natural regeneration in
the context of large-scale forest and landscape restoration in
the tropics. Biotropica 48:709–715.
Chazdon, R. L., et al. 2016. Carbon sequestration potential of
second-growth forest regeneration in the Latin American
tropics. Science Advances 2:e1501639.
Cole, R. J., K. D. Holl, C. L. Keene, and R. A. Zahawi. 2011.
Direct seeding of late-successional trees to restore tropical mon-
tane forest. Forest Ecology and Management 261:1590–1597.
Crees, J. J., and S. T. Turvey. 2015. What constitutes a “native”
species? Insights from the Quaternary faunal record. Biologi-
cal Conservation 186:143–148.
Crouzeilles, R., M. Curran, and Y. Clough. 2016a. Which land-
scape size best predicts the influence of forest cover on
restoration success? A global meta-analysis on the scale of
effect. Journal of Applied Ecology 53:440–448.
Crouzeilles, R., M. Curran, M. S. Ferreira, D. B. Lindenmayer,
C. E. Grelle, and J. M. Rey Benayas. 2016b. A global meta-
analysis on the ecological drivers of forest restoration success.
Nature Communications 7:11666.
Dean, W. 1977. Rio Claro: um sistema brasileira de grande
lavoura, 1820–1920. Paz e Terra, Rio de Janeiro, Brazil.
Ferraz, S. F. B., K. M. P. M. B. Ferraz, C. C. Cassiano, P. H. S.
Brancalion, D. T. A. da Luz, T. N. Azevedo, L. R. Tambosi,
and J. P. Metzger. 2014. How good are tropical forest patches
for ecosystem services provisioning? Landscape Ecology
29:187–200.
Garcia, L. C., R. J. Hobbs, D. B. Ribeiro, J. Y. Tamashiro, F. A.
M. Santos, and R. R. Rodrigues. 2016. Restoration over time:
Is it possible to restore trees and non-trees in high-diversity
forests? Applied Vegetation Science 19:655–666.
Gilman, A. C., S. G. Letcher, R. M. Fincher, A. I. Perez, T. W.
Madell, A. L. Finkelstein, and F. Corrales-Araya. 2016.
Recovery of floristic diversity and basal area in natural forest
regeneration and planted plots in a Costa Rican wet forest.
Biotropica 48:798–808.
Holl, K. D. 2017. Restoring tropical forests from the bottom
up. Science 355:455–456.
Holl, K. D., and T. M. Aide. 2011. When and where to actively
restore ecosystems? Forest Ecology and Management
261:1558–1563.
Holl, K. D., M. E. Loik, E. H. V. Lin, and I. A. Samuels. 2000.
Tropical montane forest restoration in Costa Rica: overcom-
ing barriers to dispersal and establishment. Restoration Ecol-
ogy 8:339–349.
Holl, K. D., V. M. Stout, J. L. Reid, and R. A. Zahawi. 2013.
Testing heterogeneity-diversity relationships in tropical forest
restoration. Oecologia 173:569–578.

Ecological Applications
384 CESAR ET AL.
Holl, K. D., and R. A. Zahawi. 2014. Factors explaining variabil-
ity in woody above-ground biomass accumulation in restored
tropical forest. Forest Ecology and Management 319:36–43.
Hooper, E., P. Legendre, and R. Condit. 2005. Barriers to forest
regeneration of deforested and abandoned land in Panama.
Journal of Applied Ecology 42:1165–1174.
Johnstone, J. F., et al. 2016. Changing disturbance regimes, eco-
logical memory, and forest resilience. Frontiers in Ecology
and the Environment 14:369–378.
Lamb, D. 1998. Degraded tropical forest lands: the potential
role of timber plantations. Restoration Ecology 6:271–279.
Locatelli, B., C. P. Catterall, P. Imbach, C. Kumar, R. Lasco,
E. Mar
ın-Spiotta, B. Mercer, J. S. Powers, N. Schwartz, and
M. Uriarte. 2015. Tropical reforestation and climate change:
beyond carbon. Restoration Ecology 23:337–343.
Martin, P. A., A. C. Newton, and J. M. Bullock. 2013. Carbon
pools recover more quickly than plant biodiversity in tropical
secondary forests. Proceedings of the Royal Society B
280:20132236.
Mayfield, M. M. 2016. Restoration of tropical forests requires
more than just planting trees, a lot more. Applied Vegetation
Science 19:553–554.
McDonald, T., J. Jonson, and K. W. Dizon. 2016. National
standards for the practice of ecological restoration in Aus-
tralia. Restoration Ecology 24:S4–S32.
McGarigal, K., and B. J. Marks. 1994. FRAGSTATS: spatial
pattern analysis program for quantifying landscapes struc-
ture. General Technical Report PNW-GTR-351. U.S. Depart-
ment of Agriculture, Forest Service, Pacific Northwest
Research Station, Portland, Oregon, USA 97331:134.
Molin, P. G., S. E. Gergel, B. S. Soares-Filho, and S. F. B.
Ferraz. 2017. Spatial determinants of Atlantic Forest loss and
recovery in Brazil. Landscape Ecology 32:857–870.
Oksanen, J. F., et al. 2016. vegan: community ecology package.
https://cran.r-project.org/web/packages/vegan/
Pasher, J., S. W. Mitchell, D. J. King, L. Fahrig, A. C. Smith,
and K. E. Lindsay. 2013. Optimizing landscape selection for
estimating relative effects of landscape variables on ecological
responses. Landscape Ecology 28:371–383.
Rezende, C. L., A. Uezu, F. R. Scarano, and D. S. D. Araujo.
2015. Atlantic Forest spontaneous regeneration at landscape
scale. Biodiversity and Conservation 24:2255–2272.
Ribeiro, M. C., J. P. Metzger, A. C. Martensen, F. J. Ponzoni,
and M. M. Hirota. 2009. The Brazilian Atlantic Forest: How
much is left, and how is the remaining forest distributed?
Implications for conservation. Biological Conservation
142:1141–1153.
Rodrigues, R. R., S. Gandolfi, A. G. Nave, J. Aronson, T. E.
Barreto, C. Y. Vidal, and P. H. S. Brancalion. 2011. Large-scale
SUPPORTING INFORMATION
Additional supporting information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/eap.1653/full
Vol. 28, No. 2
ecological restoration of high-diversity tropical forests in SE
Brazil. Forest Ecology and Management 261:1605–1613.
Rodrigues, R. R., R. A. F. Lima, S. Gandolfi, and A. G. Nave.
2009. On the restoration of high diversity forests: 30 years of
experience in the Brazilian Atlantic Forest. Biological Con-
servation 142:1242–1251.
Sansevero, J. B. B., P. V. Prieto, L. F. D. de Moraes, and P. J. P.
Rodrigues. 2011. Natural regeneration in plantations of
native trees in lowland Brazilian Atlantic forest: community
structure, diversity, and dispersal syndromes. Restoration
Ecology 19:379–389.
Shoo, L. P., K. Freebody, J. Kanowski, and C. P. Catterall.
2016. Slow recovery of tropical old-field rainforest regrowth
and the value and limitations of active restoration. Conserva-
tion Biology 30:121–132.
Sloan, S., M. Goosem, and S. G. Laurance. 2015. Tropical
forest regeneration following land abandonment is driven by
primary rainforest distribution in an old pastoral region.
Landscape Ecology 31:601–618.
Soares-filho, B., R. Raj~ao, M. Macedo, A. Carneiro, W. Costa,
M. Coe, H. Rodrigues, and A. Alencar. 2014. Cracking
Brazil’s forest code. Science 344:363–364.
Stanturf, J. A., B. J. Palik, and R. K. Dumroese. 2014. Contem-
porary forest restoration: a review emphasizing function. For-
est Ecology and Management 331:292–323.
Suganuma, M. S., and G. Durigan. 2015. Indicators of restora-
tion success in riparian tropical forests using multiple refer-
ence ecosystems. Restoration Ecology 23:238–251.
Swaine, M. D., and T. C. Whitmore. 1988. On the definition of
ecological species groups in tropical rain forests. Vegetatio
75:81–86.
Swinfield, T., R. Afriandi, F. Antoni, and R. D. Harrison. 2016.
Accelerating tropical forest restoration through the selective
removal of pioneer species. Forest Ecology and Management
381:209–216.
Vidal, E., J. Johns, J. J. Gerwing, P. Barreto, and C. Uhl. 1997.
Vine management for reduced impact logging in eastern
Amazonia. Forest Ecology and Management 98:105–114.
Wilson, S. J., and J. M. Rhemtulla. 2016. Acceleration and
novelty: community restoration speeds recovery and trans-
forms species composition in Andean cloud forest. Ecological
Applications 26:203–218.
Zahawi, R. A., J. L. Reid, and K. D. Holl. 2014. Hidden costs
of passive restoration. Restoration Ecology 22:284–287.
Zanne, A. E., G. Lopez-Gonzalez, D. A. Coomes, J. Ilic, S.
Jansen, S. L. Lewis, R. B. Miller, N. G. Swenson, M. C.
Wiemann and J. Chave. 2009. Data from: towards a world-
wide wood economics spectrum. Dryad Digital Repository.
https://doi.org/10.5061/dryad.234