INNOVATIVE VIEWPOINTS
Reframing tropical savannization: linking changes in canopy
structure to energy balance alterations that impact climate
SCOTT C. S 1,TARK , 
 DAVID D. BRESHEARS ,2,3 SUSAN ARAGO
N ,4,5,6 J 2,7UAN CAMILO VILLEGAS ,
DARIN J. L ,2 M 1 1,8 9,10AW ARIELLE N. SMITH , DAVID M. MINOR, RAFAEL LEANDRO DE ASSIS ,
D 11ANILO ROBERTI ALVES DE ALMEIDA , GABRIEL DE O ,12,13LIVEIRA SCOTT R. S ,3ALESKA
A 14 15 9 16BIGAIL L. S. SWANN , JOSE
 MAURO S. MOURA , JOSE
 LUIS CAMARGO , RODRIGO DA SILVA ,
LUIZ E. O. C. A 13,17 18RAGA~O , AND R. COSME OLIVEIRA
1Department of Forestry, Michigan State University, East Lansing, Michigan 48824 USA
2School of Natural Resources and the Environment, University of Arizona, Tucson, Arizona 85721 USA
3Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85721 USA
4Center of Integrated Studies of Amazonian Biodiversity (CENBAM), National Institute of Amazonian Research (INPA),
Manaus, Amazonas 69067-375 Brazil
5Program of Postgraduate Studies of Natural Resources of the Amazon, Federal University of Western Par
a (UFOPA),
Santar
em, Par
a, Brazil
6Institute of Environment, Territory and Renewable Energy (INTE), Pontificia Universidad Cat
olica del Peru
 (PUCP), Lima, Peru
7Grupo de Ecolog
ıa Aplicada, Universidad de Antioquia, Medell
ın, Colombia
8Department of Geographical Sciences, University of Maryland at College Park, 2181 Samuel J. LeFrak Hall, 7251 Preinkert Drive,
College Park, Maryland 20742 USA
9Biological Dynamics of Forest Fragments Project, PDBFF, Instituto Nacional de Pesquisas da Amazônia, Av. Andr
e Araujo,
2936 - Petr
opolis, Manaus, Amazonas 69067-375 Brazil
10Natural History Museum (NHM), University of Oslo (UiO), P.O. Box 1172,Blindern, Oslo 0318 Norway
11Department of Forest Sciences, “Luiz de Queiroz” College of Agriculture, University of S~ao Paulo (USP/ESALQ, Piracicaba,
S~ao Paulo, Brazil
12Department of Geography and Planning, University of Toronto, Toronto, Ontario M5S 3G3 Canada
13Remote Sensing Division, National Institute for Space Research (INPE), S~ao Jos
e dos Campos, Brazil
14Department of Atmospheric Sciences and Department of Biology, University of Washington, Seattle, Washington 98195 USA
15Interdisciplinary Training Center, Federal University of Western Para, Santar
em, Par
a 68040-255 Brazil
16Laborat
orio de F
ısica e Qu
ımica da Atmosfera - Bloco 29, Rua Vera Paz, Santar
em, Par
a 68040-260 Brazil
17College of Life and Environmental Sciences, University of Exeter, Exeter, UK
18Embrapa Amazonia Oriental, Santarem, Par
a 68020-640 Brazil
Citation: Stark, S. C. D. D. Breshears, S. Arago
n, J. C. Villegas, D. J. Law, M. N., Smith, D. M. Minor, R. L. de Assis, D. R.
A. de Almeida, G. de Oliveira, S. R. Saleska, A. L. S. Swann, J. M. S. Moura, J. L. Camargo, R. da Silva, L. E. O. C.
Arag~ao, and R. C. Oliveira. 2020. Reframing tropical savannization: linking changes in canopy structure to energy
balance alterations that impact climate. Ecosphere 11(9):e03231. 10.1002/ecs2.3231
Abstract. Tropical ecosystems are undergoing unprecedented rates of degradation from deforestation, fire,
and drought disturbances. The collective effects of these disturbances threaten to shift large portions of tropi-
cal ecosystems such as Amazon forests into savanna-like structure via tree loss, functional changes, and the
emergence of fire (savannization). Changes from forest states to a more open savanna-like structure can affect
local microclimates, surface energy fluxes, and biosphere–atmosphere interactions. A predominant type of
ecosystem state change is the loss of tree cover and structural complexity in disturbed forest. Although impor-
tant advances have been made contrasting energy fluxes between historically distinct old-growth forest and
savanna systems, the emergence of secondary forests and savanna-like ecosystems necessitates a reframing to
consider gradients of tree structure that span forest to savanna-like states at multiple scales. In this Innovative
Viewpoint, we draw from the literature on forest–grassland continua to develop a framework to assess the
consequences of tropical forest degradation on surface energy fluxes and canopy structure. We illustrate this
framework for forest sites with contrasting canopy structure that ranges from simple, open, and savanna-like
to complex and closed, representative of tropical wet forest, within two climatically distinct regions in the
 v www.esajournals.org 1 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Amazon. Using a recently developed rapid field assessment approach, we quantify differences in cover, leaf
area vertical profiles, surface roughness, albedo, and energy balance partitioning between adjacent sites and
compare canopy structure with adjacent old-growth forest; more structurally simple forests displayed lower
net radiation. To address forest–atmosphere feedback, we also consider the effects of canopy structure change
on susceptibility to additional future disturbance. We illustrate a converse transition—recovery in structure
following disturbance—measuring forest canopy structure 10 yr after the imposition of a 5-yr drought in the
ground-breaking Seca Floresta experiment. Our approach strategically enables rapid characterization of sur-
face properties relevant to vegetation models following degradation, and advances links between surface
properties and canopy structure variables, increasingly available from remote sensing. Concluding, we
hypothesize that understanding surface energy balance and microclimate change across degraded tropical for-
est states not only reveals critical atmospheric forcing, but also critical local-scale feedbacks from forest sensi-
tivity to additional climate-linked disturbance.
Key words: Amazon; climate change; Earth System Models; energy balance; forest transitions; lidar; rapid field
assessment; savannization; vegetation structure.
Received 10 October 2019; revised 5 May 2020; accepted 12 May 2020; final version received 22 June 2020. Corresponding
Editor: Debra P. C. Peters.
Copyright: © 2020 The Authors. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

 E-mail : scstark@msu.edu
INTRODUCTION many areas, where savanna is maintained by fire
feedbacks that promote grass, light-demanding
Tropical ecosystems are undergoing unprece- fire-resilient trees, and open canopy structure,
dented rates of anthropogenically caused distur- and thus fire recurrence (i.e., a fire trap; Des-
bance and conversion. For example, in Brazil in jardins et al. 1996, House et al. 2003, Hirota et al.
2019, there was a roughly 50% increase from the 2011, Ratnam et al. 2011, Staver et al. 2011, Hoff-
previous year in deforestation followed by exten- mann et al. 2012, Oliveras and Malhi 2016). Now,
sive fires (Escobar 2019), with similar forest loss a wider range of forest types are present in these
rates recently recorded for Amazônia in other regions because of changing disturbance regimes
countries such as Colombia and Bolivia (Kala- that create degraded and potentially pre-savan-
mandeen et al. 2018, Salazar et al. 2018). These nized or secondarized forest states (Barlow and
changes impact ecosystem processes at local, Peres 2008, Nepstad et al. 2008; degraded forests
regional, and global scales, including increasing we define as having lost one or more ecosystem
atmospheric greenhouse gases such as CO2, CO, functions over a period of time). To address the
CH4, and N2O (Gash et al. 2004, Davidson et al. needs of forecasting future climates with Earth
2012, de Oliveira et al. 2019). Efforts to model the system approaches (IPCC 2013, 2014) and devel-
effects of climate change suggest that portions of oping management strategies to reduce savan-
Amazon forests are threatened with conversion nization (Malhi et al. 2009), it is now essential to
into a savanna due to tree loss from deforestation, develop a detailed understanding of forest–atmo-
fire, and die-offs associated with extreme climate sphere interactions over this range of degraded
conditions and drought (Brando et al. 2008, 2019, forest types. This understanding requires expand-
Nepstad et al. 2008, Phillips et al. 2009, Arag~ao ing quantitative characterization of canopy func-
and Shimabukuro 2010, Allen et al. 2015). Savan- tion and microclimate changes within the
nization (sometimes called savannification) is the spectrum of degraded states, and knowledge of
transformation of forest to lower biomass how functional and microclimate changes will
savanna structure, associated with the emergence feedback to influence the sensitivity of forests to
of fire in the system (Silv
erio et al. 2013). Forest droughts, heatwaves, and fires that could cause
and savannas represent alternate forests states in additional degradation.
 v www.esajournals.org 2 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Changes from a mature forest state to more Rappaport et al. 2018, Shao et al. 2019, Tang
open savanna-like structure, including secondary et al. 2019, Almeida et al. 2019a).
or transitional forest, can affect the local energy Our specific objectives in this Innovative View-
fluxes, tipping the balance and associated pro- point article are to, in Section 1: Reframing, pro-
cesses (including feedbacks) toward savanniza- pose a reframing of different types of forest
tion (Cochrane and Laurance 2008, Arag~ao and transitions in tropical wet forest to highlight the
Shimabukuro 2010, Hirota et al. 2010, Ordway need to understand the impacts of forest change
and Asner 2020). Although advances have been on energy balance and the consequences for
made in quantifying energy fluxes in Amazon forest–atmosphere interactions; in Section 2: Con-
forests and in distant savannas in adjacent trasts, illustrate potential to aide in addressing
regions (da Rocha et al. 2009, Restrepo-Coupe this knowledge gap by presenting example data
et al. 2013), less is known about changes in sets of vegetation differences and associated
surface energy fluxes and microclimates associ- energy characterizations based on short field
ated with forest disturbances. Furthermore, a campaigns; and, in Section 3: Challenges and Pro-
predominant type of ecosystem state change is spects, discuss hypotheses about energy implica-
tree cover loss of different amounts in tropical tions of vegetation change to more open and less
forest due to human-related disturbances (Curtis complex structures, including the sensitivity of
et al. 2018, Bullock et al. 2020), while degraded degraded forest to fire-driven savannization and
forests are becoming more widespread and may other vegetation change feedbacks. We conclude
become the dominant mode of Amazon forests suggesting steps to improve the savannization
in the future (Bullock et al. 2020). These changes vegetation change pathways in Earth System
necessitate a reframing of tropical forest energy models (ESMs) and to develop new remote sens-
fluxes and microclimates from the current per- ing programs to closely monitor the changing
spective that primarily contrasts old-growth for- structure and function of tropical canopies in
est with savanna in a different region, to Amazon forests and elsewhere.
considering a multidimensional gradient of tree
cover and structure that spans forest to grass- SECTION 1: REFRAMING TROPICAL FOREST
land at multiple scales, including within a given SAVANNIZATION AS TRANSITIONS ALONG
location and climate. This reframing can comple- CANOPY STRUCTURE AND ENERGY BALANCE
ment ecological and biogeographic understand- GRADIENTS
ing that considers a wide spectrum of
disturbance, recovery, and historical community We propose a simple conceptual framework
assembly-related structural states that span trop- that links disturbance and forest state transitions
ical wet forests to savannas (Pennington et al. with gradients of canopy structural complexity
2000, Barlow and Peres 2008, Berenguer et al. and cover (Fig. 1A), and with surface energy bal-
2014, Chazdon 2014), as well as advances in ance components, to improve modeling of
remote detection of forest structure change (Rap- forest–atmosphere interactions (Fig. 1). The key
paport et al. 2018, Smith et al. 2019, Bullock et al. global change drivers considered that can trigger
2020). Here, we propose a program for the con- transitions among these states include degrada-
ceptual and empirical steps needed to achieve tion from deforestation, fires, and drought and
this reframing—including providing new data heatwaves, and forest regrowth (Fig. 1B). Struc-
on forest structure–energy balance and microcli- tural predictor variables that may be detected
mate contrasts—to meet these needs. We draw remotely include forest canopy cover and canopy
from previously developed dryland literature on structure, tree size distribution, and leaf area
forest–grassland continua (Breshears 2006 and index (LAI; Fig. 1C). The relationships between
references therein) and expand that framework these predictor and response variables for micro-
from a focus on cover to include canopy struc- climate and energy balance components—in-
tural complexity (canopy cover, and vertical and cluding albedo, net radiation, sensible to latent
horizontal variation), which can be retrieved heat (Bowen ratio), surface roughness and
from advanced remote sensing technologies boundary layer conductance, and near-ground
(Chambers et al. 2007, Stark et al. 2012, 2015, incoming solar radiation—are incompletely
 v www.esajournals.org 3 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Fig. 1. Conceptual figure of proposed framework for assessing the consequences of tropical forest disturbance
and forest degradation on surface energy balance, and subsequent ecological and ecosystem properties, extend-
ing a temperate drylands framework focused on woody plant cover (Breshears 2006) to consider canopy struc-
tural changes more generally. This framework is useful for a variety of gradients of forest structure, including
those related to deforestation and degradation in one direction and forest regrowth/regeneration in the other (A),
and, most relevant to tropical forest savannization, decreases in structure and cover associated with tree die-off
from droughts, heatwaves, and the emergence of fire (B). The characteristics to consider along these gradients are
those integrally linked to microclimate and canopy function—forest cover and structure, tree size distribution,
and LAI (C). As noted in Table 1, these characteristics influence energy balance components and associated
microclimates in predictable ways, even if relationships are not completely characterized (D). These functions
could increase or decrease in nonlinear as well as linear ways or could follow a peaked curve relationship where
the maxima usually falls below 50% tree cover (Breshears 2006, Villegas et al. 2014; E). For nonpeaked relation-
ships, the degree to which tree canopy units of a given structure interactively influence the area around them
determines how linear or nonlinear and threshold like the functional responses are (Breshears 2006).
 v www.esajournals.org 4 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Table 1. Hypothesized functional responses for key properties moving from closed, mature lowland tropical for-
est canopy to very open canopy structure, driven by a combination of fire, tree mortality, deforestation and
degradation.
Feedback on future Representative Representative
Functional Threshold disturbance/ temperate primarily tropical
Property response strength savannization dryland references references
Albedo Increasing with Expected Impact via Swann et al. 2012 Faria et al. 2018
loss of cover nonlinearities, alteration to energy (M), Bonan 2019 (O, c); von
and complexity possible weak budget. Increase (O), Jin et al. 2012 Randow et al.
allowing thresholds with savannization (O, g) 2004 (O, c);
surface to means mitigation Loarie et al.
reflect more of increasing 2011 (O, g);
solar radiation, sensible heat, Houghton 2018
but albedo may enhancing (M, g); de
decrease resilience Oliveira et al.
following fire 2016 (O, M, g);
because of de Oliveira
charring (Faria et al. 2019 (O,
et al. 2018) M, g); de
Oliveira and
Moraes 2013
(O, M, g); this
study
Net Decreasing with Expected With the increase of Morecroft et al. 1998 Gash and Nobre
radiation loss of cover nonlinearities, albedo after loss of (O, c); Anthoni 1997 (O, M);
(Rn) and complexity possible weak cover, total energy et al. 2000 (M, O); von Randow
(higher albedo) thresholds available in the Jin et al. 2012 (O, g) et al. 2004 (O);
canopy to be de Oliveira
partitioned into et al. 2016 (O,
sensible and latent M, g); de
heat will decrease, Oliveira et al.
while surface 2019 (O, M, g);
radiation and heat de Oliveira and
fluxes could Moraes 2013
increase. Altered (O, M, g),
microclimates Giambelluca
impact plants et al. 2009 (O,
c); this study
Bowen ratio, i.e., Potentially Nonlinear with Sensible heat Campbell and Restrepo-Coupe
sensible/latent, increasing as likely thresholds, increases can Norman 1998 (O); et al. 2013 (O,
heat flux sensible heat may be weak or increase VPD, Villegas et al. 2017 g); da Rocha
partitioning flux increases, strong promoting fire (O); Villegas et al. et al. 2009 (O,
and latent heat spread (Ray et al. 2014 (H) g); von Randow
flux decreases 2005, Cochrane et al. 2004 (O,
with loss of and Laurance c); de Oliveira
cover and 2008) and et al. 2019 (O,
complexity, but increasing forest M, g); this
more complex sensitivity. But study
relationship reduction in fuel
possible after fire can
reduce potential
for future fire
(Balch et al. 2008)
Near-ground Increasing with Nonlinear region of Higher soil Breshears 2006 (H, Galo et al. 1992
solar radiation cover and rapid change; evaporation effect g); Martens et al. (O); Bellingham
complexity loss moderate and hotter higher 2001 (M, g); Royer et al. 1996 (O);
threshold-type VPD near-ground et al. 2010 (O, g); de Oliveira
response. Weak or environments may Breshears and et al. 2016 (O,
more complex create a more Ludwig 2010 (O, g) M, g);
relationships with stressful Royer et al. 2012 Montgomery
tree size and environment for (O, g); Villegas and Chazdon
density in early/small growth et al. 2010a (O, g), 2001 (O, g); this
regeneration stages, promoting Villegas et al. 2010b study
(Montgomery and savannization (O, g)
Chazdon 2001)
 v www.esajournals.org 5 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
(Table 1. Continued.)
Feedback on future Representative Representative
Functional Threshold disturbance/ temperate primarily tropical
Property response strength savannization dryland references references
Surface roughness Increasing from Nonlinear and high An increase in Villegas et al. 2014 Tan et al. 2019
and boundary closed mature possibilities for boundary layer (H, g); Campbell (O, M, g);
layer canopy to thresholds, conductance over and Norman 1998 Dickinson and
conductance gappy open including in structure change (O); Lee and Kennedy 1992
forest, but may multiple response has the potential to Soliman 1977 (O); (M, c);
decrease regions, and increase VPD, Lee 1991a (O); Lee Spracklen and
moving to multimodality intensify water 1991b (O), Wolfe Garcia-Carreras
shorter statured stress, and increase and Nickling 1993 2015 (M, c); this
savanna; sensitivity to (O), as framed in study (via
possibly disturbance/ Breshears et al. roughness
peaked or savannafication. If 2009 (H, g); Sankey quantification)
multimodal a complex et al. 2013 (O, g)
relationship relationship, the
feedback will be
weaker/ limited
Notes: Source abbreviations are H, hypothesized; O, observations; M, modeled. Study type abbreviations are c, contrast of
two points; g, gradient of vegetation structure–function was considered.
characterized (Fig. 1D). These relationships We highlight relevant findings from dryland
could either decrease, increase, or intermediately and tropical research to illustrate this canopy
peak with decreasing forest canopy complexity structure–function framework including patterns
and cover; linear and nonlinear responses are that have been hypothesized, documented, and
possible within response types (Fig. 1E). predicted for a diverse range of properties such
In this conceptual framework, we draw on as albedo, net radiation (Rn), surface roughness,
the rich literature from forest–grassland con- and near-ground (below woody canopy) solar
tinua in temperate systems—often dryland gra- radiation patterns, among others (Table 1). For
dients from semi-arid grassland through semi- example, as woody plant cover decreases, albedo
arid forest (Breshears 2006, and references and near-ground solar radiation increase, which
therein, including Archer et al. 1988, Belsky increases the Bowen ratio (von Randow et al.
and Danham 1994, Scholes and Archer 1997, 2004, Villegas et al. 2014; citations in Table 1).
Breshears and Barnes 1999, Martens et al. 2001, Additionally, these patterns are often nonlinear,
House et al. 2003, Sankaran et al. 2004, 2005; potentially displaying a range of threshold-type
see also Wang et al. 2010, Villegas et al. 2014, rapid change responses (Fig. 1E), depending on
Villegas et al. 2015, Ratajczak et al. 2017)—that the degree of interactivity between vegetation
describe connections between microclimate, units (determining beneath canopy vs. inter-
components of energy balance, canopy cover, canopy space in Breshears 2006). Near-ground
and structural complexity. A central premise of solar radiation, for example, decreases nonlin-
this framework is that woody plants (shrubs early in plots with increasing canopy cover; in
and trees) have a disproportionately large influ- one instance, near-ground areas receiving the
ence on the microclimate and associated energy maximum amount of daily solar radiation in a
balance beneath and around them and that this semi-arid woodland made up more than a third
influence changes, often in nonlinear ways, of the ground surface at 21% tree cover, but
with the increasing prevalence of woody plants dropped to just a tenth of the area at 34% tree
and an elevated vegetation canopy. In wood- cover, and were completely absent by 41% tree
land and savanna systems, gradients in the cover (Martens et al. 2001). Maximum canopy
vegetation canopy can often be conceptualized height and vertical foliage distribution also
as changes in canopy cover, though other struc- impact near-ground radiation (Sankey et al. 2013,
tural factors may also be important (which we Villegas et al. 2014). Similarly, surface roughness
will elaborate in the context of tropical forest and associated wind flow category—impacting
below). mixing and canopy conductance (Bonan 2015)—
 v www.esajournals.org 6 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
change nonlinearly with increasing cover. While empirical and theoretical development (Wehr
natural vegetation canopy data is needed, one et al. 2017, Wu et al. 2018). However, the results
wind tunnel study using solid cylinders found presented above, and theoretical foundations
isolated wake flow at less than 14% cover, wake (Bonan 2008), suggest strong predictive quantita-
interference flow at 14–40% cover, and skimming tive links between forest canopy structure and
flow at >40% cover (Wolfe and Nickling 1993; function; with additional development, these
and see Breshears et al. 2009). Ultimately, struc- links should open transformative remote obser-
tural properties impact the full range of compo- vation data streams on degraded forest function.
nents of surface energy and its partitioning into Soil factors may also impact surface energy
sensible and latent heat fluxes. fluxes (Bonan 2008) and influence the distribu-
Gradient analyses considering elements of tion of savannas (Lloyd et al. 2008, Lehmann
canopy structure (e.g., cover or LAI) in relation et al. 2011) and canopy structure and function
to microclimate and energy balance conse- gradients (Schietti et al. 2016). Lehmann et al.
quences of these gradients, and forest change 2011, for example, identify soil fertility as a key
more broadly, remain relatively rare in Amazon factor influencing arid savanna forest transitions,
forests. However, a number of tropical site con- though no single fertility threshold was identi-
trast studies, experimental disturbance manipu- fied for savanna. Furthermore, soil structural fac-
lations, analyses tracking the after effects of tors such as sand content influence patterns of
uncontrolled disturbances, and modeling results soil water availability and plant water relations
offer insight on structure–function relationships (Sperry and Hacke 2002, Oliveras and Malhi
(Table 1). Investigations of the impact of forest 2016). Seasonally flooded forests are also at risk
structure gradients and change on albedo typi- of transitions to degraded savanna-like states as
cally contrast forest with highly altered vegeta- a result of high fire sensitivity and susceptibility
tion types canopies. For example, higher albedo due to superficial root mats and other factors
is typically found in pasture, crops, and cleared (Almeida et al. 2016, Flores et al. 2017). Together,
lands, relative to forest, at least after initial post- soil structure and nutrients likely impact the
fire increases in absorptive charred materials chances of long-term savannization following
subside (de Oliveira and Moraes 2013, de Oli- disturbance and may alter relations between veg-
veira et al. 2016, 2019, Faria et al. 2018). The etation structure and energy balance compo-
albedo of regenerating secondary forest appears nents, but more studies are needed.
to converge toward mature forest values (low Energy balance and forest structural attributes
values, i.e., ~0.1); however, seasonal differences have been shown to change with forest distur-
in albedo and net radiation appear linked to bance; however, not enough information is avail-
structure in these forests (de Oliveira and Moraes able for robust characterizations of the functional
2013, de Oliveira et al. 2016, 2019). Foundational relationships highlighted by our framework in
energy and materials flux network studies high- tropical forests. The seasonal dynamics of the
light large scale gradients in Amazon forest func- sensible to latent heat flux partitioning, the
tion, including variation in seasonal patterns and Bowen ratio, varies between wetter and drier for-
the roles of water vs. radiation limitation of pho- ests; dry forest and savanna exhibit strong
tosynthesis and latent and sensible heat fluxes increases in the Bowen ratio in the dry season,
(da Rocha et al. 2009, Restrepo-Coupe et al. while wet forest remains relatively insensitive to
2013). The broad gradients studied to date rainfall seasonality (da Rocha et al. 2009,
include major tropical forest biome transitions, Restrepo-Coupe et al. 2013). Brando et al. 2019
including forest to savanna and forest to the drier detail the impacts of 13 yr of fire impact and
cerrado vegetation, which may also be associated recovery in a controlled experiment in south cen-
with differences in canopy structure (Marselis tral Amazônia; contrasting flux measurements in
et al. 2018, Shao et al. 2019, Tang et al. 2019). Ulti- burned and unburned (control) plots, evapotran-
mately, discerning the detailed links between spiration (ET) was similar even though LAI in
canopy structure and ecosystem functions the control was 70% higher, potentially consis-
including surface energy dynamics remains an tent with a nonlinear functional response. A
area of emerging research requiring greater foundational study that imposed a drought
 v www.esajournals.org 7 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
experimentally for 5 yr in the central Amazon when disturbed forest is more susceptible to
found a 30% reduction in leaf area by the end of additional disturbance, and when ecological
measurements (Brando et al. 2008; see discussion recovery trajectories are altered. One of the most
below of our canopy resurvey of this study site). consequential feedbacks occurs when structural
Patterns of light transmission through the forest change promotes fire because of the capacity of
are also impacted by disturbance and local fire to profoundly reorganize ecosystem and
canopy structure gradients. Montgomery and community structure, including by converting
Chazdon (2001), working in Costa Rican mature wet forest to savanna (Malhi et al. 2009, Silv
erio
and secondary forests, found relationships et al. 2013), or secondarized low biomass forest
between stem densities and near-surface radia- comprised of pioneer trees (Barlow and Peres
tion transmittance in mature forest but not for 2008), or creating alternate regeneration path-
regenerating forest; also observing vertical ways (Mesquita et al. 2001, Norden et al. 2011).
heterogeneity in radiation environments, these Disturbances that open the forest canopy create
authors concluded that the multilayered hetero- conditions favorable to fire spread following
geneity of the canopy played a complex role in human-driven ignition, increasing boundary
radiation transmission dynamics. Little direct layer conductance, reducing humidity, and
work is available on the role of canopy structure increasing air speed and temperatures (Ray et al.
gradients in surface roughness and boundary 2005, Cochrane and Laurance 2008). Fire itself
layer conductances; models generally assume may cause these alterations, but an important
that deforestation will reduce roughness and caveat is that fires also reduce fuel loads, which
thus mixing (Lean and Warrilow 1989, Shukla has a negative feedback on fire spread in the near
et al. 1990, Spracklen and Garcia-Carreras 2015). term (Balch et al. 2008). Canopy opening distur-
While we do not consider larger scale climate bances such as fires promote grass establishment
processes in depth here, it is important to note which can facilitate future fire ignition and
that these processes contribute to several addi- spread, particularly if the newly established
tional near-surface atmosphere impacts includ- grass species are fire adapted (Silv
erio et al.
ing the local formation of clouds, aerosols, and 2013). The recurrence of burning promotes the
rainfall, which alter albedo and heating processes establishment of fire-resilient savanna species
in the atmosphere (Bonan 2015, Lagu€e et al. with thick bark that coexist with grasses, creating
2019). Deforestation generally decreases rainfall a positive feedback to fire susceptibility enhanc-
over the Amazon in model-based studies (Knox ing savannization (Ratnam et al. 2011, Hoffmann
et al. 2011, Davidson et al. 2012, Lawrence and et al. 2012).
Vandecar 2015, Spracklen and Garcia-Carreras The reduction of forest cover and structural
2015), including threshold-type nonlinear reduc- complexity from disturbance can promote addi-
tions and bioclimatic savannization in the south- tional canopy simplification through positive for-
ern and eastern Amazon (Pires and Costa 2013). est change feedback mechanisms. The loss of
Forest disturbance can have significant influ- cover and increased canopy openness can
ence on the future ecological state of the increase the vulnerability of trees not only to fire
impacted forest, including by enhancing the risk but also to wind throw during storms (Silv
erio
of savannization (Cochrane and Laurance 2008, et al. 2019). Increasing openness, light penetra-
Arag~ao and Shimabukuro 2010, Silva et al. 2018). tion, and understory vapor pressure deficit
Disturbance broadly impacts the landscape (VPD) were associated with edge effects in the
structure of forest and savanna mosaics, as well Biological Dynamics of Forests Fragments Project
as continua of structural gradients within vegeta- (BDFFP), which led to the establishment of light
tion types. For example, Soares-Filho et al. (2012) wooded faster growing and faster dying trees in
found that a deforestation-related feedback is the the years that followed fragmentation (Nasci-
primary driver of increased fire frequency in a mento and Laurance 2004, Almeida et al. 2019b).
data-driven model of a highly disturbed central Lianas may also increase after disturbance (Lau-
Amazonian landscape. Feedbacks are also linked rance et al. 2001) and can lead to additional tree
to microclimate and energy balance impacts fall (Hunter et al. 2015). Negative feedbacks
associated with forest structural change and arise between disturbance and canopy change may
 v www.esajournals.org 8 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
also be at play; particularly, disturbance surface properties associated with vegetation
enhances light environments in the lower strata structure, particularly those related to energy
of forests, promoting tree growth and the infill- balance and radiation fluxes (Villegas et al. 2017).
ing of gaps (via gap regeneration dynamics; Bro- Although sacrificing accuracy and temporal
kaw 1985), which can also suppress fire (Oliveras scope relative to long-term tower installation and
and Malhi 2016). Linking specific microclimate flux monitoring approaches, this approach offers
impacts and the enhancement of regeneration the potential to more rapidly assess energy bal-
potential in vertically structured canopies ance partitioning in multiple sites and potentially
remains an emerging area of research (Stark et al. multiple stages of degradation and forest-sa-
2015). However, the interplay of microclimates, vanna gradients. This capacity can help close
light-limited tree regeneration, and drought and gaps in the scope of assessment of the conse-
fire disturbances is not yet fully understood (see quences of forest structural change, offering field
also Oliveras and Malhi 2016), which precludes measurements in recently disturbed forest, and
more informed predictions about forest change degraded forests undergoing recovery or savan-
trajectories, including recovery and savanniza- nization. Ecosystem and atmospheric functions
tion, and associated longer term ecosystem assessed with this rapid approach were compa-
impacts (Fig. 1A). Recent work has suggested rable to long-term quantification of similar vege-
that tree functional composition change over suc- tation change impacts within the ecosystems
cession, in terms of differentiation on growth– studied (boreal and semi-arid North American
survival and stature–recruitment tradeoffs, can forest in Villegas et al. 2017), though additional
explain forest structural change, highlighting the validation and development may offer improve-
likely importance of community and demo- ment. Here, we applied this approach in Amazon
graphic dynamics in forest transitions (Ru€ger forests for the first time to obtain example com-
et al. 2020). Furthermore, understanding the parisons among sites, providing patterns consis-
specific long-term impacts of forest disturbances tent with our framework for the investigation of
on demographic components, recruitment, forest functional and structural gradients.
growth, and mortality is increasingly recognized To quantify structural and functional contrasts
as central to predicting climate forcing (McDow- across forest disturbance and savanna transi-
ell et al. 2020). tions, we estimated changes in structural com-
plexity (tree cover, LAI, tree size structure, and
SECTION 2: TROPICAL SITE CONTRASTS vertical leaf area heterogeneity) and energy bal-
WITHIN CANOPY STRUCTURE–ENERGY ance variables (surface roughness, Bowen parti-
GRADIENTS tioning ratio, and albedo) in paired simultaneous
measurement plots. The rapid assessment
A key challenge for understanding changes method detailed in Villegas et al. 2017, and in the
along forest structure gradients and their conse- context of this Amazon forest study, in Support-
quences for energy balance is the ability to ing Information Appendix S1, is based on the fol-
quickly detect and assess these changes across lowing components: (1) the establishment of
heterogeneous landscapes. This is particularly focal forest plots where tree surveys provided
relevant where changes from disturbance are diameter distributions and basal area; (2) the
occurring rapidly or have novel impacts relative measurement of canopy structure with a combi-
to long-term patterns of vegetation heterogene- nation of hemispherical photos (used to calculate
ity, such as in the central Amazon. The conven- direct site factor, DSF, a metric of incident radia-
tional approach to understanding these gradients tion) and a profiling lidar (PCL) to quantify the
is through the establishment of long-term tower- vertical and horizontal variation of canopy sur-
based eddy covariance flux monitoring networks face areas, of which leaves are the primary com-
(Baldocchi et al. 2001), which limits the scope of ponent (Parker et al. 2004); and, (3) the
assessment in large and quickly changing sys- quantification of surface energy dynamics with
tems. A recently developed rapid campaign incoming and outgoing above-canopy solar radi-
approach can supplement these long-term moni- ation from a net radiometer—and, in profile,
toring networks for the characterization of wind speed, humidity, and temperature—that
 v www.esajournals.org 9 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
utilized lightweight sensors and a portable mast roughness, derived from lidar data appeared
system secured above the canopy. From these higher in savanna/savannized forest (BDFFP sav-
measurements, we estimate Bowen ratio by annized forest, 2.92 m; Alter do Cha~o savanna,
assuming that vertical fluxes of sensible and 3.47 m), relative to fragments and regenerating
latent heat are proportional to vertical gradients forest (BDFFP regenerating forest, 2.74 m; Alter
of temperature and humidity (Shuttleworth forest fragment, 1.84 m), though reconstructed
2012). Soil heat flux is also measured in the tower areas were too small for spatially controlled sta-
footprint (Appendix S1). tistical comparison. Interestingly, tall vertically
We considered two study regions in the central heterogeneous canopies of old-growth forest
Amazon with gradients of structural complexity were more rugose (Manaus—Reserva Ducke,
and tree cover, including large-statured mature 4.79 m, 95% confidence intervals, CI 4.44–5.14;
rainforest, smaller-statured closed canopy forest Manaus—ZF2, 4.65 m, CI 4.16–5.13; Tapajo
s—
with disturbance histories, and open savanna (or K67, 7.55 m, CI 7.11–7.99; sites and PCL lidar
savannized forest) sites. Climate variables and, collection described in Stark et al. 2012), which
to the extent possible, soil conditions at a given highlights the possibility of multimodality of
study region were held constant. In the first rugosity over structural gradients. Similarities
region, near the confluence of the Tapajo
s and were apparent when comparing estimated verti-
Amazon rivers near the city of Santar
em (Par
a, cal leaf area density profiles; in particular, both
Brazil), in the village of Alter do Ch~ao, we con- gradients displayed tall vertically complex
ducted a rapid assessment of differences mature rainforest structure with maximum foli-
between natural savanna, nearby forest frag- age heights near 50 m, shorter intermediate com-
ments—here natural island-like patches—dis- plexity forests around 20 m tall with developed
turbed periodically by fire (Magnusson et al. upper canopy layers, and open savanna (or
2008), and mature rainforest 45 km away at the savanna-like) structure with a few trees or tree-
K67 site of the Tapajo
s National Forest (TNF; lets reaching between 10 and 15 m (Fig. 3). Both
where there is long-term tower-based eddy gradients displayed correspondence between
covariance flux monitoring; Saleska et al. 2003). stem diameter size structure and vertical leaf
We note that Alter do Ch~ao has sandy soil, while area structure (consistent with prior findings
soil is clay-rich in the TNF. In a second region Stark et al. 2012, 2015). Similar contrasts were
~50 km north of the city of Manaus (Amazônas, apparent comparing leaf area estimated from
Brazil), we investigated the impacts of recent for- lidar leaf area density profiles across the struc-
est clearing in a regrowing pasture with remnant tural gradients, wherein mature forests displayed
trees and in an ~20-yr-old secondary forest— LAI values around six, intermediate complexity
areas cleared and under study as part of the recovering forest and forest fragments, were
long-term BDFFP project (Laurance et al. 2002). slightly lower, with LAI closer to five, and savan-
We compared these disturbed areas to nearby nized/savanna sites were open-canopied with
(within 20 km) mature rainforest contiguous LAI between 0.5 and 2 (Table 2, Fig. 4). LAI and
with extensive undisturbed rainforest; here, all vertical patterns of leaf area were, thus, similar in
sites were characterized by clay-rich soil. Gradi- our two study regions, but structural differences
ents in forest structural complexity were multidi- were also apparent. Most notably, the BDFFP
mensional, but broadly similar in the two region displays a more developed mature forest
regions, as revealed by the data (and described upper canopy, as has been noted previously
below). Within each area, mature vs. disturbed (Stark et al. 2012). The historical savanna site of
forest contrasts were collected through simulta- Alter do Cha~o also appeared to differ relative to
neous deployment of rapid assessment towers the savannized forest/pasture of the BDFFP.
and instrumentation. Specifically, in the BDFFP, a thicket of a vine
Measurements revealed differences in the locally known as cipo
-de-fogo (genera Davilla
numbers and distributions of stem sizes, the ver- and Doliocarpus) formed around persistent
tical and horizontal structure of leaf area, and the woody vegetation, while tall herbaceous dicots
canopy surface layer, over vegetation contrasts dominated the vegetation at ground level. In
(Figs. 2, 3). Rugosity, a metric of surface contrast, Alter do Ch~ao was grass-dominated
 v www.esajournals.org 10 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Fig. 2. Three-dimensional canopy surfaces derived from ground-based lidar at (disturbed) Alter do Ch~ao for-
est fragment, and savanna sites (top) and Biological Dynamics of Forest Fragments Project (BDFFP) recovering
forest and savannized forest/pasture sites illustrate the impacts of forest disturbance on canopy height and rugos-
ity. Canopy height is higher in less disturbed sites, while rugosity may be higher in more disturbed sites (though
the sample size precludes statistical comparison of the rugosity metric calculated as the standard deviation of
canopy surface height).
and trees and shrubs were sparse, had open DSF (Royer et al. 2010, Villegas et al. 2010a, b)—
canopies, and represented fire-tolerant species increased in the low LAI open sites. Lower DSF,
typical of savanna (Miranda 1993, Magnusson albedo, and ground heat fluxes correspond with
et al. 2008); this translated to a lower LAI and higher available radiation/energy (for latent and
lower near-ground leaf area in Alter do Ch~ao. sensible heat flux; Rn - G) in the closed canopy
Canopy structural differences among sites had disturbed forest relative to the savanna/savan-
apparent consequences for radiation transmis- nized sites. Over the short observation window
sion, and other microclimate and surface energy of this study, we did not identify differences in
canopy functions. The progression of decreasing the sensible to latent heat fluxes (Bowen ratio).
LAI from mature to degraded to savannized for- However, available energy is partitioned to latent
est corresponded with patterns in other vari- heat with primacy over sensible heat when water
ables, in the same or opposing directions is available, as it was during our study. Since
(Table 2, Fig. 4). All contrasts were significant there was higher available net radiation in the
and consistent in direction over the comparable forest sites (particularly morning through mid-
forest structural gradients in the two regions, day, Fig. 5), corresponding with lower albedo
with the exception of the Bowen ratio (Table 2, and soil heat flux, the similar Bowen ratio values
Fig. 4). The proportion of potential near-ground mean that there were higher latent than sensible
solar radiation reaching the soil surface—the heat fluxes overall. Higher latent heat fluxes, and
 v www.esajournals.org 11 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Fig. 3. Vegetation structure—integration of stem diameter and lidar vertical vegetation information. Panels
from left to right move from mature tropical rainforest (control/comparison sites), to disturbed, and then
 v www.esajournals.org 12 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
(Fig. 3. Continued)
savanna/savannized forest. Top two rows present the Alter do Cha~o region gradient (mature forest comparison
in the Tapajo
s National Forest, K67 site), and the bottom two rows, the BDFFP gradient. Basal area profiles are
derived from the stem frequency distributions shown in each panel pair (pink). Frequency panels also include
forest type information. Lidar-derived leaf area density profiles are shown in green; ribbons display 95% confi-
dence intervals around mean values. Gray shaded regions are size class basal area plotted vs expected height of
the size group (Feldpausch et al. 2011, Stark et al. 2012), highlighting (partial) correspondence with vertical leaf
area profiles. Bottom panels for each region show the tree size frequency distributions (purple bars). Abbrevia-
tion regen. indicates regenerating.
net radiation differences, suggested greater (and suggesting a likely reduction of leaf area in
generally high) transfer of water to the atmo- upper strata at that time.
sphere in the forest fragments and regenerating Collectively, these examples highlight expected
forest relative to savanna and savannized forest. but to date rarely quantified patterns of canopy
Temporally and climatically coincident data were structural, functional change in tropical forests
not available from mature forest eddy covariance that relate directly to savannization and help
sites. Albedo values of the disturbed forest sites, reveal structure–energy relationships, though
however, were close to those reported for the expanding sampling over broader gradients is
old-growth forest sites (Arau
jo et al. 2002, von needed to quantify functional responses essential
Randow et al. 2004, Miller et al. 2011), suggesting for predictive ecosystem science (see Fig. 1).
that even with apparently slightly reduced LAI, Although these examples are insufficient to span
the low albedo of these closed canopy disturbed full gradients of canopy structure, they provide
sites was comparable to mature closed canopy initial insights into how such differences might
tropical forest. be placed within the proposed framework.
We also leveraged research infrastructure pre-
viously developed in the Seca floresta experi- SECTION 3: REFRAMING TROPICAL FOREST
ment at the TNF. We tested whether there were SAVANNIZATION, CHALLENGES AND PROSPECTS
persistent or even elevated canopy impacts 10 yr
after the cessation of 5 yr of experimental Studies in tropical forest structure and func-
drought from rain throughfall exclusion (Nep- tion can be particularly challenging due to vege-
stad et al. 2007, Brando et al. 2008). This contrast tation height, accessibility, and infrastructural
(pre- and post-drought) enabled us to consider limitations. Limited funding has further con-
temporal changes in addition to the structural strained studies of vegetation structure in Ama-
spatial contrasts illustrated above, to highlight zon Basin tropical forests. Despite the enormous
potential forest change feedback. From the end importance of the Amazon forests and concerns
of the manipulated drought in 2005 to our resur- about the potential for savannization—high-
vey in 2015, the LAI in the control plot did not lighted prominently in the recent IPCC assess-
change significantly, whereas the drought plot ment (IPCC 2013, 2014)—there are to date fewer
had a reduced LAI in 2005 that recovered to con- than 10 flux towers currently operating in this
trol levels by 2015 (Fig. 6, top row; 2015 control region. Flux towers are extremely valuable for
mean 5.64 with 95% CI 5.34–5.94, and experi- studying differences between forest types over
mental 5.71 with CI 5.45–5.97). We were also able major vegetation gradients, including forest vs.
to confirm that this recovery was complete in savanna contrasts (da Rocha et al. 2009,
terms of the height profile of vegetation, with Restrepo-Coupe et al. 2013), and in rare cases
very similar leaf area profiles occurring in the over local vegetation contrasts (Brando et al.
control and the drought plots in 2015 (Fig. 6, bot- 2019). However, distant comparisons over major
tom row). We note that though vegetation profile gradients are confounded by climate and varia-
data are not available from the experimental per- tion in vegetation types (functional composition)
iod, higher mortality of large trees in the drought when used as proxies to investigate changes that
plot was reported (Nepstad et al. 2007), might occur in the process of savannization.
 v www.esajournals.org 13 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Table 2. Means and 95% confidence intervals (parentheses) for structural, energy balance, and associated micro-
climate variables in forest contrasts.
BDFFP/Manaus Alter do Ch~ao/Santar
em
Region
Forest Mature/ Degraded/ Mature/
type old-growth secondary Savannized old-growth Degraded Savanna
LAI 6.20 (6.05–6.35) 4.60 (4.26–4.93) 1.25 (0.93–1.57) 5.73 4.97 (4.69–5.25) 0.23 (0.16–0.31)
(5.57–5.89)
DSF 0.26 (0.24–0.28) 0.64 (0.57–0.72) 0.32 (0.00–0.65) 0.93 (0.90–0.95)
Rn-G 104.42 37.21 (1.30–73.12) 144.15 31.94 (0.39–63.50)
(W/m2⋅s) (80.17–128.68) (102.62–185.68)
Bowen 0.13 (0.05–0.21) 0.72 (0.22–1.21) 1.86 (0.93–2.80) 2.77 (1.31–4.24)
Ratio
Albedo 0.100 (0.097–0.102) 0.165 (0.160–0.171) 0.092 (0.090–0.095) 0.143 (0.135–0.150)
Note: See Fig. 4.
Furthermore, flux network comparisons lack per- field campaigns such as the one presented, but
spectives on local vegetation structure gradients expanded beyond the pilot demonstration to
associated with disturbance (the local flux con- cover more complete canopy structure gradients
trasts offered by the Fazenda Tanguro fire experi- (e.g., related to age of regeneration, impact level,
ment are an exception to these generalizations; or disturbance type), while also controlling for
Brando et al. 2019). Disturbed and degraded for- environmental and climate variation. The objec-
ests account for a large fraction of Amazon Basin tive of field campaigns would be to quantify the
—recently reported by Bullock et al. 2020 as 17% functional relationships between elements of
of the ecoregion in 2017 (in contrast to 11% of canopy structure and surface energy balance
Amazon forest cleared for agriculture, and 52% and, ideally, to evaluate additional environmen-
higher than previously estimated)—while sec- tal and biogeographic dependencies, needed to
ondary growth may account for 28% of Neotrop- transform Earth system approaches. With a lar-
ical forest (Chazdon et al. 2016). Degraded forest ger investment, this could also be studied with a
types represent a significant risk for savanniza- network of flux towers, which would provide
tion (Cochrane and Laurance 2008, Arag~ao and additional longer term information on energy,
Shimabukuro 2010), which could have critical water, and carbon fluxes. The Fazenda Tanguro
impacts, including lowering the Amazon forest fire experiment contrasting paired fire impacted
carbon sink, altering global temperature (Cox and control forests partially addresses this gap
et al. 2004, Malhi et al. 2008), and altering other (Brando et al. 2019). There is a critical need, how-
climate patterns (Garcia et al. 2016). ever, to move beyond contrasts toward under-
Gradient studies are a critical source of ecolog- standing of gradients and functional responses,
ical information in tropical forests, as highlighted for which a range of structural states must be
by the BDFFP project, which investigates frag- considered (Berenguer et al. 2014, 2018). As with
ment size and distance-to-edge effects (Nasci- temperate grassland–forest gradients, many
mento and Laurance 2004, Almeida et al. 2019b). microclimate and associated energy balance
However, to date, there have not been specifi- impacts of forest structural change will display
cally designed and implemented studies that nonlinear functional responses, including
focus on canopy structure gradients of degraded decreasing, increasing or peaked relationships
tropical forest and secondary growth at risk of (Fig. 1; Breshears 2006). Nonlinearities and
savannization and the associated relationships thresholds in functional responses of energy bal-
between canopy structure, microclimate, and ance, and related microclimates, may determine
surface energy dynamics. Our proposed frame- whether a site or region will transition to a
work (Fig. 1) could help guide the development savanna or recovered forest state over the long
and installation of one or more gradients to term (Fig. 1; Staver et al. 2011, Hoffmann et al.
specifically evaluate canopy structure–energy 2012, da Silva Junior et al. 2019). In the future,
relationships. This could be accomplished with combinations of targeted gradient studies and
 v www.esajournals.org 14 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
networks of comparable field contrasts, with
detailed structural predictor and energy and
microclimate response variable measurements,
can directly reveal thresholds and functional
responses, and their associated dependencies.
Canopy structure–energy relationships in trop-
ical forests are essential to global climate func-
tion; tropical forests are not simply important for
their carbon sequestration value (IPCC 2014).
Rather, as we have illustrated (Figs. 2–6),
changes in canopy structure have pronounced
impacts on energy balance and associated local
microclimate. This has implications for locally
mediated, and large scale, vegetation change
feedbacks. First, at the local scale, sites with less
cover and simplified structure create a hotter
microclimate that can impact understory vegeta-
tion (Nascimento and Laurance 2004, Allen et al.
2015). In harsher microclimates, stress may be
amplified for remaining trees, potentially trigger-
ing additional tree mortality and creating envi-
ronments favorable to fire ignition and spread
(Nascimento and Laurance 2004, Ray et al. 2005,
Allen et al. 2015). Next, widespread vegetation
change impacts on local climates alter large scale
temperature, precipitation, and pressure gradi-
ents (Spracklen and Garcia-Carreras 2015, Ville-
gas et al. 2015), which may cause large scale
vegetation feedbacks (Friedlingstein et al. 2014).
Changes in climate within one subregion can
also alter climate teleconnections and thereby
impact vegetation elsewhere—termed an ecocli-
Fig. 4. Structural, energy balance, and associated mate teleconnection (Swann et al. 2012, Stark
microclimate variables over gradients of decreasing et al. 2016). Thus, if savannization is sufficiently
structural complexity and cover. Horizontal lines are extensive, it can potentially affect energy balance
means, while shaded bars are estimated 95% confi- and local climate in both the region of tree loss as
dence regions. All contrasts are significant at the well as in other regions via climate connections
P < 0.05 level by t-tests (or comparison of confidence (Swann et al. 2012, Medvigy et al. 2013, Garcia
region in the case of LAI), excepting Bowen ratios, et al. 2016, Stark et al. 2016, Molina et al. 2019).
which did not appear to differ. Leaf area index was Structure–energy relationships appear integral
estimated from lidar and, in this case, included the to the sensitivity and resilience of tropical forest
mature forest site data for comparison. LAI and vege- to structural change from droughts, heatwaves,
tation structure impacts on incoming solar radiation fires, deforestation, and other increasing environ-
are illustrated by direct site factor (DSF; proportion of mental disturbances at local scales, because of
annual direct solar radiation that reaches the surface microclimate alteration and the influence of
relative to the open-sky expectation)—and here, higher microclimates on disturbance impacts. However,
DSF may have contributed to higher albedo, with the we generally lack rigorous quantitative sensitiv-
ground more reflective than leaf area. Rn-G, available ity and resilience relationships that can translate
net radiation, and other energy balance components how a particular forest state will respond to a
are taken from half-hour mean values between 10 a.m. particular disturbance, under a given climate. In
and 2 p.m. in two observed daily cycles. See Table 2. this context, the unresolved sensitivity and
 v www.esajournals.org 15 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Fig. 5. Diurnal time-series of available energy at the surface (Rn-G) contrasted over forest structural gradients in
the central Amazon. Red stars represent savanna or savannized sites, while blue triangles are disturbed but closed
canopy forest sites. Note that the averages of these diurnal patterns are the values presented in Table 2, Fig. 4.
resilience of tropical forest to droughts, heat-
waves, and fires significantly limits Earth System
model forecasting of climate change (IPCC 2013,
2014, Friedlingstein et al. 2014, Kloster and Lass-
lop 2017, Fisher et al. 2018). To address this
uncertainty, we advance here the specific hypoth-
esis that canopy structure, surface energy bal-
ance, and the sensitivity of tropical forest to
droughts, heatwaves, and fires are mechanisti-
cally interrelated by the impacts of canopy struc-
ture on vegetation microclimates, and the roles
of microclimates in vegetation disturbance (e.g.,
impacting drought mortality and fire spread;
Nascimento and Laurance 2004, Ray et al. 2005,
Allen et al. 2015). It follows that the relationships
explored in our framework (Fig. 1, Table 1)
could then be extended to predict sensitivities,
and that data on both forest canopy structure
and energy balance would improve this predic-
tion (Fig. 7). To expand our conceptual frame-
work to include sensitivity, we consider two
forest change time steps, an initial degradation
event driven by deforestation or another distur-
bance (Fig. 7 left), and a second step that could
Fig. 6. Temporal contrast addressing forest resili- include forest recovery, or continued degradation
ence/sensitivity. Forest recovery 10 yr after the Seca (Fig. 7 right). The outcome of this second time
Floresta drought throughfall experiment (Nepstad step—recovery or collapse—hinges on forest sen-
et al. 2007). Drought is red and control, blue. LAI of sitivity to the disturbances likely to be present,
drought plot in 2005 reported in Brando et al. 2008 falls including drought and fire. Thus, as we have
below resampling-determined LAI confidence intervals hypothesized, a first feedback to change is possi-
(2015) suggesting signi cant recovery (denoted as 
fi ). ble from the local-scale impacts of degradation
 v www.esajournals.org 16 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Fig. 7. Hypothesis and proposed research program to address the sensitivity and resilience of tropical forest to
future change from structure, energy balance, and microclimate relationships. DT1 is a time step of initial distur-
bance, such as deforestation, that causes an old-growth tropical forest to become secondarized or otherwise
degraded, with changing energy partitioning (blue to red gradient, with more red indicating more savanna-like
energy characteristics). In this time step, rapid field assessment and remote observation of forest structure (lidar)
and energy balance (multispectral and thermal approaches) offer the development of quantitative models linking
structure and energy balance. In the DT2 second time step, forests may recover, or collapse and be further
degraded from additional disturbances. Initiated by the disturbance in time step 1, two feedback pathways medi-
ated by forest sensitivity and resilience may affect forest transitions in the second time step. The first, we hypoth-
esize, is a local-scale disturbance feedback wherein forest structure and energy balance influence forest
sensitivity to drought, heatwave, and fire disturbance because of cross-linking microclimate relationships (e.g.,
open canopies are drier and hotter, facilitating fire). The second feedback is the well-known large scale vegeta-
tion–climate feedback arising when widespread energy balance changes affect climate gradients and processes
that in turn impact the forest. Remote and field monitoring over this second time step would test the disturbance
feedback hypothesis and allow quantitative model development of forest disturbance sensitivity relationships,
that should also account for climate and functional traits.
on disturbance sensitivity. The larger scale vege- emission components of energy balance are
tation–atmosphere interactions typically already directly observable at fine (few meter)
addressed in Earth System models (Friedling- scales from remote sensing technologies (lidar
stein et al. 2014) represent a second feedback. In and multispectral radiometric imaging respec-
this case, the feedback results from regional alter- tively). From these raw remote observations,
ations of temperature and precipitation, which approaches to estimate surface and canopy
impacts the forest via its sensitivity to variations microclimates (Zellweger et al. 2019), tree size-
in these factors. Other factors influencing forest structured dynamics (Stark et al. 2012, 2015,
sensitivity and resilience to disturbance such as Smith et al. 2019), and full surface energy parti-
forest functional composition (Brum et al. 2019, tioning (de Oliveira et al. 2016) have been elabo-
Ru€ger et al. 2020) would also play a critical role. rated. Thus, an implication of our framework is
Advanced fine-grained remote sensing may that, if correct, advances in remote sensing will
allow for the rapid application of this frame- open a wide new avenue to monitor both atmo-
work to reduce uncertainty in tropical forest spheric forcing relevant to predicting large scale
transitions (highlighted in Fig. 7). The quantity climate feedbacks and changing forest sensitivi-
and arrangement of canopy surface areas, short- ties, relevant to predicting local-scale distur-
wave albedo, and thermal near infrared bance sensitivity feedbacks (Fig. 7). Field
 v www.esajournals.org 17 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
measurements such as those presented in this canopy cover (Asner et al. 2010, Tyukavina et al.
article will also be essential to develop faster 2017, Almeida et al. 2019b). Because savanniza-
high-throughput remote observation-based tion of tropical wet forests is of such concern
approaches to quantify structure, surface energy both locally and globally (IPCC 2014), we need
balance, and microclimates across forest degra- to move past historical savanna and forest con-
dation and savannization gradients. trasts to the huge areas at risk of savannization,
In the context of changes related to forest resili- to understand threshold type and nonlinear
ence (Fig. 7 right), the temporal change we docu- response to forest gradients associated with sav-
mented at the Seca Floresta experimental annization and other types of forest change.
drought study (Fig. 6) provides one example of Given that changes in tree cover and structure
forest recovery. In that study, rain throughfall have large impacts on energy balance and
was excluded from 1 ha of forest for 5 yr; our ecosystem properties, there is an urgent need to
resurvey 10 yr after cessation showed that there quantify these properties not only for primary
was no apparent impact on leaf area—that leaf forests, but also for forests with lower, less com-
area had recovered from an initially reported plex cover and structure. The proposed frame-
30% loss—and that there were no discernable work (Table 1, Fig. 1), along with the recently
differences in the vertical canopy structure. This developed field campaign protocol, provides a
forest experiences significant seasonal drought, means for achieving this broader characteriza-
with a likely stronger than average drought tion. In addition, recent advances in remote sens-
event in 2010 (Lewis et al. 2011; though, we note ing offer new opportunities to quantify and link
this region did not experience the severe drought canopy structure to components of energy bal-
of others), but, nevertheless, was able to recover. ance, and forest sensitivity and recovery from
This may be consistent with findings of Brando disturbance (Fig. 7)—these include multispectral
et al. (2008) in the initial study: The forest based approaches to mapping components of
appeared to shift allocation to maintain leaf area energy balance, microclimates, canopy functional
(estimated with litter production) during the parameters, and airborne and spaceborne lidar
drought, at the expense of wood growth. Fur- for forest three-dimensional canopy structure
thermore, wood growth appeared to recover characterization, including the new GEDI mis-
quickly in the first year after drought cessation. sion (Chambers et al. 2007, Stark et al. 2012, 2015,
More recently, detailing responses to experimen- de Oliveira et al. 2016, Shao et al. 2019, Tang
tal fire in the southern central Amazon, Brando et al. 2019, Zellweger et al. 2019). Further, many
et al. (2019) describe a surprising resilience of forest disturbance factors including intensifica-
function in forest undergoing apparent savan- tion of ENSO, and drought generally, as well as
nization—even with a 70% decrease in leaf area, direct human alterations, and related fires are
fire impacted forests were able to maintain simi- currently increasing in the Amazon (Davidson
lar transpiration as control vegetation. This high- et al. 2012). Smoke pollution from land use
lights the need to include not just physical change related fires also causes widespread
structural information but also vegetation func- adverse respiratory health impacts, which in
tional characterization. A more complete under- 2020 may worsen the COVID-19 pandemic (de
standing of structure–energy balance change and Oliveira et al. 2020). The much publicized 2019
forest resilience would link variation in leaf func- crisis of Amazon fire and land conversion (Esco-
tion and water strategies over canopy strata and bar 2019), and emerging 2020 crisis (de Oliveira
canopy microenvironments with the risks of fire, et al. 2020), in this light represent a clear call for
mortality, and other factors under droughts and action to safeguard human health and Amazon
heatwaves (McDowell et al. 2018, Smith et al. forest climate function, including by better
2019, Brum et al. 2019). understanding the factors that could contribute
In conclusion, global change-related droughts, to restoration and recovery of high biomass trop-
deforestation, and fire are rapidly converting ical forest, instead of long-term savannization
previously structurally complex mature forests (and see Barlow et al. 2020). There is need to both
into a matrix of degraded states that range in ver- quantify the consequences of degradation and
tical and horizontal variation of leaf area and understand feedbacks to longer term forest
 v www.esajournals.org 18 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
change, particularly to track and monitor ecolog- helped develop and implement the study and partici-
ical trajectories that can provide early warning pated in manuscript development. MN Smith,
for forest state transitions from high biomass for- DRAlves de Almeida, G de Oliveira conducted analy-
est to low biomass and high sensible heat ux sis, and contributed substantial conceptual develop-fl
savanna. Ultimately, energy and carbon budgets ment to the manuscript.
of disturbed and degraded forest types, and their
sensitivities to forest state transitions, must be LITERATURE CITED
included in ecosystem model forecasts of cou- Allen, C. D., D. D. Breshears, and N. G. McDowell.
pled vegetation–climate change. 2015. On underestimation of global vulnerability to
tree mortality and forest die-off from hotter
ACKNOWLEDGMENTS drought in the Anthropocene. Ecosphere 6:1–55.
Almeida, D. R. A., et al. 2019a. The effectiveness of
This work was primarily supported by NSF lidar remote sensing for monitoring forest cover
Macrosystems Biology EF-1340604 to Michigan State attributes and landscape restoration. Forest Ecol-
University, EF-1340624 to University of Arizona, and ogy and Management 438:34–43.
EF-1340649 to University of Washington. Additional Almeida, D. A., et al. 2019b. Persistent effects of frag-
support was provided for SC Stark by NASA award mentation on tropical rainforest canopy structure
#NNX17AF65G and NSF-DEB-1754357, EF-1550686, after 20 years of isolation. Ecological Applications
and DEB-1950080, 1702379 awards and by USDA 29:e01952.
NIFA. DD Breshears and DJ Law were supported by Almeida, D. A., B. W. Nelson, J. Schietti, E. B. Gorgens,
the Arizona Ag Experiment Station (AZRT-1390130- A. F. Resende, S. C. Stark, and R. Valbuena. 2016.
M12-222) and NSF-DEB 1550756 and 1824796 (DDB). Contrasting fire damage and fire susceptibility
DRA Almeida was supported by the S~ao Paulo between seasonally flooded forest and upland for-
Research Foundation (#2018/21338-3 and #2019/14697- est in the Central Amazon using portable profiling
0). S Aragon was supported by a CNPq PDJ grant LiDAR. Remote Sensing of Environment 184:153–
(150827/2013-0) and a CAPES-PNPD post-doctoral 160.
scholarship. POPA-PELD (Long Term Ecological Mon- Anthoni, P. M., B. E. Law, M. H. Unsworth, and R. J.
itoring of Western Para
) establishment was supported Vong. 2000. Variation of net radiation over hetero-
by the Programa Piloto para a Prote
c~ao das Florestas geneous surfaces: measurements and simulation in
Tropicais do Brasil– INPA/FINEP/European Union a juniper–sagebrush ecosystem. Agricultural and
64.00.0021.00, INPA-PPI-1-3010, the Programa de Pes- Forest Meteorology 102:275–286.
quisa em Biodiversidade (PPBio) and by the Centro de Arag~ao, L. E., and Y. E. Shimabukuro. 2010. The inci-
Estudos Integrados da Biodiversidade Amazônica dence of fire in Amazonian forests with implica-
(CENBAM), INPA and CNPq grant (441443/2016-8). tions for REDD. Science 328:1275–1278.
JCV received support from the Proyecto “Trayectorias Arau
jo, A. C., et al. 2002. Comparative measurements
de sistemas socio-ecolo
gicos y sus determinantes en of carbon dioxide fluxes from two nearby towers in
cuencas estrat
egicas en un contexto de cambio ambien- a central Amazonian rainforest: the Manaus LBA
tal. Co
digo 110180863961” Convocatoria 808-2018 site. Journal of Geophysical Research: Atmo-
Proyectos de ciencia, tecnolog
ıa e innovacio
n y su con- spheres. 107:LBA-58.
tribucio
n a los retos de pa
ıs- Colciencias. LEOC Ara- Archer, S., C. Scifres, C. R. Bassham, and R. Maggio.
ga~o was supported by CNPq 305054/2016-3, FAPESP 1988. Autogenic succession in a subtropical
2018/15001-6. We also wish to thank the Researchers savanna: conversion of grassland to thorn wood-
and staff of the Biological Dynamics of Forest Frag- land. Ecological Monographs 58:111–127.
ments Project (BDFFP), the Large Scale Biosphere- Asner, G. P., et al. 2010. High-resolution forest carbon
Atmosphere Experiment in the Amazon (LBA, in San- stocks and emissions in the Amazon. Proceedings
tar
em and Manaus), and PPBio of the National Insti- of the National Academy of Sciences USA
tute for Amazonian Research (INPA, Brazil), and the 107:16738–16742.
Federal University of Western Para
 (UFOPA) for Balch, J. K., D. C. Nepstad, P. M. Brando, L. M. Curran,
exceptional support. This is study 759 of the BDFFP O. Portela, O. de Carvalho Jr, and P. Lefebvre. 2008.
Technical Series. Author contributions: SC. Stark led Negative fire feedback in a transitional forest of
the study and manuscript development, DD. Bres- southeastern Amazonia. Global Change Biology
hears, S Arago
n, JC Villegas, DJ Law, DM Minor, RL 14:2276–2287.
de Assis, SR Saleska, ALS Swann, JMS Moura, R da Baldocchi, D., et al. 2001. FLUXNET: a new tool to
Silva, LEOC Arag~ao, JL Camargo, and RC Oliveira study the temporal and spatial variability of
 v www.esajournals.org 19 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
ecosystem-scale carbon dioxide, water vapor, and moisture heterogeneity for semiarid landscapes
energy flux densities. Bulletin of the American within the grassland/forest continuum: a unified
Meteorological Society 82:2415–2434. conceptual model. Landscape Ecology 14:465–478.
Barlow, J., E. Berenguer, R. Carmenta, and F. Fran
ca. Breshears, D. D., and J. A. Ludwig. 2010. Near-ground
2020. Clarifying Amazonia’s burning crisis. Global solar radiation along the grassland–forest contin-
Change Biology 26:319–321. uum: Tall-tree canopy architecture imposes only
Barlow, J., and C. A. Peres. 2008. Fire-mediated die- muted trends and heterogeneity. Austral Ecology
back and compositional cascade in an Amazonian 35:31–40.
forest. Philosophical Transactions of the Royal Soci- Breshears, D. D., J. J. Whicker, C. B. Zou, J. P. Field,
ety B: Biological Sciences 363:1787–1794. and C. D. Allen. 2009. A conceptual framework for
Bellingham, P. J., E. V. J. Tanner, P. M. Rich, and T. C. dryland aeolian sediment transport along the
R. Goodland. 1996. Changes in light below the grassland–forest continuum: effects of woody plant
canopy of a Jamaican montane rainforest after a canopy cover and disturbance. Geomorphology
hurricane. Journal of Tropical Ecology 12:699–722. 105:28–38.
Belsky, A. J., and C. D. Danham. 1994. Forest gaps and Brokaw, N. V. 1985. Gap-phase regeneration in a tropi-
isolated savanna trees. BioScience 44:77–84. cal forest. Ecology 66:682–687.
Berenguer, E., J. Ferreira, T. A. Gardner, L. E. O. C. Brum, M., M. A. Vadeboncoeur, V. Ivanov, H. Asb-
Arag~ao, P. B. De Camargo, C. E. Cerri, M. Durigan, jornsen, S. Saleska, L. F. Alves, D. Penha, J. D. Dias,
R. C. D. Oliveira, I. C. G. Vieira, and J. Barlow. L. E. Arag~ao, F. Barros, and P. Bittencourt. 2019.
2014. A large-scale field assessment of carbon Hydrological niche segregation defines forest
stocks in human-modified tropical forests. Global structure and drought tolerance strategies in a sea-
Change Biology 20:3713–3726. sonal Amazon forest. Journal of Ecology 107:318–
Berenguer, E., Y. Malhi, P. Brando, A. Cardoso Nunes 333.
Cordeiro, J. Ferreira, F. Fran
ca, L. Chesini Rossi, M. Bullock, E. L., C. E. Woodcock, C. Souza Jr., and P.
Maria Moraes de Seixas, and J. Barlow. 2018. Tree Olofsson. 2020. Satellite-based estimates reveal
growth and stem carbon accumulation in human- widespread forest degradation in the Amazon.
modified Amazonian forests following drought Global Change Biology 26:2956–2969.
and fire. Philosophical Transactions of the Royal Campbell, G. S., and J. M. Norman. 1998. An introduc-
Society B: Biological Sciences 373:20170308. tion to environmental biophysics. Second edition.
Bonan, G. B. 2008. Forests and climate change: forc- Springer, New York, New York, USA.
ings, feedbacks, and the climate benefits of forests. Chambers, J. Q., G. P. Asner, D. C. Morton, L. O.
Science 320:1444–1449. Anderson, S. S. Saatchi, F. D. Esp
ırito-Santo, M.
Bonan, G. 2015. Ecological climatology: concepts and Palace, and C. Souza Jr. 2007. Regional ecosystem
Applications. Cambridge University Press, Cam- structure and function: ecological insights from
bridge, UK. remote sensing of tropical forests. Trends in Ecol-
Bonan, G. 2019. Climate change and terrestrial ecosys- ogy & Evolution 22:414–423.
tem modeling. Cambridge University Press, Cam- Chazdon, R. L. 2014. Second growth: the promise of
bridge, UK. tropical forest regeneration in an age of deforesta-
Brando, P. M., D. C. Nepstad, E. A. Davidson, S. E. tion. University of Chicago Press, Chicago, Illinois,
Trumbore, D. Ray, and P. Camargo. 2008. Drought USA.
effects on litterfall, wood production and below- Chazdon, R. L., et al. 2016. Carbon sequestration
ground carbon cycling in an Amazon forest: results potential of second-growth forest regeneration in
of a throughfall reduction experiment. Philosophi- the Latin American tropics. Science Advances 2:
cal Transactions of the Royal Society B: Biological e1501639.
Sciences 363:1839–1848. Cochrane, M. A., and W. F. Laurance. 2008. Syner-
Brando, P. M., et al. 2019. Prolonged tropical forest gisms among fire, land use, and climate change in
degradation due to compounding disturbances: the Amazon. Ambio 37:522–527.
implications for CO2 and H2O fluxes. Global Cox, P. M., R. A. Betts, M. Collins, P. P. Harris, C. Hunt-
Change Biology 25:2855–2868. ingford, and C. D. Jones. 2004. Amazonian forest
Breshears, D. D. 2006. The grassland–forest contin- dieback under climate-carbon cycle projections for
uum: trends in ecosystem properties for woody the 21st century. Theoretical and Applied Climatol-
plant mosaics? Frontiers in Ecology and the Envi- ogy 78:137–156.
ronment 4:96–104. Curtis, P. G., C. M. Slay, N. L. Harris, A. Tyukavina,
Breshears, D. D., and F. J. Barnes. 1999. Interrelation- and M. C. Hansen. 2018. Classifying drivers of glo-
ships between plant functional types and soil bal forest loss. Science 361:1108–1111.
 v www.esajournals.org 20 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
da Rocha, H. R., et al. 2009. Patterns of water and heat forest resilience. Proceedings of the National Acad-
flux across a biome gradient from tropical forest to emy of Sciences USA 114:4442–4446.
savanna in Brazil. Journal of Geophysical Research. Friedlingstein, P., M. Meinshausen, V. K. Arora, C. D.
Biogeosciences 114:G00B12. Jones, A. Anav, S. K. Liddicoat, and R. Knutti.
da Silva Junior, C. A., G. de Medeiros Costa, F. S. 2014. Uncertainties in CMIP5 climate projections
Rossi, J. C. E. Vale, R. B. de Lima, M. Lima, J. due to carbon cycle feedbacks. Journal of Climate
F. de Oliveira-Junior, P. E. Teodoro, and R. C. 27:511–526.
Santos. 2019. Remote sensing for updating the Galo, A. T., P. M. Rich, and J. J. Ewel. 1992. Effects of
boundaries between the Brazilian Cerrado-Amazo- forest edges on the solar radiation regime in a ser-
nia biomes. Environmental Science & Policy. ies of reconstructed tropical ecosystems. Pages 98–
101:383–392. 108 in American Society for Photogrammetry and
Davidson, E. A., et al. 2012. The Amazon basin in tran- Remote Sensing Annual Meeting. Albuquerque,
sition. Nature 481:321. New Mexico, USA.
de Oliveira, G., N. Brunsell, E. Moraes, G. Bertani, T. Garcia, E. S., A. L. Swann, J. C. Villegas, D. D. Bres-
dos Santos, Y. Shimabukuro, and L. Arag~ao. 2016. hears, D. J. Law, S. R. Saleska, and S. C. Stark. 2016.
Use of MODIS sensor images combined with Synergistic ecoclimate teleconnections from forest
reanalysis products to retrieve net radiation in loss in different regions structure global ecological
Amazonia. Sensors 16:956. responses. PLOS ONE 11:e0165042.
de Oliveira, G., N. A. Brunsell, E. C. Moraes, Y. E. Shi- Gash, J. H. C., et al. 2004. Amazonian climate: results
mabukuro, T. V. dos Santos, C. von Randow, R. G. and future research. Theoretical and Applied Cli-
de Aguiar, and L. E. Aragao. 2019. Effects of land- matology 78:187–193.
cover changes on the partitioning of surface energy Gash, J. H. C., and C. A. Nobre. 1997. Climatic effects
and water fluxes in Amazonia using high-resolu- of Amazonian deforestation: some results from
tion satellite imagery. Ecohydrology 12:e2126. ABRACOS. Bulletin of the American Meteorologi-
de Oliveira, G., J. M. Chen, S. C. Stark, E. Berenguer, P. cal Society 78:823–830.
Moutinho, P. Artaxo, L. O. Anderson, and L. E. O. Giambelluca, T. W., F. G. Scholz, S. J. Bucci, F. C. Mein-
C. Araga~o. 2020. Smoke pollution’s impacts in zer, G. Goldstein, W. A. Hoffmann, A. C. Franco,
Amazonia. Science 369:634–635. and M. P. Buchert. 2009. Evapotranspiration and
de Oliveira, G., and E. C. Moraes. 2013. Validation of energy balance of Brazilian savannas with contrast-
net radiation obtained from MODIS/TERRA data ing tree density. Agricultural and Forest Meteorol-
in Amazonia with LBA surface measurements. ogy 149:1365–1376.
Acta Amazonica 43:353–363. Hirota, M., M. Holmgren, E. H. Van Nes, and M.
Desjardins, T., A. Mariotti, C. Girardin, and A. Chauvel. Scheffer. 2011. Global resilience of tropical forest
1996. Changes of the forest-savanna boundary in and savanna to critical transitions. Science 334:232–
Brazilian Amazonia during the Holocene revealed 235.
by stable isotope ratios of soil organic carbon. Hirota, M., C. Nobre, M. D. Oyama, and M. M. Busta-
Oecologia 108:749–756. mante. 2010. The climatic sensitivity of the forest,
Dickinson, R. E., and P. Kennedy. 1992. Impacts on savanna and forest–savanna transition in tropical
regional climate of Amazon deforestation. Geo- South America. New Phytologist 187:707–719.
physical Research Letters 19:1947–1950. Hoffmann, W. A., E. L. Geiger, S. G. Gotsch, D. R. Ros-
Escobar, H. 2019. Amazon fires clearly linked to defor- satto, L. C. Silva, O. L. Lau, M. Haridasan, and A.
estation, scientists say. Science 365:853. C. Franco. 2012. Ecological thresholds at the
Faria, T. D. O., T. R. Rodrigues, L. F. A. Curado, D. C. savanna-forest boundary: how plant traits,
Gaio, and J. D. S. Nogueira. 2018. Surface albedo in resources and fire govern the distribution of tropi-
different land-use and cover types in Amazon for- cal biomes. Ecology Letters 15:759–768.
est region. Revista Ambiente & A
gua 13:1. Houghton, R. A. 2018. Interactions between land-use
Feldpausch, T. R., et al. 2011. Height-diameter allome- change and climate-carbon cycle feedbacks. Cur-
try of tropical forest trees. Biogeosciences 8:1081– rent Climate Change Reports 4:115–127.
1106. House, J. I., S. Archer, D. D. Breshears, R. J. Scholes,
Fisher, R. A., et al. 2018. Vegetation demographics in and NCEAS Tree–Grass Interactions Participants.
Earth System Models: a review of progress and pri- 2003. Conundrums in mixed woody–herbaceous
orities. Global Change Biology 24:35–54. plant systems. Journal of Biogeography 30:1763–
Flores, B. M., M. Holmgren, C. Xu, E. H. van Nes, C. 1777.
C. Jakovac, R. C. Mesquita, and M. Scheffer. 2017. Hunter, M. O., M. Keller, D. Morton, B. Cook, M. Lef-
Floodplains as an Achilles’ heel of Amazonian sky, M. Ducey, S. Saleska, R. C. de Oliveira Jr, and
 v www.esajournals.org 21 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
J. Schietti. 2015. Structural dynamics of tropical rough wall turbulent boundary layers. Journal of
moist forest gaps. PLOS ONE 10:e0132144. Fluids Engineering 99:503–509.
IPCC. 2013. Climate change 2013: the physical science Lee, J. A. 1991a. Near-surface wind flow around desert
basis. In Working Group I Contribution to the Fifth shrubs. Physical Geography 12:140–146.
Assessment Report of the Intergovernmental Lee, J. A. 1991b. The role of desert shrub size and spac-
Panel on Climate Change. Cambridge University ing on wind profile parameters. Physical Geogra-
Press, Cambridge, UK. phy 12:72–89.
IPCC. 2014. Climate change 2014: impacts, adaptation, Lehmann, C. E., S. A. Archibald, W. A. Hoffmann, and
and vulnerability. Part A: global and sectoral W. J. Bond. 2011. Deciphering the distribution of
aspects. In Contribution of Working Group II to the the savanna biome. New Phytologist 191:197–209.
Fifth Assessment Report of the Intergovernmental Lewis, S. L., P. M. Brando, O. L. Phillips, G. M. van der
Panel on Climate Change. Cambridge University Heijden, and D. Nepstad. 2011. The 2010 amazon
Press, Cambridge, UK. drought. Science 331:554.
Jin, Y., J. T. Randerson, M. L. Goulden, and S. J. Goetz. Lloyd, J., M. I. Bird, L. Vellen, A. C. Miranda, E. M.
2012. Post-fire changes in net shortwave radiation Veenendaal, G. Djagbletey, H. S. Miranda, G. Cook,
along a latitudinal gradient in boreal North Amer- and G. D. Farquhar. 2008. Contributions of woody
ica. Geophysical Research Letters 39. https://doi. and herbaceous vegetation to tropical savanna
org/10.1029/2012GL051790 ecosystem productivity: a quasi-global estimate.
Kalamandeen, M., E. Gloor, E. Mitchard, D. Quincey, Tree Physiology 28:451–468.
G. Ziv, D. Spracklen, B. Spracklen, M. Adami, L. E. Loarie, S. R., D. B. Lobell, G. P. Asner, and C. B. Field.
Arag~ao, and D. Galbraith. 2018. Pervasive rise of 2011. Land-cover and surface water change drive
small-scale deforestation in Amazonia. Scientific large albedo increases in South America. Earth
Reports 8:1600. Interactions 15:1–16.
Kloster, S., and G. Lasslop. 2017. Historical and future Magnusson, W. E., A. P. Lima, A. L. K. M. Albernaz, T.
fire occurrence (1850 to 2100) simulated in CMIP5 M. Sanaiotti, and J. L. Guillaumet. 2008. Com-
Earth System Models. Global and Planetary posi
c~ao flor
ıstica e cobertura vegetal das savanas
Change 150:58–69. na regi~ao de Alter do Cha~o, Santar
em - PA. Revista
Knox, R., G. Bisht, J. Wang, and R. Bras. 2011. Precipi- Brasileira De Botânica 31:165–177.
tation variability over the forest-to-nonforest tran- Malhi, Y., L. E. Arag~ao, D. Galbraith, C. Huntingford,
sition in southwestern Amazonia. Journal of R. Fisher, P. Zelazowski, S. Sitch, C. McSweeney,
Climate 24:2368–2377. and P. Meir. 2009. Exploring the likelihood and
Lagu€e, M. M., G. B. Bonan, and A. L. Swann. 2019. mechanism of a climate-change-induced dieback of
Separating the impact of individual land surface the Amazon rainforest. Proceedings of the National
properties on the terrestrial surface energy budget Academy of Sciences USA 106:20610–20615.
in both the coupled and uncoupled land–atmo- Malhi, Y., J. T. Roberts, R. A. Betts, T. J. Killeen, W. Li,
sphere system. Journal of Climate 32:5725–5744. and C. A. Nobre. 2008. Climate change, deforesta-
Laurance, W. F., T. E. Lovejoy, H. L. Vasconcelos, E. M. tion, and the fate of the Amazon. Science 319:L169–
Bruna, R. K. Didham, P. C. Stouffer, C. Gascon, R. L172.
O. Bierregaard, S. G. Laurance, and E. Sampaio. Marselis, S. M., H. Tang, J. D. Armston, K. Calders, N.
2002. Ecosystem decay of Amazonian forest frag- Labriere, and R. Dubayah. 2018. Distinguishing
ments: a 22-year investigation. Conservation Biol- vegetation types with airborne waveform lidar
ogy 16:605–618. data in a tropical forest-savanna mosaic: a case
Laurance, W. F., D. P
erez-Salicrup, P. Delamônica, P. study in Lop
e National Park, Gabon. Remote Sens-
M. Fearnside, S. D’Angelo, A. Jerozolinski, L. Pohl, ing of Environment 216:626–634.
and T. E. Lovejoy. 2001. Rain forest fragmentation Martens, S. N., D. D. Breshears, and F. J. Barnes. 2001.
and the structure of Amazonian liana communi- Development of species dominance along an eleva-
ties. Ecology 82:105–116. tional gradient: population dynamics of Pinus edu-
Lawrence, D., and K. Vandecar. 2015. Effects of tropi- lis and Juniperus monosperma. International Journal
cal deforestation on climate and agriculture. Nat- of Plant Sciences 162:777–783.
ure Climate Change 5:27. McDowell, N. G., et al. 2020. Pervasive shifts in forest
Lean, J., and D. A. Warrilow. 1989. Simulation of the dynamics in a changing world. Science. 368:
regional climatic impact of Amazon deforestation. eaaz9463.
Nature 342:411. McDowell, N., et al. 2018. Drivers and mechanisms of
Lee, B. E., and B. F. Soliman. 1977. An investigation of tree mortality in moist tropical forests. New Phy-
the forces on three dimensional bluff bodies in tologist 219:851–869.
 v www.esajournals.org 22 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Medvigy, D., R. L. Walko, M. J. Otte, and R. Avissar. Proceedings of the National Academy of Sciences
2013. Simulated changes in northwest US climate USA 117:7863–7870.
in response to Amazon deforestation. Journal of Parker, G. G., D. J. Harding, and M. L. Berger. 2004. A
Climate 26:9115–9136. portable LIDAR system for rapid determination of
Mesquita, R. C., K. Ickes, G. Ganade, and G. B. Wil- forest canopy structure. Journal of Applied Ecol-
liamson. 2001. Alternative successional pathways ogy 41:755–767.
in the Amazon Basin. Journal of Ecology 89:528– Pennington, T. R., D. E. Prado, and C. A. Pendry. 2000.
537. Neotropical seasonally dry forests and Quaternary
Miller, S. D., M. L. Goulden, L. R. Hutyra, M. Keller, S. vegetation changes. Journal of Biogeography
R. Saleska, S. C. Wofsy, A. M. S. Figueira, H. R. Da 27:261–273.
Rocha, and P. B. De Camargo. 2011. Reduced Phillips, O. L., et al. 2009. Drought sensitivity of the
impact logging minimally alters tropical rainforest Amazon rainforest. Science 323:1344–1347.
carbon and energy exchange. Proceedings of the Pires, G. F., and M. H. Costa. 2013. Deforestation
National Academy of Sciences USA 108:19431– causes different subregional effects on the Amazon
19435. bioclimatic equilibrium. Geophysical Research Let-
Miranda, I. S. 1993. Estrutura do estrato arbo
reo do ters 40:3618–3623.
cerrado amazônico em Alter do Ch~ao, Para
, Brasil. Rappaport, D. I., D. C. Morton, M. Longo, M. Keller,
Revista Brasileira De Botânica 16:143–150. R. Dubayah, and M. N. dos-Santos. 2018. Quantify-
Molina, R. D., J. F. Salazar, J. A. Mart
ınez, J. C. Ville- ing long-term changes in carbon stocks and forest
gas, and P. A. Arias. 2019. Forest-induced exponen- structure from Amazon forest degradation. Envi-
tial growth of precipitation along climatological ronmental Research Letters 13:065013.
wind streamlines over the Amazon. Journal of Ratajczak, Z., P. D’Odorico, J. B. Nippert, S. L. Collins,
Geophysical Research: Atmospheres 124:2589– N. A. Brunsell, and S. Ravi. 2017. Changes in spa-
2599. tial variance during a grassland to shrubland state
Montgomery, R. A., and R. L. Chazdon. 2001. Forest transition. Journal of Ecology 105:750–760.
structure, canopy architecture, and light transmit- Ratnam, J., W. J. Bond, R. J. Fensham, W. A. Hoffmann,
tance in tropical wet forests. Ecology 82:2707–2718. S. Archibald, C. E. Lehmann, M. T. Anderson, S. I.
Morecroft, M. D., M. E. Taylor, and H. R. Oliver. 1998. Higgins, and M. Sankaran. 2011. When is a ‘forest’
Air and soil microclimates of deciduous woodland a savanna, and why does it matter? Global Ecology
compared to an open site. Agricultural and Forest and Biogeography 20:653–660.
Meteorology 90:141–156. Ray, D., D. Nepstad, and P. Moutinho. 2005. Microme-
Nascimento, H. E., and W. F. Laurance. 2004. Biomass teorological and canopy controls of fire susceptibil-
dynamics in Amazonian forest fragments. Ecologi- ity in a forested Amazon landscape. Ecological
cal Applications 14:127–138. Applications 15:1664–1678.
Nepstad, D. C., C. M. Stickler, B. S. Filho, and F. Merry. Restrepo-Coupe, N., et al. 2013. What drives the sea-
2008. Interactions among Amazon land use, forests sonality of photosynthesis across the Amazon
and climate: prospects for a near-term forest tip- basin? A cross-site analysis of eddy flux tower
ping point. Philosophical Transactions of the Royal measurements from the Brazil flux network. Agri-
Society B: Biological Sciences 363:1737–1746. cultural and Forest Meteorology 182:128–144.
Nepstad, D. C., I. M. Tohver, D. Ray, P. Moutinho, and Royer, P. D., D. D. Breshears, C. B. Zou, N. S. Cobb,
G. Cardinot. 2007. Mortality of large trees and lia- and S. A. Kurc. 2010. Ecohydrological energy
nas following experimental drought in an Amazon inputs in semiarid coniferous gradients: responses
forest. Ecology 88:2259–2269. to management-and drought-induced tree reduc-
Norden, N., R. C. Mesquita, T. V. Bentos, R. L. Chaz- tions. Forest Ecology and Management 260:1646–
don, and G. B. Williamson. 2011. Contrasting com- 1655.
munity compensatory trends in alternative Royer, P. D., D. D. Breshears, C. B. Zou, J. C. Villegas,
successional pathways in central Amazonia. Oikos N. S. Cobb, and S. A. Kurc. 2012. Density-depen-
120:143–151. dent ecohydrological effects of pin~on–juniper
Oliveras, I., and Y. Malhi. 2016. Many shades of green: woody canopy cover on soil microclimate and
the dynamic tropical forest–savannah transition potential soil evaporation. Rangeland Ecology &
zones. Philosophical Transactions of the Royal Management 65:11–20.
Society B: Biological Sciences 371:20150308. Ru€ger, N., R. Condit, D. H. Dent, S. J. DeWalt, S. P.
Ordway, E. M., and G. P. Asner. 2020. Carbon declines Hubbell, J. W. Lichstein, O. R. Lopez, C. Wirth, and
along tropical forest edges correspond to heteroge- C. E. Farrior. 2020. Demographic trade-offs predict
neous effects on canopy structure and function. tropical forest dynamics. Science 368:165–168.
 v www.esajournals.org 23 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
Salazar, A., et al. 2018. The ecology of peace: preparing light environment in an Amazon forest. New Phy-
Colombia for new political and planetary climates. tologist 222:1284–1297.
Frontiers in Ecology and the Environment 16:525– Soares-Filho, B., et al. 2012. Forest fragmentation, cli-
531. mate change and understory fire regimes on the
Saleska, S. R., et al. 2003. Carbon in Amazon forests: Amazonian landscapes of the Xingu headwaters.
unexpected seasonal fluxes and disturbance-in- Landscape Ecology 27:585–598.
duced losses. Science 302:1554–1557. Sperry, J. S., and U. G. Hacke. 2002. Desert shrub water
Sankaran, M., et al. 2005. Determinants of woody relations with respect to soil characteristics and
cover in African savannas. Nature 438:846. plant functional type. Functional Ecology 16:367–
Sankaran, M., J. Ratnam, and N. P. Hanan. 2004. Tree– 378.
grass coexistence in savannas revisited–insights Spracklen, D. V., and L. Garcia-Carreras. 2015. The
from an examination of assumptions and mechanisms impact of Amazonian deforestation on Amazon
invoked in existingmodels. EcologyLetters 7:480–490. basin rainfall. Geophysical Research Letters
Sankey, J. B., D. J. Law, D. D. Breshears, S. M. Munson, 42:9546–9552.
and R. H. Webb. 2013. Employing lidar to detail Stark, S. C., et al. 2012. Amazon forest carbon
vegetation canopy architecture for prediction of dynamics predicted by profiles of canopy leaf area
aeolian transport. Geophysical Research Letters and light environment. Ecology Letters 15:1406–
40:1724–1728. 1414.
Schietti, J., et al. 2016. Forest structure along a 600 km Stark, S. C., et al. 2016. Toward accounting for ecocli-
transect of natural disturbances and seasonality mate teleconnections: intra-and inter-continental
gradients in central-southern Amazonia. Journal of consequences of altered energy balance after vege-
Ecology 104:1335–1346. tation change. Landscape Ecology 31:181–194.
Scholes, R. J., and S. R. Archer. 1997. Tree-grass interac- Stark, S. C., B. J. Enquist, S. R. Saleska, V. Leitold, J.
tions in savannas. Annual Review of Ecology and Schietti, M. Longo, L. F. Alves, P. B. Camargo, and
Systematics 28:517–544. R. C. Oliveira. 2015. Linking canopy leaf area and
Shao, G., S. C. Stark, D. R. de Almeida, and M. N. light environments with tree size distributions to
Smith. 2019. Towards high throughput assessment explain Amazon forest demography. Ecology Let-
of canopy dynamics: the estimation of leaf area ters 18:636–645.
structure in Amazonian forests with multitemporal Staver, A. C., S. Archibald, and S. A. Levin. 2011. The
multi-sensor airborne lidar. Remote Sensing of global extent and determinants of savanna and for-
Environment 221:1–13. est as alternative biome states. Science 334:230–232.
Shukla, J., C. Nobre, and P. Sellers. 1990. Amazon Swann, A. L., I. Y. Fung, and J. C. Chiang. 2012. Mid-
deforestation and climate change. Science latitude afforestation shifts general circulation and
247:1322–1325. tropical precipitation. Proceedings of the National
Shuttleworth, W. J. 2012. Terrestrial hydrometeorology. Academy of Sciences USA 109:712–716.
Wiley, Chichester, UK. Tan, Z. H., et al. 2019. Surface conductance for evapo-
Silva, C. V., et al. 2018. Drought-induced Amazonian transpiration of tropical forests: calculations, varia-
wildfires instigate a decadal-scale disruption of for- tions, and controls. Agricultural and Forest
est carbon dynamics. Philosophical Transactions of Meteorology 275:317–328.
the Royal Society B: Biological Sciences Tang, H., J. Armston, S. Hancock, S. Marselis, S. Goetz,
373:20180043. and R. Dubayah. 2019. Characterizing global forest
Silv
erio, D. V., P. M. Brando, J. K. Balch, F. E. Putz, D. canopy cover distribution using spaceborne lidar.
C. Nepstad, C. Oliveira-Santos, and M. M. Busta- Remote Sensing of Environment 231:111262.
mante. 2013. Testing the Amazon savannization Tyukavina, A., M. C. Hansen, P. V. Potapov, S. V. Steh-
hypothesis: fire effects on invasion of a neotropical man, K. Smith-Rodriguez, C. Okpa, and R. Aguilar.
forest by native cerrado and exotic pasture grasses. 2017. Types and rates of forest disturbance in
Philosophical Transactions of the Royal Society B: Brazilian Legal Amazon, 2000–2013. Science
Biological Sciences 368:20120427. Advances 3:e1601047.
Silv
erio, D. V., P. M. Brando, M. M. Bustamante, F. E. Villegas, J. C., et al. 2015. Sensitivity of regional evapo-
Putz, D. M. Marra, S. R. Levick, and S. E. Trum- transpiration partitioning to variation in woody
bore. 2019. Fire, fragmentation, and windstorms: a plant cover: insights from experimental dryland
recipe for tropical forest degradation. Journal of tree mosaics. Global Ecology and Biogeography
Ecology 107:656–667. 24:1040–1048.
Smith, M. N., et al. 2019. Seasonal and drought-related Villegas, J. C., et al. 2017. Prototype campaign assess-
changes in leaf area profiles depend on height and ment of disturbance-induced tree loss effects on
 v www.esajournals.org 24 September 2020 v Volume 11(9) v Article e03231
INNOVATIVE VIEWPOINTS STARK ETAL.
surface properties for atmospheric modeling. Eco- Wang, L., K. K. Caylor, J. C. Villegas, G. A. Barron-Gaf-
sphere 8:e01698. ford, D. D. Breshears, and T. E. Huxman. 2010. Par-
Villegas, J. C., D. D. Breshears, C. B. Zou, and D. J. titioning evapotranspiration across gradients of
Law. 2010a. Ecohydrological controls of soil evapo- woody plant cover: assessment of a stable isotope
ration in deciduous drylands: how the hierarchical technique. Geophysical Research Letters 37.
effects of litter, patch and vegetation mosaic cover https://doi.org/10.1029/2010GL043228
interact with phenology and season. Journal of Wehr, R., R. Commane, J. W. Munger, J. B. McMa-
Arid Environments 74:595–602. nus, D. D. Nelson, M. S. Zahniser, S. R. Saleska,
Villegas, J. C., D. D. Breshears, C. B. Zou, and P. D. and S. C. Wofsy. 2017. Dynamics of canopy
Royer. 2010b. Seasonally pulsed heterogeneity in stomatal conductance, transpiration, and evapora-
microclimate: phenology and cover effects along tion in a temperate deciduous forest, validated by
deciduous grassland–forest continuum. Vadose carbonyl sulfide uptake. Biogeosciences 14:389–
Zone Journal 9:537–547. 401.
Villegas, J. C., J. E. Espeleta, C. T. Morrison, D. D. Bres- Wolfe, S. A., and W. G. Nickling. 1993. The protective
hears, and T. E. Huxman. 2014. Factoring in canopy role of sparse vegetation in wind erosion. Progress
cover heterogeneity on evapotranspiration parti- in Physical Geography 17:50–68.
tioning: beyond big-leaf surface homogeneity Wu, J., et al. 2018. Biological processes dominate sea-
assumptions. Journal of Soil and Water Conserva- sonality of remotely sensed canopy greenness in an
tion 69:78A–83A. Amazon evergreen forest. New Phytologist
von Randow, C., et al. 2004. Comparative measure- 217:1507–1520.
ments and seasonal variations in energy and car- Zellweger, F., P. De Frenne, J. Lenoir, D. Rocchini, and
bon exchange over forest and pasture in South D. Coomes. 2019. Advances in microclimate ecol-
West Amazonia. Theoretical and Applied Climatol- ogy arising from remote sensing. Trends in Ecology
ogy 78:5–26. & Evolution 34:327–341.
SUPPORTING INFORMATION
Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.
3231/full
 v www.esajournals.org 25 September 2020 v Volume 11(9) v Article e03231