Glob Change Biol. 2019;00:1–15.	 wileyonlinelibrary.com/journal/gcb	 	 | 	1© 2019 John Wiley & Sons Ltd
1  | INTRODUC TION
Understanding tropical forest responses to atmospheric and cli‐
mate change is critical to modeling the future global carbon cycle. 
Repeated censuses of forest inventory plots have found increasing 
stocks of tree carbon in apparently undisturbed tropical forests in 
Amazonia (Baker et al., 2004; Phillips et al., 1998), Africa (Lewis 
et al., 2013), and Borneo (Qie et al., 2017). Globally, analysis of 
long‐term forest inventories suggests that old‐growth trop‐
ical forests are currently carbon sinks, sequestering an average 
 
Received:	20	April	2019  |  Accepted:	27	August	2019
DOI: 10.1111/gcb.14833  
P R I M A R Y  R E S E A R C H  A R T I C L E
Testing for changes in biomass dynamics in large‐scale forest 
datasets
Ervan Rutishauser1  |   Stuart J. Wright1 |   Richard Condit2 |   Stephen P. Hubbell3 |   
Stuart J. Davies4,5 |   Helene C. Muller‐Landau1
1Smithsonian Tropical Research Institute, 
Ancon, Panama
2Morton Arboretum, Lisle, IL, USA
3Department of Ecology and Evolutionary 
Biology, University of California, Los 
Angeles, CA, USA
4Center for Tropical Forest Science‐Forest 
Global Earth Observatory, Smithsonian 
Tropical Research Institute, Panama City, 
Panama
5Department of Botany, National Museum of 
Natural History, Washington, DC, USA
Correspondence
Ervan Rutishauser, Smithsonian Tropical 
Research Institute, Box 0843‐03092 Balboa, 
Ancon, Panama.
Email: er.rutishauser@gmail.com
Funding information
US Department of Energy; National Science 
Foundation; Smithsonian Tropical Research 
Institute; MacArthur Foundation
Abstract
Tropical forest responses to climate and atmospheric change are critical to the future 
of the global carbon budget. Recent studies have reported increases in estimated 
above‐ground biomass (EAGB) stocks, productivity, and mortality in old‐growth 
tropical forests. These increases could reflect a shift in forest functioning due to 
global change and/or long‐lasting recovery from past disturbance. We introduce a 
novel approach to disentangle the relative contributions of these mechanisms by de‐
composing changes in whole‐plot biomass fluxes into contributions from changes in 
the distribution of gap‐successional stages and changes in fluxes for a given stage. 
Using 30 years of forest dynamic data at Barro Colorado Island, Panama, we inves‐
tigated temporal variation in EAGB fluxes as a function of initial EAGB (EAGBi) in 
10 × 10 m quadrats. Productivity and mortality fluxes both increased strongly with 
initial quadrat EAGB. The distribution of EAGB (and thus EAGBi) across quadrats 
hardly varied over 30 years (and seven censuses). EAGB fluxes as a function of EAGBi 
varied largely and significantly among census intervals, with notably higher produc‐
tivity in 1985–1990 associated with recovery from the 1982–1983 El Niño event. 
Variation in whole‐plot fluxes among census intervals was explained overwhelmingly 
by variation in fluxes as a function of EAGBi, with essentially no contribution from 
changes in EAGBi distributions. The high observed temporal variation in productivity 
and mortality suggests that this forest is very sensitive to climate variability. There 
was no consistent long‐term trend in productivity, mortality, or biomass in this forest 
over 30 years, although the temporal variability in productivity and mortality was so 
strong that it could well mask a substantial trend. Accurate prediction of future tropi‐
cal forest carbon budgets will require accounting for disturbance‐recovery dynamics 
and understanding temporal variability in productivity and mortality.
K E Y W O R D S
biomass dynamic, carbon fluxes, long‐term change, tropical forests
2  |     RUTISHAUSER ET Al.
of 0.34 Mg C ha−1 year−1 (confidence intervals [CIs] = 0.17–0.47; 
Muller‐Landau, Detto, Chisholm, Hubbell, & Condit, 2014), albeit 
with a reduction in sink strength over time (Brienen et al., 2015; 
Qie et al., 2017). The causes of this increase and its consistency 
in space and time remain uncertain (Lewis, Lloyd, Sitch, Mitchard, 
& Laurance, 2009; Wright, 2010, 2013). Current knowledge of 
changes in old‐growth tropical forest carbon stocks is hampered 
by limited spatial and temporal coverage of monitoring plots, sen‐
sitivity to measurement errors and data correction procedures, 
and the confounding influences of disturbance‐recovery dynamics 
(Clark, Clark, et al., 2017; Muller‐Landau et al., 2014).
There are multiple pathways through which anthropogenic 
global change could increase or decrease carbon stocks and fluxes in 
old‐growth tropical forests. Increasing atmospheric carbon dioxide 
could enhance tropical tree growth and forest productivity through 
increased photosynthesis and water use efficiency (Lloyd & Farquhar, 
2008), and thereby increase forest carbon stocks (Figure 1a–c). 
Alternatively, rising temperatures and increasing drought frequency 
and/or intensity could increase tree mortality and thereby decrease 
forest carbon stocks (McDowell et al., 2018). The “Bigger and Faster” 
hypothesis proposes that growth, recruitment, and mortality are 
all increasing (Clark, Clark, et al., 2017); in this case, the net effect 
on biomass depends on the relative magnitude of the increases in 
productivity and mortality (Figure 1d–f). Some but not all studies 
have found evidence of increasing productivity (Brienen et al., 2015; 
Feeley, Wright, Supardi, Kassim, & Davies, 2007; Körner, 2015) and 
increasing mortality (Brando et al., 2014; Brienen et al., 2015; Liu 
et al., 2017; McDowell et al., 2018; Phillips et al., 2010).
It is also possible that the observed increase in tropical forest bio‐
mass is due not to global change influences but to long‐lasting recovery 
from past disturbances in some (but not all) tropical forests (Figure 1g–i; 
Wright, 2010). There is growing evidence that past human‐induced 
or climatic disturbances continue to profoundly shape current forest 
composition (Brncic, Willis, Harris, & Washington, 2006; Levis et al., 
2017; Magnabosco Marra et al., 2018), functioning (Doughty et al., 
2015; McMichael, Matthews‐Bird, Farfan‐Rios, & Feeley, 2017; Oslisly 
et al., 2013), and extent (Mayle, Burbridge, & Killeen, 2000) in tropical 
regions. Forest succession after disturbance would generate a pattern 
of increasing total biomass mortality and biomass productivity over 
time, similar to that expected under the “Bigger and Faster” hypothesis 
(Clark, Clark, et al., 2017), and constitutes a null hypothesis or non‐
global change hypothesis. Forest recovery post‐disturbance poses 
a particular challenge, as it may act at various spatial and temporal 
scales, and in only a subset of tropical forests.
We propose to separate out the contributions of disturbance‐re‐
covery dynamics to temporal variation in forest carbon stocks by 
decomposing changes in whole‐plot biomass fluxes into contribu‐
tions from changes in the distribution of gap‐successional stages and 
changes in fluxes for a given stage (Figure 2). An old‐growth tropical 
forest is a shifting mosaic of patches at different stages of recovery 
(Watt, 1947; Whitmore, 1989), of which most are increasing in biomass 
at any given time (Körner, 2003a). Dividing a plot into small quad‐
rats provides a simple operational way to characterize this mosaic of 
stages, and quadrat initial EAGB (EAGBi) provides a useful proxy for 
age/time since disturbance (Bormann & Likens, 1979; Denslow, Ellison, 
& Sanford, 1998). Plot‐level EAGB fluxes (Figure 2c) then reflect the 
integration of functions for fluxes as a function of EAGBi (Figure 2a) 
with the distribution of EAGBi (Figure 2b). Both quadrat productiv‐
ity and mortality fluxes (Mg/ha) are expected to increase with EAGBi 
(Figure 2a), and consistent global change influences should be evident 
in directional changes in these functions over time—that is, in increases 
or decreases in fluxes when controlling for EAGBi. Climate variation 
could also contribute to temporal variation in productivity and mor‐
tality drivers, and thus in fluxes for a given EAGBi (Dong et al., 2012). 
In contrast, the distribution of quadrat EAGBi depends largely on the 
disturbance history at a site (Espírito‐Santo et al., 2014; Fisher, Hurtt, 
F I G U R E  1   Expected patterns of quadrat biomass fluxes in 
relation to initial estimated above‐ground biomass (EAGBi) for three 
possible scenarios (columns) consistent with increasing biomass 
stocks in tropical forests. In every panel, the solid line shows the 
pattern expected for an earlier census interval, and the dashed 
line for a later census interval. Increasing productivity due to 
“fertilization” (first column), means higher productivity for a given 
EAGBi in the second census interval (a, dashed line) than in the 
first one (a, solid line), no systematic difference in mortality after 
controlling for EAGBi (b), and thus higher net change in EAGB per 
EAGBi in the second census interval (c). Increasing productivity 
and mortality, as expected under “speeding up” (second column), 
means higher productivity and higher mortality in the second 
census interval (d, e) and diverse possible patterns for net change 
(f). Finally, recovery from disturbance (final column) is expected to 
show no systematic change in fluxes when controlling for EAGBi 
(g–i). Note that local quadrat productivity increases with initial 
quadrat EAGB in all scenarios and census intervals (first row, blue), 
as does the local mortality flux (second row, yellow). Quadrat‐level 
biomass net change (third row, black) is the difference between 
productivity and mortality fluxes, and is positive for low EAGBi, and 
becomes negative for high EAGBi
     |  3RUTISHAUSER ET Al.
Thomas, & Chambers, 2008), and can fluctuate from a dominance 
of early successional (i.e., low EAGBi) stages soon after disturbance 
(Figure 2b, time 1) to higher frequency of late successional quadrats 
longer after disturbance (Figure 2b, time 2).
Using a large‐scale, long‐term forest plot dataset for Barro 
Colorado Island (BCI), Panama, we seek to disentangle the influences 
of disturbance recovery from potential global change by quantifying 
how local biomass dynamics vary with initial biomass, and evaluating 
whether these relationships change over time. Specifically, we tested 
for temporal variation in quadrat productivity, loss, and net change in 
biomass when controlling for initial biomass. To assess the degree to 
which similar biomass patches were consistently associated with sim‐
ilar forest structure across census intervals, we also examined how 
quadrat initial biomass was related to tree density (per area) and mean 
tree size at each census and compared these relationships over time.
2  | METHODS
2.1 | Study site
Barro Colorado Island is located in central Panama (9°08' N, 
79°510' W). Temperature averages 27°C and annual rainfall av‐
erages 2,657 mm (for 1926–2017; Paton, 2018), with a dry sea‐
son between January and April. The vegetation is moist tropical 
forest. The 50 ha forest dynamics plot was established in 1982 
(Condit, 1998; Hubbell et al., 1999), and recensused in 1985 and 
every 5 years subsequently (Figure S1). All free‐standing woody 
stems	with	a	diameter	at	breast	height	(DBH)	≥1	cm	were	mapped,	
tagged, and identified to species, and were measured in diameter 
at every census in which they remained alive. Diameter measure‐
ments were taken at 1.3 m height or above any buttress or major 
stem deformity. We omitted the initial census of 1982–1983 from 
analyses of biomass and biomass dynamics due to major differ‐
ences in the field measurement methods for large trees (Condit, 
1998), leaving seven censuses and six 5 year intervals between 
1985 and 2015. We omitted 1.2 ha of swamp and 1.9 ha of young 
forest from the analysis due to differences in species composition, 
structure, and dynamics (Condit, Hubbell, & Foster, 1996; Harms, 
Condit, Hubbell, & Foster, 2001), leaving 46.9 ha for analysis. (We 
note, however, that our results are robust to the inclusion of the 
young forest, as we would expect; see Supporting information C1). 
Supplemental materials provide additional details on the census 
methods (Supporting information A).
2.2 | Individual tree biomass calculation
For all stems of dicot (non‐palm) species, we estimated above‐
ground biomass (EAGB) from individual DBH and species‐specific 
F I G U R E  2   Whole plot estimated above‐ground biomass (EAGB) fluxes (right column) in a given time period depends on the combination 
of how small‐scale fluxes vary with initial EAGB (EAGBi) (left column) and how EAGBi varies spatially across the plot and over time (middle 
column). Here, we illustrate this for productivity; the same integration holds for whole‐plot mortality. To distinguish the contribution of 
flux–EAGBi relationships from the contribution of EAGBi distributions to plot‐level biomass dynamics, we estimated what whole‐plot fluxes 
would be in each census interval if either the flux–EAGBi relationship or the EAGBi distribution were constant across census intervals. To 
calculate a single constant flux–EAGBi relationship (d) and a single constant EAGBi distribution (h), we averaged across all census intervals 
weighted equally. We then integrated the empirical flux–EAGBi relationship over the EAGBi distribution for each census interval under three 
scenarios: (i) interval‐specific flux–EAGBi relationships and interval‐specific EAGBi distributions (a–c); (ii) the average flux–EAGBi relationship 
and interval‐specific EAGBi distributions (d–f); (iii) interval‐specific flux–EAGBi relationships and the average EAGBi distribution (g–i)
4  |     RUTISHAUSER ET Al.
wood density (WD) using a generic tropical forest biomass equation 
(Chave et al., 2014). For stems measured at heights other than 1.3 m, 
we first calculated taper‐corrected equivalent diameters at 1.3 m 
height using a generic taper correction equation for BCI:
where D is diameter measured at height of measurement (HOM; in 
m; equation 2, Cushman, Muller‐Landau, Condit, & Hubbell, 2014). 
We applied this taper correction to avoid time‐dependent biases 
in plot‐level biomass stocks and fluxes associated with changes 
over time in the proportion of trees measured above 1.3 m and 
average measurement heights (Cushman et al., 2014). We also es‐
timated missing diameter measurements (1,192 of the 2,607,014 
records) by simple linear interpolation between consecutive mea‐
surements. We then EAGB (in Mg dry biomass) of each stem in 
each census from its DBH, using the updated version of equation 
7 of Chave et al. (2014), incorporating species‐level WD (in g/cm3) 
collected near the study site (Wright et al., 2010 and unpublished) 
and a climatic index that represents environmental influences on 
tree height allometries (E = 0.05645985 for BCI):
EAGB=exp
(
−2.024−0.896×E+0.920× log (WD)
+2.795× log (DBH)−0.0461× log (DBH)2
)
 
(see Supporting information S2 for details of wood density 
assignment).
For palms, we estimated biomass using a palm‐specific allome‐
tric equation based on DBH. For palm species that do not grow in 
DBH (all local palm species except Socratea exorrhiza), we first as‐
signed species‐specific median DBH to each palm stem in each cen‐
sus (Table S1). We EAGB (in Mg dry biomass) of each palm stem as
based on the family‐wise specific allometric equation of Goodman 
et al. (2013), modified to include the log‐transformation correction 
factor (Baskerville, 1972). We recognize that this procedure results 
in systematic errors in biomass fluxes for palms that do not grow in 
diameter, because it fails to consider their height growth, and that 
better estimates would be possible if height data were available. 
Overall, palms other than Socratea account for 0.9% of the woody 
biomass and 1.8% of the estimated woody productivity on BCI.
For multi‐stemmed trees, individual tree biomass was calculated 
as the sum of biomass of all live stems. Large (DBH > 50 cm) individ‐
uals of strangler fig species (Ficus costaricana, Ficus obtusifolia, Ficus 
popenoei, and Ficus trigonata), for which diameter measurements are 
unreliable measures of size, were excluded from the analysis by ex‐
cluding associated quadrats (Table S2).
2.3 | Quadrat‐level biomass stocks and fluxes
Biomass stocks and fluxes were analyzed at the scale of 10 m quad‐
rats and for the plot as a whole. Analysis at the quadrat level aimed 
to quantify how local biomass fluxes vary with local initial biomass 
for each census interval, and to compare these relationships across 
census intervals. Initial biomass was calculated as the sum of bio‐
mass of all trees alive at the initial census. We calculated biomass 
fluxes per time in productivity, mortality, and net change using 
simple arithmetic estimators. That is, we calculated fluxes for each 
quadrat by summing over relevant trees (i) and dividing by the time 
interval measured separately for each tree:
Productivity was calculated as the sum of the EAGB changes of sur‐
viving trees and the EAGB of recruits (including resprouts). Because 
the census dataset includes all stems >1 cm in diameter, biomass in‐
crement accounts for the vast majority of the woody productivity 
(97.5%; Figure S2). Biomass loss was calculated as the sum of EAGBi 
of all trees that died by the next census, as well as of large individual 
stems that died on multi‐stemmed individuals (see Supporting infor‐
mation A for details on the treatment of multi‐stemmed trees). For 
recruits and dead stems, the time interval used was the mean time 
between measurements for that quadrat and census interval. These 
arithmetic estimators systematically underestimate fluxes because 
they miss contributions from trees that recruited and died during 
the census interval, with larger biases for longer census intervals 
(Kohyama, Kohyama, & Sheil, 2019). However, all our census inter‐
vals were very similar in length (Figure S1), and thus, these biases 
do not confound our analysis. Estimated biomass stocks and fluxes 
are expressed on a quadrat basis (100 m−2), and can be converted to 
units of Mg/ha by multiplying by 100.
We quantified the relationship of quadrat‐level biomass fluxes 
to initial biomass for each census interval using local regression and 
power function fits. We first illustrate the relationships for produc‐
tivity, mortality, and net change using local polynomial regression, 
specifically the locfit() function in R (Loader, 1999) with smoothing 
parameter (alpha) set to 0.7. To formally test for variation in EAGB 
productivity and loss among census intervals, each of these fluxes 
was modeled as power functions of EAGBi, with a power variance 
function structure (using package lme4; Bates, Mächler, Bolker, & 
Walker, 2015). We tested for differences among census intervals 
in power function “slopes” (exponents) and normalized intercepts, 
with these intercepts calculated as the predictions for the median 
EAGB value of 125.0 Mg/ha. CIs were computed by bootstrapping 
over quadrats (1,000 times). We evaluated the sensitivity of the 
results to different local regression models, and the consistency 
between the power function and local regression results.
For each census interval, our analyses of quadrat‐level fluxes 
excluded quadrats affected by any one or more of three prob‐
lems. First, analyses excluded quadrats if any stem had a change 
in HOM >10 cm during that census interval. Second, analyses ex‐
cluded quadrats that ever had a strangler fig larger than 50 cm DBH. 
Finally, analyses excluded quadrats where an individual stem had an 
unreasonably large change in DBH in that census interval, such that 
its absolute biomass change was >0.98 Mg/ha (this threshold was 
DBH=D×exp (0.0247× (HOM−1.3)),
AGB=0.0417565×DBH
2.7483
,
EAGB change=
∑ (EAGBfinal i−EAGBinitial i)
timei
.
     |  5RUTISHAUSER ET Al.
chosen because it is 20% of the mean productivity per hectare com‐
puted over all quadrats and census intervals that met the first two 
criteria). We excluded between 389 and 1,288 of the 4,669 possible 
quadrats (i.e., not in swamps or in young forest), depending on the 
census interval, mostly because of the first rule removing quadrats 
with a change in measurement height (Table S2). We did not remove 
or correct unlikely but less influential diameter changes, because in 
a dataset of this size, these random errors will tend to cancel out, 
whereas attempts to filter or correct such changes have strong po‐
tential to introduce systematic biases (because it is easier to detect 
some types of errors than others, Muller‐Landau et al., 2014). The 
results were robust to varying filtering methods (Figure S7).
We evaluated consistency in or differences among censuses in 
the probability distribution of quadrat‐level EAGB using empirical 
cumulative density functions. We specifically tested for differences 
among censuses in selected percentiles (5th, 10th, 25th, 50th, 75th, 
90th, and 95th), by evaluating whether CIs overlapped. We esti‐
mated CIs on the percentiles by bootstrapping over quadrats.
We evaluated the sensitivity of our results to the spatial grain of 
quadrats, applying the same methods for smaller (8.33 m) and larger 
(12.5, 15.15 m) quadrats (Figures S8–S10).
2.4 | Disentangling sources of temporal variation in 
whole‐plot fluxes
To distinguish the contribution of flux–EAGBi relationships from the 
contribution of EAGBi distributions to plot‐level biomass dynamics, 
we estimated what whole‐plot fluxes would be in each census inter‐
val if either the flux–EAGBi relationship or the EAGBi distribution was 
constant across census intervals. To calculate a single constant flux–
EAGBi relationship and a single constant EAGBi distribution, we aver‐
aged across all census intervals weighted equally. We then integrated 
the empirical flux–EAGBi relationship over the EAGBi distribution 
(Figure 2) for each census interval under three scenarios: (a) inter‐
val‐specific flux–EAGBi relationships and the average EAGBi distri‐
bution; (b) the average flux–EAGBi relationship and interval‐specific 
EAGBi distributions; and (c) interval‐specific flux–EAGBi relation‐
ships and interval‐specific EAGBi distributions. For each flux–EAGBi 
relationship, fluxes for the central 90% of the EAGBI distribution 
(from the 5th to 95th percentile) were estimated from EAGBi using 
fits of the locfit function to the full distribution, as above (α = 0.7). 
In the tails of the distribution, fluxes were estimated as equal to the 
means for the corresponding interval (i.e., the 0–5th percentile inter‐
val or the 95th–100th percentile interval), to avoid undue influence 
of a few points.
2.5 | Forest structure
For each census interval, forest structure was quantified by analyzing 
the	distributions	over	10	×	10	m	quadrats	of	the	numbers	of	trees	≥	1,	
10, and 60 cm, and of quadratic mean diameter (Dg=
�∑
DBH
2
∕n, 
in cm). We quantified how these structure measures varied with 
quadrat EAGB in each census interval by fitting local regressions (R 
function locfit with α = 0.7), and calculating values expected for se‐
lect quadrat EAGB (corresponding to the 10th, 25th, 50th, 75th, and 
90th percentiles computed for all censuses combined). We tested 
whether there were changes over time in the values expected for 
selected quadrat EAGB or in the plot‐level means, by comparing CIs 
obtained by bootstrapping over quadrats.
3  | RESULTS
3.1 | Quadrat‐level biomass fluxes
Mean quadrat EAGB productivity and EAGB mortality loss both in‐
creased with initial quadrat EAGB in each census interval, with uncer‐
tainty rising in parallel (Figures 3 and 4a,b; Figure S2). Both fluxes were 
well‐fit by power functions of EAGBi (Figure 4a,b; Figures S5 and S6). 
Mean EAGB net change was positive for low EAGBi and became 
negative in quadrats with EAGB> ~3 Mg/100 m2 (equivalently 
300 Mg/ha; Figure 4c).
There was substantial variation among census intervals in EAGB 
fluxes as a function of EAGBi. This is visually evident from examination 
of the differences in EAGB‐specific fluxes from the mean flux over cen‐
sus intervals (Figure 4d–f), and is quantified by differences in parameter 
values of the power function fits (Figure 5; power functions were good 
fits to the data, as is evident from their concordance with local regres‐
sions in Figures S5 and S6). Productivity fluxes in the 1985–1990 census 
interval averaged 10%–30% higher than the mean over census intervals 
for all but the highest values of EAGBi (Figure 4d, blue line). The next 
highest productivity fluxes were in the most recent census interval, and 
the lowest productivity fluxes were in 2000–2005 (Table S6). EAGB loss 
also varied among census intervals, with above‐average EAGB loss for 
lower EAGBi in some census intervals (2005–2010 and 2010–2015) and 
for higher EAGBi in other intervals (1995–2000). The power function 
fits reflected these patterns, with significant variation among census 
intervals in the normalized intercepts, the estimated fluxes at the mid‐
point of the EAGBi distribution (here 125 Mg/ha; Figure 5a,b; Table S6). 
For productivity, the first census interval had by far the largest inter‐
cept, with subsequent declines to 2000–2005, and then increases since 
2005 (Figure 5a). For mortality, the intercept tended to increase after 
the 1990–1995 census interval. The slopes (exponents) of the power 
functions showed little variation among census intervals (mostly over‐
lapping CIs, Figure 5c,d). CIs for EAGB loss and net flux were wide 
(Figures S3 and S4), limiting the power to test for significant differences 
among census intervals (Figure 4d–f; Figure S3 and S4). Results were 
qualitatively consistent for other quadrat sizes (Figures S8–S10).
3.2 | Whole plot biomass stocks and fluxes
The probability distribution of EAGB density over 10 × 10 m quad‐
rats varied only slightly among censuses, and showed no clear direc‐
tional change (Figure 6a). The 5th–50th percentiles of quadrat EAGB 
increased slightly from 1985 to 2000, and then decreased again to 
values similar to initial ones (Figure 6c). In contrast, the 90th percen‐
tile increased slightly over the study period (Figure 6c).
6  |     RUTISHAUSER ET Al.
F I G U R E  3   Quadrat‐level mean annual estimated above‐ground biomass (EAGB) productivity (Mg 100 m−2 year−1) as a function of 
initial quadrat biomass (EAGB in Mg/100 m2) for each 5 year census interval, for individual 10 × 10 m quadrats (points), together with 
local polynomial regression lines for individual census intervals (colored solid lines) and for all census intervals combined (black dashed 
lines). Confidence envelopes were computed from bootstrapping over quadrats. For each census interval, analyses excluded quadrats with 
problematic EAGB change measurements (see Section 2 for details); N is the number of quadrats included for each census interval. The 
range of data shown here is truncated, but all points were included in the analyses. Note that EAGB stocks and fluxes can be converted to 
units of Mg/ha by multiplying by 100
F I G U R E  4   Variation among census intervals in quadrat estimated above‐ground biomass (EAGB) productivity (a), loss (b), and net change 
(c) as a function of above‐ground biomass (EAGB) at the initial census for 10 × 10 m quadrats, together with the corresponding differences 
from the mean (per initial EAGB) across all census intervals (d–f). Patterns were quantified using local polynomial regression; regression lines 
are displayed as solid lines for initial EAGB values spanning the 2nd–98th percentiles of the respective census interval, and as dotted lines 
outside this range. Confidence envelopes were computed by bootstrapping over quadrats, and are shown for the 2nd–98th percentiles of 
initial EAGB using shading. Analyses excluded quadrats with problematic EAGB change measurements (see methods for details). Note that 
productivity (a) and loss (b) are shown on log scales, and their differences from the mean are expressed in percentages (d,e) whereas the net 
change (c) and its difference from the mean (f) are shown on linear scales. Note that EAGB stocks and fluxes can be converted to units of 
Mg/ha by multiplying by 100
     |  7RUTISHAUSER ET Al.
Temporal variation in whole‐plot EAGB fluxes as estimated 
by integrating temporal variation in flux–EAGBi relationships and 
EAGBi distributions was due almost entirely to temporal vari‐
ation in flux–EAGBi relationships for a given EAGBi, with almost 
no contribution of temporal variation in the EAGBi distribution 
(Figure 7; Table S4 scenario b). Temporal patterns in whole‐plot 
fluxes thus roughly mirrored those in fluxes as a function of EAGBi. 
Productivity flux was highest in 1985–1990, then dropped steeply, 
and has increased over the last three census intervals. Mortality 
loss flux dropped between the first and second census interval, 
and tended to increase since then, albeit CIs are wide. These pat‐
terns in estimated whole‐plot fluxes mirrored calculated fluxes 
from summing over quadrats, although the exact pattern in pro‐
ductivity is highly sensitive to data cleaning and correction proce‐
dures (Figures S7–S17).
3.3 | Changes in forest structure
The mean density of stems >1 cm dbh increased modestly after the 
1982–1983 drought, then thinned rapidly from 1990 to 2005, and 
was stable from 2005 to 2015 (Figure 8a, black points and lines). The 
same pattern was evident when controlling for focal quadrat biomass 
density	(Figure	8a,	colored	lines).	The	density	of	stems	≥10	cm	dbh	
in the plot as a whole increased slowly to 1995 and then decreased 
to 2010, with parallel patterns in all but the lowest biomass quadrats 
(Figure	8b).	The	density	of	 large	stems	 (≥60	cm	dbh)	 remained	es‐
sentially constant over time (Figure 8c; Table S5). Quadratic mean 
F I G U R E  5   Variation among census intervals in the intercepts (a, b) 
and slopes (c, d) of power function relationships of quadrat‐level 
productivity (a, c) and losses to mortality (b, d) to initial above‐
ground biomass. Intercepts are normalized, and represent the 
values at the overall midpoint estimated above‐ground biomass 
(EAGB) value (1.25 Mg/100 m2, i.e., an EAGB density of 125 Mg/ha).  
Confidence intervals were computed from bootstrapping over 
quadrats (see Table S2 for parameter values). Analyses excluded 
quadrats with problematic EAGB change measurements (see 
Section 2 for details)
F I G U R E  6   Probability density distribution (a) and cumulative 
probability distribution (b) of estimated above‐ground biomass 
(EAGB) density across all 10 × 10 m quadrats, with insets in 
(b) magnified to show inter‐census variation for cumulative 
probabilities of 0.05–0.1 and 0.9–0.95. (c) Variation over time in 
seven specific percentiles of the EAGB distribution over quadrats 
(colored points and lines) and in the overall mean EAGB density at 
the 50 ha scale (black solid points and lines). In panel (c), vertical 
lines show the 95% confidence intervals (CIs) from bootstrapping 
over quadrats, and horizontal gray shading shows the CIs for 
values in 2015 to enable easy visual assessment of how other years 
compare. The illustrated curves in panels (a) and (b) are truncated 
at the 0.2th and 99.8th percentiles of the distribution, but all values 
were included in analyses. Note that the EAGB densities can be 
converted to units of Mg/ha by multiplying by 100
8  |     RUTISHAUSER ET Al.
stem DBH increased from 1985 to 2000, and then leveled off, with 
parallel patterns for all quadrat biomass densities (Figure 8d).
4  | DISCUSSION
Here, we proposed and applied a straightforward way to control for 
successional stage in analyses of forest biomass change, by compar‐
ing forest patches at similar stages of gap recovery, as inferred by 
similar EAGBi density. Our analyses revealed strong temporal varia‐
tion in forest woody productivity and loss fluxes both at the whole 
plot level and when controlling for local (10 × 10 m) gap phase. There 
was no consistent long‐term trend in productivity, mortality, or bio‐
mass in this forest over 30 years, although the interannual variability 
in productivity and mortality is so strong that it could well mask a 
substantial trend. Woody productivity varied ~25% among 5 year 
census intervals, and woody losses ~15%, between 1985 and 2015. 
Because this variability is consistent even after controlling for gap 
phase, we conclude it arises from interannual variation in driving fac‐
tors, most obviously climate.
4.1 | Temporal climate variation in the tropics and 
its impacts
Tropical forests experience biologically important interannual 
climate variation in temperature, solar radiation, water availabil‐
ity, and the frequency and intensity of major storms. Interannual 
climate variation in tropical forests is related in part to irregu‐
larly periodic climate oscillations, including the El Niño Southern 
Oscillation (ENSO) and the Atlantic Multidecadal Oscillation 
(AMO). In much of the tropics, the El Niño phase of ENSO is as‐
sociated with drier, sunnier, and hotter conditions, whereas the 
La Niña phase is associated with wetter, cloudier, and cooler con‐
ditions (Holmgren, Scheffer, Ezcurra, Gutiérrez, & Mohren, 2001; 
Marengo, Tomasella, Alves, Soares, & Rodriguez, 2011). The AMO 
also influences climate variation at our site and other areas in the 
region (Elder, Balling, Cerveny, & Krahenbuhl, 2014). Tropical tree 
recruitment, growth, and mortality all vary among years in relation 
to such local climate variation, and thus, stand‐level productivity 
and mortality fluxes vary as well.
Extended periods of drier than normal conditions—that is, 
droughts—are associated with increased mortality and decreased 
F I G U R E  7   Time series of hypothetical whole‐plot productivity 
(a), loss (b), and net change (c) expected with and without variation 
among census intervals in functions specifying how estimated 
above‐ground biomass (EAGB) fluxes vary with initial biomass, and 
with and without variation among census intervals in initial EAGB 
distributions (EAGBi). See Section 2 for details. Vertical lines show 
95% confidence intervals obtained by bootstrapping over quadrats. 
The points are placed at the mid‐points of the census intervals, and 
jiggered horizontally to increase readability. Values are given in Table S3.  
Note that EAGB stocks and fluxes can be converted to units of  
Mg/ha by multiplying by 100. PDF, probability density function
F I G U R E  8   Variation among censuses in forest size structure 
at the whole plot level (black) and for different initial EAGB in 
10 × 10 m quadrats (colors). 95% Confidence intervals (vertical 
lines) were obtained from bootstrapping over quadrats, but are 
generally smaller than the dot size. The density of large stems and 
mean diameter are not reported for 1982 due to inconsistencies 
in measurement methods (Condit, 1998). DBH, diameter at breast 
height
     |  9RUTISHAUSER ET Al.
productivity in many tropical forests (Feldpausch et al., 2016; Phillips 
et al., 2009). The severe 1982–1983 El Niño made the dry season on 
BCI longer and harsher than normal, and was associated with higher 
tree mortality rates. Similarly, many sites in the Amazon and south‐
east Asia experienced elevated mortality during and after El Niño 
droughts (e.g., Brando et al., 2014; McDowell et al., 2018; Phillips 
et al., 2010; Qie et al., 2017; Rowland et al., 2015). El Niño droughts 
in the Amazon were also associated with decreased leaf area index, 
decreased photosynthesis (attributed to both stomatal closure 
and lower leaf area), and decreased woody productivity during the 
drought periods (Aguilos, Hérault, Burban, Wagner, & Bonal, 2018; 
Asner, Townsend, & Braswell, 2000; Rowland et al., 2014; Santos 
et al., 2018). Tree‐ring studies generally find that woody productivity 
is positively correlated with annual precipitation (Alfaro‐Sánchez, 
Muller‐Landau, Wright, & Camarero, 2017; Schippers, Sterck, 
Vlam, & Zuidema, 2015), consistent with the idea that dry years 
are bad for productivity. Temporal variation in woody growth in 
relation to water availability could be explained in part by shifts 
in allocation (rather than simply differences in total GPP and 
NPP), with trees allocating to increased xylem growth in wet pe‐
riods after droughts in order to replace drought‐damaged xylem 
(Trugman et al., 2018).
Although most studies have emphasized the negative effects 
of drier periods, drier years also bring higher solar radiation, and 
this can lead to more favorable conditions for forest productivity 
and tree survival if the dryness is not too extreme. On BCI, the 
drier periods of the 1987–1988 and 1997–1998 El Niños (Figure S18) 
were associated with elevated fruit production (Detto, 
Wright, Calderón, & Muller‐Landau, 2018), and the 5 year cen‐
sus periods encompassing these events featured above‐average 
woody productivity (this study, Meakem et al., 2017). El Niño 
events are associated with reduced rainfall during the wet season 
across northern South American and southern Central America 
(Ropelewski & Halpert, 1987). Moisture availability remains ample 
and reduced cloud cover and great solar radiation relieves light 
limitation (Graham, Mulkey, Kitajima, Phillips, & Wright, 2003), 
providing a straightforward explanation for observed increases 
in productivity (Detto et al., 2018; Wright & Calderón, 2006). 
Consistent with these findings, a previous analysis of multiple 
moist and wet tropical forests, including BCI, found higher tree 
growth in census intervals with higher solar radiation (Dong et al., 
2012). There are also reasons why dry years might exhibit lower 
rather than higher mortality. Reduced light limitation in sunnier 
years might reduce mortality rates for shaded understory tree, 
and fewer and smaller storms may reduce mortality due to wind‐
throws, lightning, and landslides (Aubry‐Kientz, Rossi, Wagner, & 
Hérault, 2015; Espírito‐Santo et al., 2010; Guzzetti, Peruccacci, 
Rossi, & Stark, 2008; Yanoviak, Gora, Burchfield, Bitzer, & Detto, 
2017).
Interannual variation in temperature may also contribute to in‐
terannual variation in tropical forest carbon dynamics. Warmer 
years have been associated with lower tropical tree growth in sev‐
eral studies (Clark, Piper, Keeling, & Clark, 2003; Dong et al., 2012; 
Schippers et al., 2015). Temperature directly affects photosynthesis 
and respiration rates, with photosynthesis responding unimodally to 
temperature, and respiration increasing continuously with tempera‐
ture (Lloyd & Farquhar, 2008; Slot & Winter, 2017). Furthermore, 
high leaf temperatures increase vapor pressure deficits and water 
demand, leading to stomatal closure, and it is this effect on water 
relations that appears to be the main contributor to temperature 
limitation on productivity (Aguilos et al., 2018; Slot & Winter, 2017). 
Higher atmospheric carbon dioxide concentrations increase plant 
water use efficiency, and may thereby partly mitigate the negative 
effects of higher temperatures and drier conditions on productivity 
(Holtum & Winter, 2010).
In addition to the direct effects of climate on plant physiological 
function and mortality risk, climate may also influence forest car‐
bon cycles indirectly by causing shifts in allocation, stand structure, 
functional composition, and influences of interacting species. Plants 
may increase or decrease allocation to woody growth, including add‐
ing to or drawing down carbon stores, depending on past and cur‐
rent climate conditions (Clark, Clark, & Oberbauer, 2013; Doughty 
et al., 2014; Trugman et al., 2018). Past mortality events from cli‐
mate alter tree age and size distributions, gap phase distributions, 
and thereby stand‐level carbon budgets (Anderegg, Schwalm, 
et al., 2015; Chazdon, 2003; Clark, 2007). Differential recruitment, 
survival, and growth of plant functional types under particular cli‐
mate conditions also alters functional composition, and thereby 
the response of forests to future climate. For example, mortality of 
drought‐intolerant species under drought can shift composition to‐
ward more drought‐tolerant species, with implications for forest car‐
bon budgets and resilience to future droughts (Esquivel‐Muelbert 
et al., 2019; McDowell et al., 2008; Ouédraogo, Mortier, Gourlet‐
Fleury, Freycon, & Picard, 2013). The proliferation of pathogens and 
insect herbivores, and thus their effects on trees, is also highly sen‐
sitive to climate conditions (Anderegg, Hicke, et al., 2015; Van Bael 
et al., 2004).
4.2 | Detecting and attributing long‐term change
Tropical forests are experiencing global atmospheric and climate 
change. Mean atmospheric CO2 concentration is increasing continu‐
ously, with an increase of 15% over the duration of our study (from 
346 ppm in 1985 to 400 ppm in 2015, Figure S18g). Temperatures 
are increasing in most tropical regions, and precipitation patterns 
are changing in region‐specific ways (Buckley & Huey, 2016; Malhi & 
Wright, 2004). At our site, the focal 30 year period exhibited trends 
of +0.15°C per decade in daily maximum temperature, +0.04°C per 
decade in daily minimum temperature, +5 mm/year in annual pre‐
cipitation,	+0.05%	soil	moisture	per	year,	and	−1.3	W	m−2 year−1 in 
annual solar radiation. Trends in annual means of these BCI climate 
metrics over the 30 year period alone were not statistically signifi‐
cant after Bonferroni correction (individual p‐values of .075, .44, 
.66, .033, and .48, respectively), reflecting the relative strength of 
interannual variation. This interannual variation limits the ability to 
detect long‐term trends—but failure to reject the null hypothesis of 
10  |     RUTISHAUSER ET Al.
no increase should not be misinterpreted as support for constant 
climate conditions.
These atmospheric and climate changes are hypothesized to be 
changing tropical forest carbon dynamics (Lewis, Malhi, & Phillips, 
2004). CO2 is needed for photosynthesis, and physiological models 
suggest that the observed increase in CO2 should increase growth 
(Phillips & Lewis, 2014), although there is uncertainty about the 
degree to which growth responses are limited by availability of 
other nutrients, and within‐plant carbon demands (Körner, 2009, 
2003b; Wright et al., 2011). Spatial variation in climate among 
tropical forests is associated with variation in tropical forest car‐
bon stocks and fluxes (Álvarez‐Dávila et al., 2017; Becknell, Kissing 
Kucek, & Powers, 2012; Poorter et al., 2016; Taylor et al., 2017), 
suggesting that directional climate changes should also change 
forest carbon budgets in the long term. As discussed above, tropi‐
cal forests are also sensitive to temporal variation in climate condi‐
tions, further suggesting that climate change will matter. However, 
the relationship of short‐term to long‐term climate responses is 
not straightforward, because the importance of particular mecha‐
nisms shifts with the timescale. Allocational shifts to/from woody 
productivity, reproductive output, and carbohydrate stores play a 
large role in variation over seasonal to annual time scales (Dickman 
et al., 2018; Doughty et al., 2014), but are likely to play little role 
at decadal and longer time scales. In contrast, shifts in functional 
composition of the tree community are likely to play a large role 
in responses to long‐term climate variation (van der Sande et al., 
2016), but very little role in short‐term responses. Such composi‐
tional shifts are in general expected to decrease the sensitivity of 
forest carbon fluxes to climate variation, as the forest becomes in‐
creasingly dominated by species that do relatively well in the new 
climate conditions (Esquivel‐Muelbert et al., 2019; Fauset et al.., 
2015; Feldpausch et al., 2016).
However, evidence of long‐term changes in tropical forest 
carbon cycling is mixed, with divergent patterns across stud‐
ies. Increasing woody productivity and mortality were found in 
analyses of the RAINFOR network of Amazonian plots (Brienen 
et al., 2015), but not in site‐specific studies of BCI (this study) or 
La Selva (Clark, Asao, et al., 2017; Clark et al., 2013; Clark, Clark, 
et al., 2017). On average, studies report increased biomass den‐
sities in old‐growth tropical forests (Brienen et al., 2015; Chave 
et al., 2008, Lewis et al., 2009; Muller‐Landau et al., 2014; Qie 
et al., 2017; Rutishauser, Wagner, Herault, Nicolini, & Blanc, 2010), 
but this pattern too is not universal (this study, Chave et al., 2008; 
Clark, Clark, et al., 2017; Feeley et al., 2007) and the causes of 
observed increases continue to be debated (Clark, 2004; Lewis 
et al., 2009; Muller‐Landau, 2009; Wright, 2013). It is notable that 
two of the most intensive long‐term studies of focal sites, this 30 
year study of BCI and Clark, Clark, et al. (2017)'s 40 year study of 
La Selva, find multidecadal stability in forest structure and dy‐
namics, in striking contrast to the “Bigger and Faster” hypothesis.
The large uncertainty surrounding estimated fluxes and their 
long‐term trends observed here, even with such a large‐scale 
long‐term study, highlights the difficulty of accurately capturing 
changes in forest dynamics from field data (Clark, Asao, et al., 
2017). Wagner, Rutishauser, Blanc, and Herault (2010) calculated 
that estimating even the mean EAGB loss within 20% error with 
95% confidence in a particular forest would require ~200 ha‐years 
of monitoring, far more than the usual size and duration of most 
forest monitoring experiments. The effort required to accurately 
estimate a long‐term trend is much higher still—even a trend as 
large as 1% a year amounts to only 10% a decade, within the error 
limits of the Wagner et al. (2010) calculation. Detection of long‐
term trends is further complicated by sensitivity of estimated 
trends to the precise methods of data quality assessment and qual‐
ity control (QAQC; Figure S18). Many common QAQC procedures 
result in systematic biases in whole‐plot statistics (Cushman et al., 
2014; Muller‐Landau et al., 2014). Our analysis followed current 
best practices for calculating EAGB fluxes, including standardizing 
effective POM (Cushman et al., 2014) and avoiding data “clean‐
ing” procedures that can systematically bias whole‐plot statistics 
(Muller‐Landau et al., 2014).
4.3 | Conclusion and future directions
The tropical forest on BCI exhibited large temporal variability in 
stand‐level productivity and mortality fluxes over 30 years, even 
after controlling for disturbance‐recovery dynamics, but no long‐
term trend. Observed temporal variability was aligned to some 
degree with ENSO‐related climate variation, but with a markedly 
different pattern than is usually reported, highlighting the complex‐
ity of tropical forest climate responses. While the drier conditions 
of El Niño events have been associated with increased tree mor‐
tality and reduced productivity in the Amazon (Feldpausch et al., 
2016; Lewis, Brando, Phillips, Heijden, & Nepstad, 2011), the drier 
and sunnier conditions during El Niño events were associated with 
enhanced productivity and no change in mortality on BCI (Detto 
et al., 2018; Meakem et al., 2017). No long‐term directional trend in 
productivity, mortality, or biomass was evident in our data, but such 
a trend could well be obscured by the strong interannual variation.
A better understanding of tropical forest responses to climate 
variation and long‐term trends requires both more data and in‐
tegration with mechanistic models. Although climate and atmo‐
spheric change is global, climate effects are locally variable, and 
forest responses are dependent on a myriad of other local factors, 
including species composition and disturbance history, resulting 
in highly variable EAGB dynamics (Clark et al., 2013). Future mul‐
tisite analyses of large plot datasets using the methods developed 
here could provide insights into temporal and spatial variation in 
tropical forest carbon fluxes (Chave et al., 2008; van der Sande 
et al., 2016; Wagner et al., 2016). However, any studies based on 
extant plot datasets are likely to fall short of what is needed to 
quantify overall trends given high temporal and spatial variability, 
and the nonsystematic placement of sampling plots (Marvin et al., 
2014; McMichael et al., 2017; Saatchi et al., 2015; Wright, 2006). 
Improvements in remote sensing ultimately promise more compre‐
hensive and consistent measurements of tropical forest structure 
     |  11RUTISHAUSER ET Al.
and function, and their change over time (Bastin et al., 2018; Rödig 
et al., 2018; Saatchi et al., 2011), although the contribution of re‐
covery from past disturbances remains to be resolved (McMichael 
et al., 2017; Palace et al., 2017). Studies integrating mechanis‐
tic models with data on climate drivers and forest dynamics are 
uniquely well suited to gleaning insights into the ultimate mech‐
anisms behind observed patterns and to disentangling the roles 
of temperature, rainfall, atmospheric carbon dioxide, soils, species 
composition, and site history (Levine et al., 2016; Schippers et al., 
2015).
ACKNOWLEDG EMENTS
E.R. and S.J.D. were supported by the Next Generation 
Ecosystem Experiments‐Tropics, funded by the US Department 
of Energy, Office of Science, Office of Biological and Environ‐
mental Research. We gratefully acknowledge the contributions 
to the Barro Colorado Island 50 ha plot of Robin Foster, co‐
founder of the plot; Rolando Pérez and Salomón Aguilar for 
species identification; Suzanne Lao for data management; 
Steven Dolins for database design; and hundreds of field‐work‐
ers over the years. The BCI forest dynamics plot dataset was 
made possible by the financial support of the National Science 
Foundation, the Smithsonian Tropical Research Institute, and 
the MacArthur Foundation. We thank Deborah A. Clark and 
an anonymous reviewer for constructive comments on this 
manuscript.
CONFLIC T OF INTERE S T
All authors declare no conflict of interest.
ORCID
Ervan Rutishauser  https://orcid.org/0000‐0003‐1182‐4032 
R E FE R E N C E S
Aguilos, M., Hérault, B., Burban, B., Wagner, F., & Bonal, D. (2018). What 
drives long‐term variations in carbon flux and balance in a tropical 
rainforest in French Guiana? Agricultural & Forest Meteorology, 253–
254, 114–123. https ://doi.org/10.1016/j.agrfo rmet.2018.02.009
Alfaro‐Sánchez, R., Muller‐Landau, H. C., Wright, S. J., & Camarero, J. J. 
(2017). Growth and reproduction respond differently to climate in 
three Neotropical tree species. Oecologia, 184, 531–541. https ://doi.
org/10.1007/s00442‐017‐3879‐3
Álvarez‐Dávila, E., Cayuela, L., González‐Caro, S., Aldana, A. M., 
Stevenson, P. R., Phillips, O., … Rey‐Benayas, J. M. (2017). Forest 
biomass density across large climate gradients in northern South 
America is related to water availability but not with tempera‐
ture. PLoS ONE, 12(3), e0171072. https ://doi.org/10.1371/journ 
al.pone.0171072
Anderegg, W. R. L., Hicke, J. A., Fisher, R. A., Allen, C. D., Aukema, J., 
Bentz, B., … Zeppel, M. (2015). Tree mortality from drought, insects, 
and their interactions in a changing climate. New Phytologist, 208, 
674–683. https ://doi.org/10.1111/nph.13477 
Anderegg, W. R. L., Schwalm, C., Biondi, F., Camarero, J. J., Koch, G., 
Litvak, M., … Pacala, S. (2015). Pervasive drought legacies in forest 
ecosystems and their implications for carbon cycle models. Science, 
349, 528–532. https ://doi.org/10.1126/scien ce.aab1833
Asner, G. P., Townsend, A. R., & Braswell, B. H. (2000). Satellite obser‐
vation of El Niño effects on Amazon Forest phenology and pro‐
ductivity. Geophysical Research Letters, 27, 981–984. https ://doi.
org/10.1029/1999G L011113
Aubry‐Kientz, M., Rossi, V., Wagner, F., & Hérault, B. (2015). Identifying 
climatic drivers of tropical forest dynamics. Biogeosciences, 12, 5583–
5596. https ://doi.org/10.5194/bg‐12‐5583‐2015
Baker, T. R., Phillips, O. L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore, A., 
… Vasquez Martinez, R. (2004). Increasing biomass in Amazonian 
forest plots. Philosophical Transactions of the Royal Society of London. 
Series B: Biological Sciences, 359, 353–365. https ://doi.org/10.1098/
rstb.2003.1422
Baskerville, G. (1972). Use of logarithmic regression in the estimation of 
plant biomass. Canadian Journal of Forest Research, 2, 49–53. https ://
doi.org/10.1139/x72‐009
Bastin, J.‐F., Rutishauser, E., Kellner, J. R., Saatchi, S., Pélissier, R., Hérault, 
B., … Zebaze, D. (2018). Pan‐tropical prediction of forest structure 
from the largest trees. Global Ecology and Biogeography, 27, 1366–
1383. https ://doi.org/10.1111/geb.12803 
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear 
mixed‐effects models using lme4. Journal of Statistical Software, 
67(1), 1–48. https ://doi.org/10.18637/ jss.v067.i01
Becknell, J. M., Kissing Kucek, L., & Powers, J. S. (2012). Aboveground 
biomass in mature and secondary seasonally dry tropical for‐
ests: A literature review and global synthesis. Forest Ecology and 
Management, 276, 88–95.
Bormann, F. H., & Likens, G. E. (1979). Pattern and process in a forested 
ecosystem: Disturbance, development and the steady state based 
on the Hubbard Brook Ecosystem Study (1st ed.). New York, NY: 
Springer‐Verlag.
Brando, P. M., Balch, J. K., Nepstad, D. C., Morton, D. C., Putz, F. E., Coe, 
M. T., … Soares‐Filho, B. S. (2014). Abrupt increases in Amazonian 
tree mortality due to drought‐fire interactions. Proceedings of the 
National Academy of Sciences of the United States of America, 111, 
6347–6352. https ://doi.org/10.1073/pnas.13054 99111 
Brienen, R. J. W., Phillips, O. L., Feldpausch, T. R., Gloor, E., Baker, T. 
R., Lloyd, J., … Zagt, R. J. (2015). Long‐term decline of the Amazon 
carbon sink. Nature, 519, 344–348. https ://doi.org/10.1038/natur 
e14283
Brncic, T. M., Willis, K. J., Harris, D. J., & Washington, R. (2006). Culture 
or climate? The relative influences of past processes on the com‐
position of the lowland Congo rainforest. Philosophical Transactions 
of the Royal Society B: Biological Sciences, 362, 229–242. https ://doi.
org/10.1098/rstb.2006.1982
Buckley, L. B., & Huey, R. B. (2016). Temperature extremes: Geographic 
patterns, recent changes, and implications for organismal vulnerabil‐
ities. Global Change Biology, 22, 3829–3842. https ://doi.org/10.1111/
gcb.13313 
Chave, J., Condit, R., Muller‐Landau, H. C., Thomas, S. C., Ashton, P. S., 
Bunyavejchewin, S., … Losos, E. C. (2008). Assessing evidence for 
a pervasive alteration in tropical tree communities. PLoS Biology, 6, 
455–462. https ://doi.org/10.1371/journ al.pbio.0060045
Chave, J., Réjou‐Méchain, M., Búrquez, A., Chidumayo, E., Colgan, M. S., 
Delitti, W. B., … Vieilledent, G. (2014). Improved allometric models to 
estimate the aboveground biomass of tropical trees. Global Change 
Biology, 20, 3177–3190. https ://doi.org/10.1111/gcb.12629 
Chazdon, R. L. (2003). Tropical forest recovery: Legacies of human impact 
and natural disturbances. Perspectives in Plant Ecology, Evolution and 
Systematics, 6, 51–71. https ://doi.org/10.1078/1433‐8319‐00042 
Clark, D. (2004). Sources or sinks? The responses of tropical forests to cur‐
rent and future climate and atmospheric composition. Philosophical 
12  |     RUTISHAUSER ET Al.
Transactions of the Royal Society of London. Series B: Biological Sciences, 
359, 477–491. https ://doi.org/10.1098/rstb.2003.1426
Clark, D. A. (2007). Detecting tropical forests' responses to global climatic 
and atmospheric change: Current challenges and a way forward. 
Biotropica, 39, 4–19. https ://doi.org/10.1111/j.1744‐7429.2006. 
00227.x
Clark, D. A., Asao, S., Fisher, R., Reed, S., Reich, P. B., Ryan, M. G., … 
Yang, X. (2017). Field data to benchmark the carbon cycle models 
for tropical forests. Biogeosciences, 14(20), 4663–4690. https ://doi.
org/10.5194/bg‐14‐4663‐2017
Clark, D. A., Clark, D. B., & Oberbauer, S. F. (2013). Field‐quantified re‐
sponses of tropical rainforest aboveground productivity to increas‐
ing CO2 and climatic stress, 1997–2009: Tropical rainforest produc‐
tivity responses. Journal of Geophysical Research: Biogeosciences, 118, 
783–794. https ://doi.org/10.1002/jgrg.20067 
Clark, D. B., Clark, D. A., Oberbauer, S. F., & Kellner, J. R. (2017). 
Multidecadal stability in tropical rain forest structure and dynamics 
across an old‐growth landscape. PLoS ONE, 12, e0183819. https ://
doi.org/10.1371/journ al.pone.0183819
Clark, D. A., Piper, S. C., Keeling, C. D., & Clark, D. B. (2003). Tropical 
rain forest tree growth and atmospheric carbon dynamics linked to 
interannual temperature variation during 1984–2000. Proceedings of 
the National Academy of Sciences of the United States of America, 100, 
5852–5857. https ://doi.org/10.1073/pnas.09359 03100 
Condit, R. (1998). Tropical forest census plots: Methods and results from 
Barro Colorado Island, Panama and a comparison with other plots. 
Berlin, Germany: Springer Science & Business Media.
Condit, R., Hubbell, S. P., & Foster, R. B. (1996). Changes in tree species 
abundance in a Neotropical forest: Impact of climate change. Journal 
of Tropical Ecology, 12, 231–256. https ://doi.org/10.1017/S0266 
46740 0009433
Cushman, K. C., Muller‐Landau, H. C., Condit, R. S., & Hubbell, S. P. 
(2014). Improving estimates of biomass change in buttressed trees 
using tree taper models. Methods in Ecology and Evolution, 5, 573–
582. https ://doi.org/10.1111/2041‐210X.12187 
Denslow, J. S., Ellison, A. M., & Sanford, R. E. (1998). Treefall gap 
size effects on above‐ and below‐ground processes in a trop‐
ical wet forest. Journal of Ecology, 86, 597–609. https ://doi.
org/10.1046/j.1365‐2745.1998.00295.x
Detto, M., Wright, S. J., Calderón, O., & Muller‐Landau, H. C. (2018). 
Resource acquisition and reproductive strategies of tropical forest in 
response to the El Niño‐Southern Oscillation. Nature Communications, 
9, 913. https ://doi.org/10.1038/s41467‐018‐03306‐9
Dickman, L. T., McDowell, N. G., Grossiord, C., Collins, A. D., Wolfe, 
B. T., Detto, M., … Chambers, J. Q. (2018). Homoeostatic mainte‐
nance of nonstructural carbohydrates during the 2015–2016 El 
Niño drought across a tropical forest precipitation gradient. Plant, 
Cell and Environment, 42(5), 1705–1714. https ://doi.org/10.1111/
pce.13501 
Dong, S. X., Davies, S. J., Ashton, P. S., Bunyavejchewin, S., Supardi, M. 
N., Kassim, A. R., … Moorcroft, P. R. (2012). Variability in solar ra‐
diation and temperature explains observed patterns and trends in 
tree growth rates across four tropical forests. Proceedings of the 
Royal Society B: Biological Sciences, 279, 3923–3931. https ://doi.
org/10.1098/rspb.2012.1124
dos Santos, V. A. H. F., Ferreira, M. J., Rodrigues, J. V. F. C., Garcia, M. 
N., Ceron, J. V. B., Nelson, B. W., & Saleska, S. R. (2018). Causes of 
reduced leaf‐level photosynthesis during strong El Niño drought in a 
Central Amazon forest. Global Change Biology, 24, 4266–4279. https 
://doi.org/10.1111/gcb.14293 
Doughty, C. E., Malhi, Y., Araujo‐Murakami, A., Metcalfe, D. B., Silva‐
Espejo, J. E., Arroyo, L., … Ledezma, R. (2014). Allocation trade‐
offs dominate the response of tropical forest growth to seasonal 
and interannual drought. Ecology, 95, 2192–2201. https ://doi.
org/10.1890/13‐1507.1
Doughty, C. E., Metcalfe, D. B., Girardin, C. A. J., Amézquita, F. F., 
Cabrera, D. G., Huasco, W. H., … Malhi, Y. (2015). Drought impact on 
forest carbon dynamics and fluxes in Amazonia. Nature, 519, 78–82. 
https ://doi.org/10.1038/natur e14213
Elder, R. C., Balling, R. C., Cerveny, R. S., & Krahenbuhl, D. (2014). Regional 
variability in drought as a function of the Atlantic Multidecadal 
Oscillation. Caribbean Journal of Science, 48, 31–44. https ://doi.
org/10.18475/ cjos.v48i1.a5
Espírito‐Santo, F. D. B., Gloor, M., Keller, M., Malhi, Y., Saatchi, S., 
Nelson, B., … Phillips, O. L. (2014). Size and frequency of natural 
forest disturbances and the Amazon forest carbon balance. Nature 
Communications, 5, 3434. https ://doi.org/10.1038/ncomm s4434 
Espírito‐Santo, F. D. B., Keller, M., Braswell, B., Nelson, B. W., Frolking, 
S., & Vicente, G. (2010). Storm intensity and old‐growth forest dis‐
turbances in the Amazon region. Geophysical Research Letters, 37, 
L11403. https ://doi.org/10.1029/2010g l043146
Esquivel‐Muelbert, A., Baker, T. R., Dexter, K. G., Lewis, S. L., Brienen, 
R. J. W., Feldpausch, T. R., … Phillips, O. L. (2019). Compositional re‐
sponse of Amazon forests to climate change. Global Change Biology, 
25, 39–56. https ://doi.org/10.1111/gcb.14413 
Fauset, S., Johnson, M. O., Gloor, M., Baker, T. R., Monteagudo, A. 
M., Brienen, R. J. W., … Phillips, O. L. (2015). Hyperdominance in 
Amazonian forest carbon cycling. Nature Communications, 6, 6857. 
https ://doi.org/10.1038/ncomm s7857 
Feeley, K. J., Wright, J., Supardi, N., Kassim, A. R., & Davies, S. J. (2007). 
Decelerating growth in tropical forest trees. Ecology Letters, 10, 461–
469. https ://doi.org/10.1111/j.1461‐0248.2007.01033.x
Feldpausch, T. R., Phillips, O. L., Brienen, R. J. W., Gloor, E., Lloyd, J., 
Lopez‐Gonzalez, G., … Dávila, E. Á. (2016). Amazon forest response 
to repeated droughts. Global Biogeochemical Cycles, 30, 964–982. 
https ://doi.org/10.1002/2015G B005133
Fisher, J. I., Hurtt, G. C., Thomas, R. Q., & Chambers, J. Q. (2008). 
Clustered disturbances lead to bias in large‐scale estimates based 
on forest sample plots. Ecology Letters, 11, 554–563. https ://doi.
org/10.1111/j.1461‐0248.2008.01169.x
Goodman, R. C., Phillips, O. L., del Castillo Torres, D., Freitas, L., Cortese, 
S. T., Monteagudo, A., & Baker, T. R. (2013). Amazon palm biomass 
and allometry. Forest Ecology and Management, 310, 994–1004. https 
://doi.org/10.1016/j.foreco.2013.09.045
Graham, E. A., Mulkey, S. S., Kitajima, K., Phillips, N. G., & Wright, S. J. 
(2003). Cloud cover limits net CO2 uptake and growth of a rainfor‐
est tree during tropical rainy seasons. Proceedings of the National 
Academy of Sciences of the United States of America, 100, 572–576. 
https ://doi.org/10.1073/pnas.01330 45100 
Guzzetti, F., Peruccacci, S., Rossi, M., & Stark, C. P. (2008). The rain‐
fall intensity–duration control of shallow landslides and debris 
flows: An update. Landslides, 5, 3–17. https ://doi.org/10.1007/
s10346‐007‐0112‐1
Harms, K. E., Condit, R., Hubbell, S. P., & Foster, R. B. (2001). 
Habitat associations of trees and shrubs in a 50‐ha Neotropical 
forest plot. Journal of Ecology, 89, 947–959. https ://doi.
org/10.1046/j.0022‐0477.2001.00615.x
Holmgren, M., Scheffer, M., Ezcurra, E., Gutiérrez, J. R., & Mohren, G. M. 
J. (2001). El Niño effects on the dynamics of terrestrial ecosystems. 
Trends in Ecology & Evolution, 16, 89–94. https ://doi.org/10.1016/
S0169‐5347(00)02052‐8
Holtum, J. A., & Winter, K. (2010). Elevated [CO2] and forest vegetation: 
More a water issue than a carbon issue? Functional Plant Biology, 37, 
694–702. https ://doi.org/10.1071/FP10001
Hubbell, S. P., Foster, R. B., O'Brien, S. T., Harms, K. E., Condit, R., 
Wechsler, B., … De Lao, S. L. (1999). Light‐gap disturbances, recruit‐
ment limitation, and tree diversity in a neotropical forest. Science, 
283, 554–557. https ://doi.org/10.1126/scien ce.283.5401.554
Kohyama, T. S., Kohyama, T. I., & Sheil, D. (2019). Estimating net biomass 
production and loss from repeated measurements of trees in forests 
     |  13RUTISHAUSER ET Al.
and woodlands: Formulae, biases and recommendations. Forest 
Ecology and Management, 433, 729–740. https ://doi.org/10.1016/ 
j.foreco.2018.11.010
Körner, C. (2003a). Slow in, rapid out – Carbon flux studies and Kyoto 
targets. Science, 300, 1242–1243. https ://doi.org/10.1126/scien 
ce.1084460
Körner, C. (2003b). Carbon limitation in trees. Journal of Ecology, 91, 
4–17. https ://doi.org/10.1046/j.1365‐2745.2003.00742.x
Körner, C. (2009). Responses of humid tropical trees to rising CO2. 
Annual Review of Ecology Evolution and Systematics, 40, 61–79. https 
://doi.org/10.1146/annur ev.ecols ys.110308.120217
Körner, C. (2015). Paradigm shift in plant growth control. Current 
Opinion in Plant Biology, 25, 107–114. https ://doi.org/10.1016/ 
j.pbi.2015.05.003
Levine, N. M., Zhang, K., Longo, M., Baccini, A., Phillips, O. L., Lewis, S. L., 
… Moorcroft, P. R. (2016). Ecosystem heterogeneity determines the 
ecological resilience of the Amazon to climate change. Proceedings of 
the National Academy of Sciences of the United States of America, 113, 
793–797. https ://doi.org/10.1073/pnas.15113 44112 
Levis, C., Costa, F. R. C., Bongers, F., Peña‐Claros, M., Clement, C. R., 
Junqueira, A. B., … Ter Steege, H. (2017). Persistent effects of pre‐
Columbian plant domestication on Amazonian forest composition. 
Science, 355, 925–931. https ://doi.org/10.1126/scien ce.aal0157
Lewis, S. L., Brando, P. M., Phillips, O. L., van der Heijden, G. M. F., & 
Nepstad, D. (2011). The 2010 Amazon drought. Science, 331, 554. 
https ://doi.org/10.1126/scien ce.1200807
Lewis, S. L., Lloyd, J., Sitch, S., Mitchard, E. T. A., & Laurance, W. F. (2009). 
Changing ecology of tropical forests: Evidence and drivers. Annual 
Review of Ecology, Evolution, and Systematics, 40, 529–549. https ://
doi.org/10.1146/annur ev.ecols ys.39.110707.173345
Lewis, S. L., Malhi, Y., & Phillips, O. L. (2004). Fingerprinting the impacts 
of global change on tropical forests. Philosophical Transactions of 
the Royal Society B: Biological Sciences, 359, 437–462. https ://doi.
org/10.1098/rstb.2003.1432
Lewis, S. L., Sonke, B., Sunderland, T., Begne, S. K., Lopez‐Gonzalez, 
G., van der Heijden, G. M. F., … Zemagho, L. (2013). Above‐ground 
biomass and structure of 260 African tropical forests. Philosophical 
Transactions of the Royal Society B: Biological Sciences, 368, 20120295–
20120295. https ://doi.org/10.1098/rstb.2012.0295
Liu, J., Bowman, K. W., Schimel, D. S., Parazoo, N. C., Jiang, Z., Lee, M., … 
Eldering, A. (2017). Contrasting carbon cycle responses of the trop‐
ical continents to the 2015–2016 El Niño. Science, 358, eaam5690. 
https ://doi.org/10.1126/scien ce.aam5690
Lloyd, J., & Farquhar, G. D. (2008). Effects of rising temperatures and 
[CO2] on the physiology of tropical forest trees. Philosophical 
Transactions of the Royal Society B: Biological Sciences, 363, 1811–
1817. https ://doi.org/10.1098/rstb.2007.0032
Loader, C. (1999). Local regression and likelihood. New York, NY: Springer 
Science & Business Media.
Magnabosco Marra, D., Trumbore, S. E., Higuchi, N., Ribeiro, G. H. P. M., 
Negrón‐Juárez, R. I., Holzwarth, F., … Wirth, C. (2018). Windthrows 
control biomass patterns and functional composition of Amazon for‐
ests. Global Change Biology, 24, 5867–5881. https ://doi.org/10.1111/
gcb.14457 
Malhi, Y., & Wright, J. (2004). Spatial patterns and recent trends in the 
climate of tropical rainforest regions. Philosophical Transactions of the 
Royal Society of London. Series B: Biological Sciences, 359, 311–329. 
https ://doi.org/10.1098/rstb.2003.1433
Marengo, J. A., Tomasella, J., Alves, L. M., Soares, W. R., & Rodriguez, D. 
A. (2011). The drought of 2010 in the context of historical droughts 
in the Amazon region. Geophysical Research Letters, 38. https ://doi.
org/10.1029/2011G L047436
Marvin, D. C., Asner, G. P., Knapp, D. E., Anderson, C. B., Martin, R. E., 
Sinca, F., & Tupayachi, R. (2014). Amazonian landscapes and the bias 
in field studies of forest structure and biomass. Proceedings of the 
National Academy of Sciences of the United States of America, 111, 
E5224–E5232. https ://doi.org/10.1073/pnas.14129 99111 
Mayle, F. E., Burbridge, R., & Killeen, T. J. (2000). Millennial‐scale dynam‐
ics of southern Amazonian Rain Forests. Science, 290, 2291–2294. 
https ://doi.org/10.1126/scien ce.290.5500.2291
McDowell, N., Allen, C. D., Anderson‐Teixeira, K., Brando, P., Brienen, R., 
Chambers, J., … Xu, X. (2018). Drivers and mechanisms of tree mor‐
tality in moist tropical forests. New Phytologist, 219, 851–869. https ://
doi.org/10.1111/nph.15027 
McDowell, N., Pockman, W. T., Allen, C. D., Breshears, D. D., Cobb, N., 
Kolb, T., … Yepez, E. A. (2008). Mechanisms of plant survival and 
mortality during drought: Why do some plants survive while oth‐
ers succumb to drought? New Phytologist, 178, 719–739. https ://doi.
org/10.1111/j.1469‐8137.2008.02436.x
McMichael, C. N. H., Matthews‐Bird, F., Farfan‐Rios, W., & Feeley, K. 
J. (2017). Ancient human disturbances may be skewing our under‐
standing of Amazonian forests. Proceedings of the National Academy 
of Sciences of the United States of America, 114, 522–527. https ://doi.
org/10.1073/pnas.16145 77114 
Meakem, V., Tepley, A. J., Gonzalez‐Akre, E. B., Herrmann, V., Muller‐
Landau, H. C., Wright, S. J., … Anderson‐Teixeira, K. J. (2017). Role 
of tree size in moist tropical forest carbon cycling and water deficit 
responses. New Phytologist, 219, 947–958. https ://doi.org/10.1111/
nph.14633 
Muller‐Landau, H. C. (2009). Carbon cycle: Sink in the African jungle. 
Nature, 457, 969–970. https ://doi.org/10.1038/457969a
Muller‐Landau, H. C., Detto, M., Chisholm, R. A., Hubbell, S. P., & Condit, 
R. (2014). Detecting and projecting changes in forest biomass from 
plot data. In D. A. Coomes, D. F. R. P. Burslem, & W. D. Simonson 
(Eds.), Forests and global change, ecological review (pp. 381–416). 
Cambridge, UK: Cambridge University Press.
Oslisly, R., White, L., Bentaleb, I., Favier, C., Fontugne, M., Gillet, J.‐F., & 
Sebag, D. (2013). Climatic and cultural changes in the west Congo 
Basin forests over the past 5000 years. Philosophical Transactions of 
the Royal Society B: Biological Sciences, 368, 20120304. https ://doi.
org/10.1098/rstb.2012.0304
Ouédraogo, D.‐Y., Mortier, F., Gourlet‐Fleury, S., Freycon, V., & Picard, 
N. (2013). Slow‐growing species cope best with drought: Evidence 
from long‐term measurements in a tropical semi‐deciduous moist 
forest of Central Africa. Journal of Ecology, 101, 1459–1470. https ://
doi.org/10.1111/1365‐2745.12165 
Palace, M. W., McMichael, C. N. H., Braswell, B. H., Hagen, S. C., Bush, M. 
B., Neves, E., … Frolking, S. (2017). Ancient Amazonian populations 
left lasting impacts on forest structure. Ecosphere, 8, e02035. https 
://doi.org/10.1002/ecs2.2035
Paton, S. (2018). 2017 Meteorological and hydrological summary for Barro 
Colorado Island. Balboa, Panama: Smithsonian Tropical Research Institute.
Phillips, O. L., Aragao, L. E. O. C., Lewis, S. L., Fisher, J. B., Lloyd, J., 
Lopez‐Gonzalez, G., … Torres‐Lezama, A. (2009). Drought sensitiv‐
ity of the Amazon Rainforest. Science, 323, 1344–1347. https ://doi.
org/10.1126/scien ce.1164033
Phillips, O. L., & Lewis, S. L. (2014). Recent changes in tropical forest 
biomass and dynamics. In D. A. Coomes, D. F. R. P. Burslem, & W. 
D. Simonson (Eds.), Forests and global change, ecological review (pp. 
77–108). Cambridge, UK: Cambridge University Press.
Phillips, O. L., Malhi, Y., Higuchi, N., Laurance, W., Nunez, P. V., Vasquez, 
R. M., … Grace, J. (1998). Changes in the carbon balance of tropical 
forest: Evidence from long‐term plots. Science, 282, 439–442. https 
://doi.org/10.1126/scien ce.282.5388.439
Phillips, O. L., van der Heijden, G., Lewis, S. L., López‐González, G., 
Aragão, L. E. O. C., Lloyd, J., … Vilanova, E. (2010). Drought–mortal‐
ity relationships for tropical forests. New Phytologist, 187, 631–646. 
https ://doi.org/10.1111/j.1469‐8137.2010.03359.x
Poorter, L., Bongers, F., Aide, T. M., Almeyda Zambrano, A. M., Balvanera, 
P., Becknell, J. M., … Rozendaal, D. M. A. (2016). Biomass resilience of 
14  |     RUTISHAUSER ET Al.
Neotropical secondary forests. Nature, 530(7589), 211–214. https ://
doi.org/10.1038/natur e16512
Qie, L., Lewis, S. L., Sullivan, M. J. P., Lopez‐Gonzalez, G., Pickavance, G. C., 
Sunderland, T., … Phillips, O. L. (2017). Long‐term carbon sink in Borneo's 
forests halted by drought and vulnerable to edge effects. Nature 
Communications, 8(1), 1966. https ://doi.org/10.1038/s41467‐017‐01997‐0
Rödig, E., Cuntz, M., Rammig, A., Fischer, R., Taubert, F., & Huth, A. 
(2018). The importance of forest structure for carbon flux estimates 
in the Amazon rainforest. Environmental Research Letters, 13, 054013. 
https ://doi.org/10.1088/1748‐9326/aabc61
Ropelewski, C. F., & Halpert, M. S. (1987). Global and regional scale 
precipitation patterns associated with the El Niño/Southern 
Oscillation. Monthly Weather Review, 115, 1606–1626. https ://doi.
org/10.1175/1520‐0493(1987)115<1606:GARSP P>2.0.CO;2
Rowland, L., Lobo‐do‐Vale, R. L., Christoffersen, B. O., Melém, E. A., 
Kruijt, B., Vasconcelos, S. S., … Meir, P. (2015). After more than a de‐
cade of soil moisture deficit, tropical rainforest trees maintain photo‐
synthetic capacity, despite increased leaf respiration. Global Change 
Biology, 21, 4662–4672. https ://doi.org/10.1111/gcb.13035 
Rowland, L., Malhi, Y., Silva‐Espejo, J. E., Farfán‐Amézquita, F., Halladay, 
K., Doughty, C. E., … Phillips, O. L. (2014). The sensitivity of wood 
production to seasonal and interannual variations in climate in a 
lowland Amazonian rainforest. Oecologia, 174, 295–306. https ://doi.
org/10.1007/s00442‐013‐2766‐9
Rutishauser, E., Wagner, F., Herault, B., Nicolini, E. A., & Blanc, L. (2010). 
Contrasting above‐ground biomass balance in a Neotropical rain 
forest. Journal of Vegetation Science, 21, 672–682. https ://doi.
org/10.1111/j.1654‐1103.2010.01175.x
Saatchi, S. S., Harris, N. L., Brown, S., Lefsky, M., Mitchard, E. T. A., Salas, 
W., … Morel, A. (2011). Benchmark map of forest carbon stocks in 
tropical regions across three continents. Proceedings of the National 
Academy of Sciences of the United States of America, 108, 9899–9904. 
https ://doi.org/10.1073/pnas.10195 76108 
Saatchi, S., Mascaro, J., Xu, L., Keller, M., Yang, Y., Duffy, P., … Schimel, 
D. (2015). Seeing the forest beyond the trees. Global Ecology and 
Biogeography, 24, 606–610. https ://doi.org/10.1111/geb.12256 
Schippers, P., Sterck, F., Vlam, M., & Zuidema, P. A. (2015). Tree growth 
variation in the tropical forest: Understanding effects of tempera‐
ture, rainfall and CO2. Global Change Biology, 21, 2749–2761. https ://
doi.org/10.1111/gcb.12877 
Slot, M., & Winter, K. (2017). In situ temperature relationships of bio‐
chemical and stomatal controls of photosynthesis in four lowland 
tropical tree species. Plant, Cell and Environment, 40, 3055–3068. 
https ://doi.org/10.1111/pce.13071 
Taylor, P. G., Cleveland, C. C., Wieder, W. R., Sullivan, B. W., Doughty, 
C. E., Dobrowski, S. Z., & Townsend, A. R. (2017). Temperature and 
rainfall interact to control carbon cycling in tropical forests. Ecology 
Letters, 20(6), 779–788. https ://doi.org/10.1111/ele.12765 
Trugman, A. T., Detto, M., Bartlett, M. K., Medvigy, D., Anderegg, W. R. 
L., Schwalm, C., … Pacala, S. W. (2018). Tree carbon allocation ex‐
plains forest drought‐kill and recovery patterns. Ecology Letters, 21, 
1552–1560. https ://doi.org/10.1111/ele.13136 
Van Bael, S. A., Aiello, A., Valderrama, A., Medianero, E., Samaniego, M., & 
Wright, S. J. (2004). General herbivore outbreak following an El Nino‐re‐
lated drought in a lowland Panamanian forest. Journal of Tropical Ecology, 
20, 625–633. https ://doi.org/10.1017/s0266 46740 4001725
van der Sande, M. T., Arets, E. J. M. M., Peña‐Claros, M., de Avila, 
A. L., Roopsind, A., Mazzei, L., … Poorter, L. (2016). Old‐growth 
Neotropical forests are shifting in species and trait composition. 
Ecological Monographs, 86, 228–243. https ://doi.org/10.1890/ 
15‐1815.1
Wagner, F. H., Hérault, B., Bonal, D., Stahl, C., Anderson, L. O., Baker, 
T. R., … Aragão, L. E. O. C. (2016). Climate seasonality limits 
leaf carbon assimilation and wood productivity in tropical for‐
ests. Biogeosciences, 13, 2537–2562. https ://doi.org/10.5194/
bg‐13‐2537‐2016
Wagner, F., Rutishauser, E., Blanc, L., & Herault, B. (2010). Assessing ef‐
fects of plot size and census interval on estimates of tropical for‐
est structure and dynamics. Biotropica, 42, 664–671. https ://doi.
org/10.1111/j.1744‐7429.2010.00644.x
Watt, A. S. (1947). Pattern and process in plant ecology. Journal of 
Ecology, 35, 1–22.
Whitmore, T. C. (1989). Canopy gaps and the two major groups of forest 
trees. Ecology, 70, 536–538. https ://doi.org/10.2307/1940195
Wright, S. J. (2006). Response to Lewis et al.: The uncertain response of 
tropical forests to global change. Trends in Ecology & Evolution, 21(4), 
174–175. https ://doi.org/10.1016/j.tree.2006.02.003
Wright, S. J. (2010). The future of tropical forests. Annals of the New 
York Academy of Sciences, 1195, 1–27. https ://doi.org/10.1111/j.1749‐ 
6632..2010.05455.x
Wright, S. J. (2013). The carbon sink in intact tropical forests. 
Global Change Biology, 19(2), 337–339. https ://doi.org/10.1111/
gcb.12052 
Wright, S. J., & Calderón, O. (2006). Seasonal, El Nino and longer term 
changes in flower and seed production in a moist tropical forest. 
Ecology Letters, 9, 35–44. https ://doi.org/10.1111/j.1461‐0248.2005. 
00851.x
Wright, S. J., Kitajima, K., Kraft, N. J. B., Reich, P. B., Wright, I. J., Bunker, 
D. E., … Zanne, A. E. (2010). Functional traits and the growth–mor‐
tality trade‐off in tropical trees. Ecology, 91, 3664–3674. https ://doi.
org/10.1890/09‐2335.1
Wright, S. J., Yavitt, J. B., Wurzburger, N., Turner, B. L., Tanner, E. V. 
J., Sayer, E. J., … Corre, M. D. (2011). Potassium, phosphorus, or 
nitrogen limit root allocation, tree growth, or litter production 
in a lowland tropical forest. Ecology, 92, 1616–1625. https ://doi.
org/10.1890/10‐1558.1
Yanoviak, S. P., Gora, E. M., Burchfield, J. M., Bitzer, P. M., & Detto, M. 
(2017). Quantification and identification of lightning damage in 
tropical forests. Ecology and Evolution, 7, 5111–5122. https ://doi.
org/10.1002/ece3.3095
SUPPORTING INFORMATION
Additional supporting information may be found online in the 
Supporting Information section at the end of the article.  
How to cite this article: Rutishauser E, Wright SJ, Condit R, 
Hubbell SP, Davies SJ, Muller‐Landau HC. Testing for 
changes in biomass dynamics in large‐scale forest datasets. 
Glob Change Biol. 2019;00:1–15. https ://doi.org/10.1111/
gcb.14833