Biogeochemistry 52: 115?131, 2001.
? 2001 Kluwer Academic Publishers. Printed in the Netherlands.
Respiration from coarse wood litter in central Amazon
forests
JEFFREY Q. CHAMBERS1;2;3, JOSHUA P. SCHIMEL1 &
ANTONIO D. NOBRE2
1Department of Ecology, Evolution and Marine Biology, University of California, Santa
Barbara CA 93106, U.S.A.; 2present address: Instituto Nacional de Pesquisas da Amaz?nia
(INPA) C.P. 478, Manaus, Amazonas, Brazil 69011-970 (e-mail: jeff@inpa.gov.br; fax:
55-92- 642-4068); 3Biological Dynamics of Forest Fragments Project INPA, C.P. 478
Manaus, Amazonas, Brazil 69011-970
Key words: carbon cycling, heterotrophic respiration, net ecosystem exchange, predictive
model, tropical forest, wood decomposition
Abstract. Respiration from coarse litter (trunks and large branches > 10 cm diameter)
was studied in central Amazon forests. Respiration rates varied over almost two orders of
magnitude (1.003?0.014 g C g?1 C min?1, n = 61), and were significantly correlated
with wood density (r2adj = 0.42), and moisture content (r2adj = 0.39). Additional samples
taken from a nearby pasture indicated that wood moisture content was the most important
factor controlling respiration rates across sites (r2adj = 0.65). Based on average coarse litter
wood density and moisture content, the mean long-term carbon loss rate due to respiration
was estimated to be 0.13 yr?1 (range of 95% prediction interval (PI) = 0.11?0.15 yr?1).
Comparing mean respiration rate with mean mass loss (decomposition) rate from a previous
study, respiratory emissions to the atmosphere from coarse litter were predicted to be 76%
(95% PI = 65?88%) of total carbon loss, or about 1.9 (95% PI = 1.6?2.2) Mg C ha?1 yr?1.
Optimum respiration activity corresponded to about 2.5 g H2O g?1 dry wood, and severely
restricted respiration to < 0.5 g H2O g?1 dry wood. Respiration from coarse litter in central
Amazon forests is comparable in magnitude to decomposing fine surface litter (e.g. leaves,
twigs) and is an important carbon cycling component when characterizing heterotrophic
respiration budgets and net ecosystem exchange (NEE).
Introduction
Net ecosystem exchange (with the atmosphere) (NEE) is the difference
between gross primary production (GPP), and total respiration (Rt = hetero-
trophic + autotrophic). Because NEE is the difference between two relatively
large fluxes, NEE is typically an order of magnitude less that either GPP or
Rt. Eddy covariance measurements have indicated that Amazon forests are
acting as a net carbon sink (negative NEE) on the order of 1.0?5.9 Mg C
116
ha?1 yr?1 (Mahli et al. 1998; Grace et al. 1995; Fan et al. 1990). This meas-
ured sink has been corroborated by forest inventory plot data (Phillips et al.
1998). However, given the many potential sources of error in these large-scale
measurements (Mahli et al. 1998; Goulden et al. 1996; Keller et al. 1996), a
95% confidence interval about the magnitude of the sink strength for Amazon
forests would probably not preclude carbon balance. Independent ground
measurements of all significant carbon fluxes (e.g. > 0.5 Mg C ha?1 yr?1)
are important to help evaluate eddy covariance results, and for understanding
the factors that control flux variability.
Respiration exerts a strong control over short- and long-term variation in
NEE (Goulden 1997). Thus, to more thoroughly evaluate the hypothesis that
tropical forests are acting as net carbon sinks, detailed studies for separate
respiratory fluxes are required (e.g. soil organic matter, surface litter), along
with factors that control flux variability. There are respiration data for fine
surface litter, soil, and live wood from tropical regions (e.g., Sampaio et al.
1993; Ryan et al. 1994; Trumbore et al. 1995; Davidson & Trumbore 1995),
but investigations of respiration from coarse litter (dead trunks and branches
> 10 cm diameter) are lacking. This is a notable deficiency because wood
decay is the primary source of carbon flux to the atmosphere from land-use
in the Brazilian Amazon, and this flux may largely offset the proposed carbon
sink from undisturbed forests (Houghton et al. 2000).
Because eddy covariance results cannot offer insight into the future beha-
vior of carbon balance, data that characterize how individual carbon pools
respond to environmental changes are needed to develop predictive models
(Amthor 1989). Without data describing controls over carbon pools and
fluxes, however, models are often based on paramaterizations that may or
may not be accurate. In the CASA model, for example, all surface litter
losses are as CO2 respired to the atmosphere (Potter 1993), whereas in the
Century model, a large fraction of wood carbon is incorporated into soil
organic matter (Parton 1987, 1988). This study provides important coarse
litter data for tropical forests.
Most litter studies focus on leaves, twigs, fruits and other small organic
matter components (fine litter). Tree mortality and damage results in the
production of coarse litter whose role in carbon cycling is often overlooked,
especially in the tropics (Harmon et al. 1986). Central Amazon forests have
coarse litter inputs that are at least 30% of total surface litter production, with
a average decomposition rate of about 0.17 yr?1 (Chambers et al. 2000). Also,
based on forest inventory data, Chambers (1998) demonstrated that year-to-
year variability in tree mortality results in coarse litter stocks with a strong
patchy distribution. Thus, decomposition of coarse litter probably results in
117
a substantial respiratory flux to the atmosphere with considerable spatial and
temporal variability.
Decomposition can be defined as respiration, fragmentation, and leaching
of organic matter which comprise all the processes that result in mass loss
(Swift et al. 1979). Respiration from surface litter results in CO2 emissions
to the atmosphere, whereas the other two processes result in organic matter
inputs to soil or streams. Most wood decomposition studies quantify rates
as either mass loss or density change per unit time, and decomposition is
treated as an aggregate process. To characterize ecosystem carbon balance,
losses must be partitioned into constituent components (e.g. atmosphere,
soil, streams). Studies isolating respiratory losses from coarse litter are few,
and are limited to temperate old-growth (Carpenter et al. 1988; Marra &
Edmonds 1994) and clear-cut coniferous forest (Marra & Edmonds 1996),
and a temperate evergreen oak forest (Yoneda 1985). No coarse litter respir-
ation studies have been published for tropical forests. Also, previous studies
used static chamber methods which have been shown to overestimate low flux
rates, severely underestimate high flux rates, and are generally less accurate
than dynamic chamber methods (Lund et al. 1999).
This study was carried out in central Amazon forests. The objectives were
to: (i) develop a methodology for using a dynamic chamber to quantify coarse
litter respiration rates; (ii) measure rates from a representative sample of dead
trees; (iii) investigate relationships between respiration rate, wood moisture
content, and wood density; (iv) estimate the fraction of decomposition as
respiratory loss to the atmosphere; and (v) compare the magnitude of coarse
litter respiration rates with other heterotrophic sources of CO2.
Materials and methods
Sites
Work was carried out on permanent plots established by the Biological
Dynamics of Forest Fragments Project (BDFFP) (Lovejoy & Bierregaard
1990), a joint project between Brazil?s National Institute for Amazon
Research (Instituto Nacional de Pesquisas da Amaz?nia ? INPA) and
the Smithsonian Institution, and INPA?s Biomass and Nutrient Experiment
(BIONTE) (Higuchi et al. 1997). Data from permanent plots included dates
when trees died, their diameter at breast height (DBH, at 1.3 m), and
taxonomic information. Plots are located within a 1,200 km2 area (2
300S, 60W) approximately 60 km north of Manaus. Vegetation is closed-
canopy old-growth forest, with some of the largest trees living over 1000
years (Chambers et al. 1998). Mean annual rainfall is about 2200 mm and
118
mean annual temperature is 26.7 C (National Climatic Data Center). There
is a distinct dry season during July, August, and September with usually
< 100 mm of rain per month. The terrain is undulating, with soils comprising
Oxisols on plateaus, Utilsols on slopes, and Spodosols associated with small
streams in valley bottoms (Bravard & Righi 1989). Surface (to 5 cm) clay
content decreases from about 75% to 5%, and sand content increases from
about 10% to 85%, when moving from plateau to valley (Ferraz et al. 1998).
Nearly 1,200 tree species have been identified in nearby forests (Ribeiro et al.
1999), most with densities of less than one individual per hectare (Rankin-De
Merona et al. 1992).
Permanent forest inventory plots were established in the central Amazon
during the 1980s to study the effects of logging practices (BIONTE, 3 1-
ha plots) and fragmentation (BDFFP, 18 1-ha plots) on forest structure and
functioning. All trees larger than 10 cm DBH were tagged, measured, and,
when possible, identified, and during subsequent inventories, recruitment,
growth of surviving trees, and mortality were documented. Mortality records
from control plots in the primary forest were used to select dead trees for
sampling. Sampled trees had been dead for 3 to 12 years, and in most cases,
census intervals allowed establishing the time of death to within  1.5 years.
Field methods
Boles from 155 dead trees were previously sampled for a coarse litter decom-
position study (Chambers et al. 2000). Briefly, these 155 samples were
randomly chosen from a larger population of 880 dead trees using mortality
data from the BDFFP and BIONTE plots. This population was stratified by
DBH, wood density, and time since death, to ensure adequate representation
of large trees, low and high wood densities, and decomposition time intervals.
Cross-sections (usually 3) were removed from the boles with a chainsaw, or
a machete for boles in advanced stages of decomposition. Decomposition
(mass loss) rate was estimated using the original mass, the mass at the time
of sampling, and the time decomposing, assuming first order decay. Rates
were inversely correlated with wood density and bole diameter.
For this study, two additional cross-sections were removed from 61 of
the previously sampled boles at least 1 m from previous cuts. A wedge was
removed from the cross-sections using a machete. The sample was placed
in a plastic container (3.8 L, Rubbermaid) and the seal was coated with a
small amount of high-vacuum grease (Dow Corning) to minimize leaking.
The change in chamber headspace CO2 concentration was measured as ppm
CO2 s?1 (or mmol CO2 mol?1 air s?1,1CO2) using an infra-red gas analyzer
(IRGA, LiCor 6200). The chamber was modified to create an air-tight seal
with the IRGA sensor head. Measurement of 1CO2 began once the chamber
119
Figure 1. CO2 emission rates from coarse litter samples removed from boles and immedi-
ately placed in the measurement chamber. Respiration rates initially declined rapidly, and
then leveled off, as high CO2 concentrations in the wood pore matrix equilibrated with the
atmosphere. After approximately 3 hours, there was no significant decline in CO2 emissions
rate over time, suggesting partial pressure equilibrium with the atmosphere.
humidity stabilized (about 1 minute), and the measurement interval spanned
1?2 minutes. Since coarse litter is not actively transpiring, a correction (flow
through the desiccant) for a rise in chamber humidity during the measure-
ment interval was not required. The leak rate of the chamber was determined
by creating a CO2 partial pressure gradient using a soda lime scrubber, and
was orders of magnitude lower than typical respiration rates. The volume of
the wood sample was estimated (water displacement) to correct the chamber
volume, and to determine wood density.
To explore how carbon emissions vary with large changes in environ-
mental conditions, wood samples were also cut from 10 boles located in
an approximately 15 year old pasture (one cross-section per log). Respira-
tion rates, wood density and moisture for these samples were calculated as
described above.
Atmospheric equilibration
Initially, samples were placed in the chamber immediately after cutting.
Rates of 1CO2 were very high and declined rapidly (Figure 1). We assumed
that this sharp decline was caused by CO2 partial pressure equilibration
between high CO2 concentrations in the wood pore matrix and the atmo-
120
sphere. Thacker and Good (1952), For example, found that CO2 concentration
was typically > 10% (vol.) in boles of decomposing sugar maple. To avoid
the steepest portion of the decline curve when measuring the initial rate
(Figure 1), we allowed samples to equilibrate for 5?10 minutes before placing
them in the chamber.
To estimate the average time required for pore matrix CO2 to equilibrate
with the atmosphere, 1CO2 rates for 41 samples, representing a wide range
of moisture contents and wood densities, were measured a minimum of 4
times after the sample was removed from the log. Since the initial absolute
rates varied from 3.59?0.0004 ppm s?1, relative rates were calculated as the
fraction of the initial rate for each sample. A regression of relative rates vs.
time demonstrated that after about 3 hours, there was no significant decline in
1CO2 over time. Based on this analysis, the respiration rate for each wedge
was estimated by averaging all 1CO2 measurements taken 180 minutes
after the sample was removed. At this point, the respiration rate averaged
65% of the initial rate. For some samples in the forest (n = 10), it was only
possible to measure the initial rate, and in these cases, respiration rates were
estimated as 0.65 of the initial rate.
Calculations
A mass based respiration rate (g C g?1 C min?1) was calculated using the
rate of CO2 accumulation in the chamber, and the chamber volume. Mass
was determined by drying the sample at 70 C to a constant weight. Carbon
content was measured for samples from 70 boles from the decomposition
study (Chambers et al. 2000) with a Fisons C/N auto-analyzer, and averaged
49%. Moisture content was expressed gravimetrically (g H2O g?1 dry wood).
Wood density was calculated from oven dry weight and water displacement
volume. To compare respiration rates with long-term decomposition rates
(Chambers et al. 2000), 1C was also calculated as an annual carbon loss
rate assuming that a constant fraction of material is lost per unit time (Olson
1963) from kc D ? ln..mc ? 1C/=mc/ (Eqn. 1), where 1C is in units of g
C yr?1, mc is the mass of carbon in the wood, and kc is the carbon loss rate
(fraction yr?1).
Statistical analyses
Statistics were performed using SAS (v. 6.12). Relationships between wood
moisture content, wood density, and respiration rates were investigated
using multiple linear regression. The x and y variables were transformed
to calculate best-fit regressions with residuals exhibiting no curvilinearity
(homoscedasticity). The site effect (forest vs. pasture) was examined using
121
Table 1. Comparison of wood density and moisture content for coarse litter from two
separate studies at the same sites as described in the text. Using single-factor ANOVAs,
no significant differences were detected between the two studies. For the test, to increase
normality, moisture values were log10 transformed. Combined average wood moisture
and density were 0.96 (1.05?0.87 = 95% C.I., n = 149) g H2O g?1 dry wood, and 0.51
(0.48?0.54 = 95% C.I., n = 105) g cm?3. Combined moisture content was higher than
either study separately because some boles were sampled twice, and these were averaged
Variable Study n Average 95% C.I. min max P-value
moisture respiration 61 0.89 0.80?1.00 0.245 2.291 0.659
decomposition 137 0.93 0.84?1.02 0.266 5.000
density respiration 61 0.53 0.50?0.56 0.272 1.054 0.297
decomposition 64 0.49 0.47?0.52 0.196 0.993
site as an indicator variable (Neter et al. 1996, p. 455). Transformations to
increase the probability of a normal distribution were performed on variables
that exhibited skewed distributions when performing ANOVAs.
Large-scale fluxes
To assess whether coarse litter respiration relationships were applicable over
larger temporal and spatial scales, we had to determine if the range in
wood density and moisture content were representative of larger spatial and
temporal scales. This study took place during the transition from wet to
dry season of 1997 (June?August), whereas sampling from the decomposi-
tion study (Chambers et al. 2000) was carried out during both the dry and
wet seasons of 1996?97, and covered a larger area. Single-factor ANOVAs
showed no significant differences in wood moisture content (log-transformed)
or wood density between the two studies (Table 1). Despite being carried out
in the dry season, high rainfall often coincided with sampling days during
this study, and may explain why the moisture regime of coarse litter was not
significantly drier than for sampling that took place during both wet and dry
seasons.
Average hectare-scale fluxes were predicted using a mean coarse litter
respiration rate, and standing stocks. Average wood density and moisture
(Table 1) were used to predict an average coarse litter respiration rate based
on regression relationships. Error was expressed as a 95% prediction interval
(PI) for the mean response from regression relationships. Because there is
also error associated with estimates for average coarse litter moisture and
122
wood density, a 95% PI was also estimated for error in these predictor
variables.
Results
Coarse litter respiration rates for the forest varied by almost two orders of
magnitude (1.003?0.014 g C g?1 C min?1, n = 61), and varied to a lesser
extent in the pasture (0.213?0.004 g C g?1 C min?1, n = 10). Respira-
tion rates exhibited a strongly skewed distribution, and log- transformations
significantly increased the probability of a normal distribution. Averages
from log-transformed data were 0.192 g C g?1 C min?1 for the forest
and 0.023 g C g?1 C min?1 for the pasture. Respiration rates were almost
an order of magnitude lower in the pasture, and this difference was highly
significant (ANOVA, p < 0.0001).
Comparing relationships across sites, respiration rates were highly correl-
ated with both wood density (Figure 2(b)) and wood moisture content
(Figure 2(c)). There was also a strong relationship between wood density
and moisture content (Figure 2(a)). The relationship between respiration rate
and wood density, and between moisture content and wood density, showed
a significant site effect. There was no site effect for the relationship between
respiration rate and moisture content. Analysis of the forest data alone indic-
ated outliers (n = 2) when regressing respiration rates vs. wood density. The
outliers were two snags (standing dead trees) that had much lower moisture,
and respiration rates, than for downed boles with similar wood density. The
mean and range in annual respiration rate (Eqn. 1) was 0.110 (0.887?0.007)
yr?1 for the forest, and 0.012 (0.119?0.002) yr?1 for the pasture.
95% prediction intervals (PIs) were calculated using both error in the
predictor variable (i.e. wood density and moisture content), and error in
the coefficients from the regression equations (Figure 2, Eqns. (4) & (5)).
Using the average coarse litter moisture and 95% CI (Table 1), the average
respiration rate (Figure 2, Eqn. (5)) and 95% PI was 0.12 (0.11?0.14) yr?1.
Using average coarse litter wood density and 95% CI (Table 1), the average
respiration rate (Figure 2, Eqn. (4)) and 95% PI was also 0.12 (0.11?0.14)
yr?1. For comparison, Summers (1998) measured average coarse litter wood
density of 0.46  0.03 (95% CI) g cm?3, which gave an average respiration
rate (Figure 2, Eqn. (4)) and 95% PI of 0.15 (0.13?0.18) yr?1. Analysis
of error associated with the predicted coefficients of the regression equa-
tions gave similar results. Average respiration rates and 95% PIs for average
wood density, moisture content, and Summers (1998) average wood density
were 0.12 (0.11?0.14) yr?1, 0.12 (0.11?0.14) yr?1, and 0.15 (0.13?0.17)
123
Figure 2. Relationships between wood moisture (M, g H2O g?1 dry wood), wood density,
(, g dry weight cm?3), and respiration rate (kr , g C g?1 wood C min?1). The curves are
best-fit regressions with residuals exhibiting homoscedasticity (SAS v. 6.12) giving statistic-
ally unbiased estimates. Open circles represent wood sampled in pastures and solid circles in
forest. (a) M and  were highly correlated with a significant site effect (multiple regression
and ANOVA,
p
M = 1.14 + 0.268[site = forest] ? 0.841, p < 0.0001, r2adj = 0.67)(1). (b)
In predicting respiration rates, there was a significant site effect for  (Log[kr ] = ?0.379
+ 0.562[site = forest] ? 1.71, p < 0.0001, r2adj = 0.62) (2), with pasture logs respiring at
lower rates. (c) However, with respect to M, there was no site effect, and moisture alone was
a good predictor of respiration rates for both pasture and forest sites (Log[kr ] = ?0.650 +
1.592Log[M], p < 0.0001, r2adj = 0.65)(3). Predicting annual carbon loss rates (kc, see text),
regression analysis of the forest data alone gave Log[kc] = ?1.788, (r2adj = 0.42, y-intercept
not significant, p = 0.91)(4), for density, and Log[kc] = ?0.889 + 1.42Log[M],(r2adj = 0.39)(5),
for moisture. An annual loss rate for both forest and pasture was given by Log[kc] = ?0.886 +
1.66Log[M],(r2adj = 0.64) (6).
124
yr?1, respectively. Thus, error analysis showed similar PIs (error bars) when
calculated for both the x (predictor) and y (response) variables.
We estimated an average coarse litter respiration rate and PI of 0.13
(0.11?0.15) yr?1 by averaging error in the wood density and moisture error
estimates. Chambers et al. (2000) estimated an average decomposition (mass
loss) rate of 0.17 yr?1 for the same sites studied here. Thus, we estimate that
76% (95% PI = 65?88%) of carbon mass loss from coarse litter decompos-
ition is respiratory. Mean standing-stocks of coarse litter in 3 ha of central
Amazon forests ranged from 5.7?26.3 Mg C ha?1, and averaged 14.9 Mg
C ha?1 (Summers 1998). Thus, as a first approximation, using the average
coarse litter respiration rate and 95% PI estimated here, the mean flux of
carbon to the atmosphere is estimated to be 1.9 (1.6?2.2) Mg C ha?1 yr?1.
Discussion
Methodological considerations
The time required for a sample removed from a decomposing bole to reach
CO2 atmospheric equilibrium was about three hours. An important question is
whether the rate measured after three hours is indicative of the actual respir-
ation rate when the sample was located within the intact bole. Two factors
that can influence the activity of fungi are CO2 and O2 concentrations. The
interior of decaying boles are characterized by high CO2 and low O2 pressures
(Thacker and Good 1952) that may restrict the growth of fungi. Removing a
sample from the bole may enhance respiration rates by reducing CO2, and
increasing O2, concentrations.
A number of studies have shown, however, that fungal growth is not
inhibited by CO2 concentrations as high as 15% (vol.) (Thacker & Good
1952), and is not markedly reduced until O2 concentrations fall below about
1.5% (vol.) (Scheffer 1986). The highest CO2 concentrations found in the
heartwood of dead sugar maple trees was about 17%, and O2 concentrations
ranged from 1?14% (Thacker and Good 1952). O2 concentration above 5%
(vol.) have little effect on increasing fungal growth rates (Scheffer 1986).
With respect to high CO2 concentrations, growth of some wood-decay fungi
can be slightly stimulated above 10% (vol.) CO2, and most show no metabolic
decline with high CO2 (Dix and Webster 1995). This suggests that changes in
CO2 and O2 concentrations caused by removing samples from decaying boles
will not significantly alter fungal metabolism rates. However, these studies
were carried out in environments very different from tropical forests, and
additional studies are needed to determine the sensitivity of fungal activity to
the sampling methodology used here.
125
Another important consideration is the timing of CO2 equilibration with
the atmosphere. When a wood sample is removed from a decaying bole,
the surface area exposed to the atmosphere increases, and a large amount
of trapped CO2 can escape. As this high concentration equilibrates with the
atmosphere, there is a large pulse of CO2. Regression analysis indicated that
about three hours were required until there was no significant decline in
CO2 emission rate. We assumed that this stable rate was respiratory activity.
Although not directly comparable, Sampaio et al. (1993) found that about
90% of fine litter losses in a humid tropical forest were respiratory. In combin-
ation with our results, it appears that a large fraction of litter decomposition
mass loss is respiratory.
Controls over respiration rates
This study showed that moisture alone accounted for a large part of the
variation in coarse litter respiration rates (Figure 2(c)). This relationship
appeared to hold for coarse litter decomposing in both the forest and pasture.
Wood density was also highly correlated with respiration rates, although this
relationship appears to have been ultimately controlled by the relationship
between wood moisture and wood density (Figure 2(a), 2(b)). Wood density
changes slowly over time, whereas wood moisture can rapidly respond to
temporal changes in precipitation. Low density wood that experiences dry
conditions, for example, will exhibit relatively low respiration rates, but when
wet-up by precipitation, will probably experience a rapid increase in rates
(Figure 2(c)).
Fungi do not respond to moisture content but to water potential, and the
water potential of decayed wood is higher than intact wood at the same mois-
ture content (Dix and Webster 1995). As wood decays, cell wall polymers are
hydrolized, and the matrix potential at a given water content rises, increasing
the availability of moisture to microbial communities (Dix 1985). At very
low moisture contents, the water potential for decayed wood can be > 5 times
higher than intact wood at the same moisture content (Dix 1985). Fungi can
also actively regulate wood moisture content, creating a more favorable decay
environment (Rayner and Boddy, 1988). Fungi responsible for wood decay
respond rapidly to changes in water potential, and growth is severely limited
below ?4.0 Mpa (Boddy 1983a; Dix 1984).
Optimal water potential values are at least ?1.0 Mpa, and probably
higher in many cases (Boddy 1983a). Averaging the regression relationships
between moisture content (g H2O g?1 dry wood) and water potential Dix
(1985) predicts that decayed wood moisture contents of 2.5 and 0.5 g H2O g?1
dry wood should correspond to optimal (?1.0 Mpa) and severely limited (?4.0
MPa) fungal growth rates, respectively. These limits coincide well with the
126
relationship found between wood moisture and respiration rates in this study
(Figure 2(c)). Wood-decay fungi are affected by relatively small changes
in moisture content (Dix and Webster 1995, and references therein), and a
strong relationship between moisture content and respiration rate is expected.
Due to the relationships between wood density, moisture content, and respir-
ation rates found in this study (Figure 2), there appears to be a non-linear
response, with respiration rates continuously increasing as decomposition
proceeds, wood density declines, and water potential increases. Moisture
contents above 2 g H2O g?1, for example, were associated with low wood
densities and high respiration rates (Figure 2). From these relationships, a
simple model can be developed to explore the time course of respiratory
carbon losses from coarse litter.
If we assume that fungal hyphae colonize a log from the periphery and
move inward toward the core at a constant rate (kh), then the volumetric
fraction of the log that is colonized is given by: fc D 2kht=r ? k2ht2=r2
(Eqn. 2), where t is the time since initial colonization and r is the radius
of the log. Assuming a linear relationship between kh and wood density ()
(high density wood is colonized more slowly), and employing the regression
equations for moisture content vs. wood density, respiration rate vs. moisture
content, we can simulate how respiration rates would vary with time.
The resulting time-series predict that the course of decomposition should
differ between density classes. High density wood exhibits a relatively long
colonization phase (Figure 3(d)), whereas low density wood follows first-
order decay (single-exponential) more closely (Figure 3(a)). In support of
this model, Harmon et al. (1995) found that decomposing wood from four
tropical trees species exhibited time-series curves similar to Figure 3, with
the high density species exhibiting a longer colonization phase (M. Harmon,
personal communication). Most studies assume first-order decay and use
a single-exponential model when calculating organic matter decomposition
rates (Wieder and Lang 1982). Deviations from the single-exponential model
are important for determining the timing of respiratory losses. However, when
estimating long-term rates, a single-exponential model makes reasonable
predictions (i.e., the curves in Figure 3 converge over time).
The relationships described above concern downed coarse litter in the
forest interior. These relationships may change in contrasting environments.
High mortality on the edges of forest fragments, for example, result in
dramatically increased coarse litter production rates (Laurance et al. 1997).
Also, in pastures, coarse litter moisture contents and respiration rates were
significantly lower for the same wood density values (Figure 2(a)). Snags
(standing dead trees) also exhibited lower moisture contents (0.245 and
0.270 g H20 g?1 dry wood, n = 2) and respiration rates (0.008 and 0.036 yr?1),
127
Figure 3. Simulated timing of mass loss due to respiration for increasing values of initial
wood density (, g cm?3). The qualitative model used to generate these curves assumes that
the rate of fungal penetration (kh, cm yr?1) is an inverse linear function of wood density,
with higher density wood being colonized more slowly ( and kh varying from 0.90?0.30
and 2.0?8.0, a?d, respectively). High density wood is predicted to experience a lag-time in
respiration corresponding to a colonization phase. Non-linear mass loss curves (solid line) are
compared with those predicted using a single-exponential model (dotted line).
and were identified as outliers in the regression analysis of the forest data.
Coarse litter that is exposed to increased solar radiation will be drier than
coarse litter in the forest interior, and will also experience larger fluctu-
ations in moisture content (Boddy 1983b; Dix and Webster 1995). This is
an important area for further research because coarse litter in disturbed areas
has been identified as the main contributor of the land-use flux of carbon to
the atmosphere for the Brazilian Amazon (Houghton et al. 2000).
Large-scale carbon fluxes
The loss of carbon from decomposing wood must either enter the atmosphere
as CO2 or be redistributed within the ecosystem (e.g. soils, streams). In this
study we estimated that about 76% of carbon is lost to the atmosphere (1.9
Mg C ha?1 yr?1), and 24% is redistributed. In comparison, fluxes to the
atmosphere from soil respiration (surface fine litter and soil, including root
respiration) in forests in the vicinity of the BIONTE plots were 26.0  6.8
(95% C.I.) Mg C ha?1 yr?1 (Sotta 1998). For a drier evergreen forest in the
eastern Amazon, soil surface fluxes were estimated to be 1.2 Mg C ha?1 yr?1
128
for fine surface litter, and 6.2 Mg C ha?1 yr?1 for soil organic matter to 5 m
depth (Trumbore et al. 1995). Total soil carbon flux in the eastern Amazon,
including root respiration, was estimated to be 23.6 Mg C ha?1 yr?1. If soil
carbon fluxes in the central Amazon are partitioned similarly to forests in
the eastern Amazon, the magnitude of the carbon flux from coarse litter is
comparable to that from fine surface litter.
When calculating NEE, net primary production (NPP) is balanced against
total heterotrophic respiration (Rh). Over large spatial scales, NEE fluxes as
small as 0.5 Mg C ha?1 yr?1, can have important implications for the global
carbon cycle (Law et al. 1999; Goulden et al. 1996). A number of studies
have indicated that tropical forests are acting as net carbon sinks (Malhi
et al. 1998; Phillips et al. 1998; Grace et al. 1995; Fan et al. 1990). When
simply scaled over the entire Amazon basin (5.0  108 ha) these proposed
sinks (0.5?3.0 Pg yr?1) represent a sizable portion of emissions from fossil
fuels (5.7 Pg yr?1) and deforestation (2 Pg yr?1) (Houghton 1991). Thus,
independent measures of source pools fluxes are important to help assess
the accuracy of eddy covariance results. Also, understanding the biological
and physical controls over the source and magnitude of respiratory fluxes
are important for developing predictive models. We have demonstrated that
respiration from coarse litter is a significant carbon flux, and that wood mois-
ture content explains a large portion of flux variability both in forests and
pastures.
Acknowledgements
We thank Sue Trumbore, Allan Stewart-Oaten, and three anonymous
reviewers for comments on previous versions of this manuscript. We are
grateful for the logistics and technical support of the BIONTE and BDFFP
projects. Work was supported in part by Lawrence Livermore National
Laboratory?s Laboratory Directed Research and Development program (95-
DI-005) under the auspices of the U.S. Department of Energy subordinate
to Contract No. W-7405-Eng-48 with the University of California, and by
NASA?s terrestrial ecology program (LBA-ecology).
References
Amthor JS (1989) Respiration and Crop Productivity. Springer-Verlag, New York
Boddy L (1983a) Effect of temperature and water potential on growth rate of wood-rotting
Basidiomycetes. Trans. Brit. Mycol. Soc. 80: 141?149
Boddy L (1983b) Microclimate and moisture dynamics of wood decomposing in terrestrial
ecosystems. Soil Biol. Biochem. 15: 149?157
129
Bravard S & Righi D (1989) Geochemical differences in an Oxisol-Spodosol toposequence of
Amazonia, Brazil. Geoderma 44: 29?42
Carptenter SE, Harmon ME, Ingham ER, Kelsey RG, Lattin JD & Schowalter TD (1988)
Early patterns of heterotrophic activity in conifer logs. Proceedings of the Royal Society
of Edinburgh 94B: 33?43
Chambers JQ, N Higuchi & JP Schimel (1998) Ancient trees in Amazonia. Nature 391: 135?
136
Chambers JQ (1998) The role of large wood in the carbon cycle of central Amazon rain forest.
Ph.D. thesis, University of California, Santa Barbara, USA
Chambers JQ, N Higuchi, LV Ferreira, JM Melack & JP Schimel (2000) Decomposition
and carbon cycling of dead trees in tropical evergreen forests of the central Amazon.
Oecologia, in press
Davidson EA & SE Trumbore (1995) Gas diffusivity and the production of CO2 in deep soils
of the eastern Amazon. Tellus 47B: 550?565
Dix NJ (1984) Minimum water potentials for growth of some litter-decomposing agarics and
other basidiomycetes. Trans. Brit. Mycol. Soc. 83: 152?153
Dix NJ (1985) Changes in relationship between water content and water potential after decay
and its significance for fungal successions. Trans. Brit. Mycol. Soc. 85: 649?653
Dix NJ & J Webster (1995) Fungal Ecology. Chapman & Hall, London
Fan S-C, SC Wofsy, PS Bakwin & DJ Jacob (1990) Atmosphere-Biosphere exchange of CO2
and O3 in the Central Amazon forest. Journal of Geophysical Research 95: 16851?16864
Ferraz J, S Ohta & PC Sales (1998) Distribu??o dos solos ao longo de dois transectos em
floresta prim?ria ao norte de Manaus Pesquisas Florestais para a Conserva??o da Floresta
e Reabilita??o de ?reas Degradadas da Amaz?nia (pp 109?144). MCT-INPA/JICA,
Manaus, Brasil
Goulden ML, BC Daube, S-M Fan, DJ Sutton, A Bazzaz, JW Munger & SC Wofsy (1997)
Gross CO2 uptake by a black spruce forest. Journal of Geophysical Research 102: 28,
987?28, 996
Goulden ML, JW Munger, SM Fan, BC Daube & SC Wofsy (1996) Measurements of carbon
sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy.
Glob. Change Biol. 2: 169?182
Grace J, J Lloyd, J McIntyre, AC Miranda, P Meir, HS Miranda, C Nobre, J Moncrieff, J
Massheder, Y Malhi, I Wright & J Gash (1995) Carbon dioxide uptake by an undisturbed
tropical rain forest in Southwest Amazonia, 1992?1993. Science 270: 778?780
Harmon ME, JF Franklin, FJ Swanson, P Sollins, SV Gregory, JD Lattin, NH Anderson, SP
Cline, NG Aumen, JR Sedell, GW Lienkaemper, K Cromack & KW Cummins (1986)
Ecology of coarse woody debris in temperate ecosystems. In: Macfadyen A & Ford ED
(eds) Advances in Ecological Research (pp 133?302). Academic Press, London
Harmon ME, DF Whigham, J Sexton & I Olmsted (1995) Decomposition and mass of
woody detritus in the dry tropical forests of the northeastern Yucatan Peninsula, Mexico.
Biotropica 27: 305?316
Higuchi N, JBS Ferraz, L Antony, F Luiz?o, R Luiz?o, Y Biot, I Hunter, J Proctor & S
Ross (1997) Biomassa e Nutrients Florestais: Projeto BIONTE Relatorio Final. Instituto
Nacional de Pesquisas da Amaz?nia, Manaus, Brasil
Houghton RA (1991) Tropical deforestation and atmospheric carbon dioxide. Clim. Change
19: 99?118
Houghton RA, DL Skole, CA Nobre, JL Hackler, KT Lawrence & WH Chomentowski (2000)
Annual fluxes of carbon from deforestation and regrowth in the Brazilian Amazon. Nature
403: 301?304
130
Laurance WF, SG Laurance, LV Ferreira, JM Rankin-de Merona, C Gascon & TE Lovejoy
(1997) Biomass collapse in Amazonian forest fragments. Science 278: 1117?1118
Law BE, MG Ryan & PM Anthoni (1999) Seasonal and annual respiration of a ponderosa
pine ecosystem. Glob. Change Biol. 5: 169?182
Lovejoy T & R Bierregaard (1990) Central Amazonian forests and the Minimum Critical Size
of Ecosystem project. In: Gentry A (Ed) Four Neotropical Rainforests (pp 60?71). Yale
University Press, New Haven, Connecticut, USA
Lund CP, WJ Riley, LL Pierce & CB Field (1999) The effects of chamber pressurization on
soil-surface CO2 flux and the implications for NEE measurements under elevated CO2.
Glob. Change Biol. 5: 269?281
Malhi Y, AD Nobre, J Grace, B Kruijt, MGP Pereira, A Culf & S Scott (1998) Carbon dioxide
transfer over a Central Amazonian rain forest. J. of Geophys. Res.
Marra JL & RL Edmonds (1994) Coarse woody debris and forest floor respiration in an old-
growth coniferous forest on the Olympic Peninsula, Washington, USA. Can. J. For. Res.
24: 1811?1817
Marra JL & RL Edmonds (1996) Coarse woody debris and soil respiration in a clearcut on the
Olympic Peninsula, Washington, USA. Can. J. For. Res. 26: 1337?1345
Neter J, MH Kutner, CJ Nachtsheim & W Wasserman (1996) Applied Linear Statistical
Models. Irwin, Chicago
Olson JS (1963) Energy storage and the balance of producers and decomposers in ecological
systems. Ecology 44: 322?331
Parton WJ, DS Schimel, CV Cole & DS Ojima (1987) Analysis of factors controlling soil
organic matter levels in Great Plains Grasslands. Soil Sci. Soc. Am. J. 51
Parton WJ, JWB Stewart & C v Cole (1988) Dynamics of C, N, P and S in grassland soils: a
model. Biogeochem. 5: 109?131
Phillips OL, Y Malhi, N Higuchi, WF Laurance, PV N??ez, RM V?squez, SG Laurance, LV
Ferreira, M Stern, S Brown & J Grace (1998) Changes in the carbon balance of tropical
forests: evidence from long-term plots. Science 282: 439?442
Potter CS, JT Randerson, CB Field, PA Matson, PM Vitousek, HA Mooney & SA Klooster
(1993) Terrestrial ecosystem production: a process model based on global satellite and
surface data. Glob. Biogeochem. Cycl. 7: 811?841
Rankin-De Merona JM, GT Prance, RW Hutchings, MFD Silva, WA Rodrigues & ME Uehling
(1992) Preliminary results of a large-scale tree inventory of upland rain forest in the central
Amazon. Acta Amazon. 22: 493?534
Rayner ADM & L Boddy (1988) Fungal Decomposition of Wood: Its Biology and Ecology.
John Wiley & Sons
Ribeiro JELS, MJG Hopkins, A Vicentini, CA Sothers, MAS Costa, JM Brito, MAD Souza,
LHP Martins, LG Lohmann, PACL Assun??o, EC Pereira, CF Silva, MR Mesquita & LC
Proc?pio (1999) Flora da Reserva Ducke. Manaus, Amazonas, Brasil
Ryan MG, RM Hubbard, DA Clark & RL Sanford, Jr. (1994) Woody-tissue respiration for
Simarouba amara and Minquartia guianensis, two tropical wet forest trees with different
growth habits. Oecologia (Berlin) 100: 213?220
Sampaio EVSB, A Dall?Olio, KS Nunes & EEP Lemos (1993) A model of litterfall, litter
layer losses and mass transfer in a humid tropical forest at Pernambuco, Brazil. Journal of
Tropical Ecology 9: 291?301
Scheffer TC (1986) O2 requirements for growth and survival of wood-decaying and sapwood
staining fungi. Can. J. Bot. 64: 1957?1963
Sotta ED (1998) Fluxo de CO2 entre solo e atmosfera em floresta tropical ?mida da Amaz?nia
central. M.S., Instituto Nacional de Pesquisas da Amaz?nia, Manaus
131
Summers PM (1998) Estoque, Decomposi??o e Nutrients da Liteira Grossa em Floresta
de Terra-Firme, na Amaz?nia Central. MS thesis, Instituto Nacional de Pesquisas da
Amaz?nia, Manaus, Brasil
Thacker DG & HM Good (1952) The composition of air in trunks of sugar maple in relation
to decay. Can. J. Bot. 30: 475?485
Trumbore SE, EA Davidson, P Barbosa de Carmago, DC Nepstad & LA Martinelli (1995)
Belowground cycling of carbon in forests and pastures of Eastern Amazonia. Glob.
Biogeochem. Cycl. 9: 515?528
Wieder RK & GE Lang (1982) A critique of the analytical methods used in examining
decomposition data. Ecology 63: 1636?1642
Yoneda T (1985) Relation of wood diameter to the rates of dry weight loss and CO2 evolution
of wood litter in evergreen oak forests. Jap. J. Ecol. 35: 57?66