Aboveground biomass and structure of rainforests in the southwestern Brazilian Amazon

The biomass of intact tropical forests must be known in order to quantify C pools and emissions arising from biomass burning associated with deforestation, land conversion, or fragmentation. To address this need, the study
quantified the total aboveground biomass (TAGB) and forest structure in 20 intact tropical forest sites in western Brazil. The sites were located in open, dense, and ecotone (to savanna) forest types. The TAGB of open
forest ranged from 288 to 346 Mg ha-1, with a mean of 313 Mg ha-1; dense forest TAGB ranged from 298 to 533 Mg ha-1, with a mean of 377 Mg ha-1; and ecotone forests TAGB ranged from 298 to 422 Mg ha-1, with a
mean of 350 Mg ha-1. Mean TAGB for all 20 sites was 341 Mg ha-1. "live trees" (broad-leaved trees) comprised most of TAGB, averaging 280 Mg ha-1. Mean aboveground biomass of trees 10 cm diameter at breast
height (dbh) differed between open (239 Mg ha-1) and dense forests (307 Mg ha-1). Mean biomass of live "non-tree" components (predominantly palms) for all 20 sites was 22 Mg ha-1. The combined biomass of coarse
wood debris, forest floor (litter/rootmat), and standing dead plants (trees, palms and vines) averaged 38 Mg ha-1 or 12% of the TAGB. Forest structure and biomass distribution were not uniform among sites or forest
types. For example, non-tree components ranged from 41% of the TAGB in one ecotone forest to as low as 7% in a dense forest site. Non-tree components comprised 22% of TAGB. This is noteworthy because the
non-tree components are often omitted from forest biomass/carbon pool estimates.
 

Tropical rainforests are a significant global terrestrial C pool, thus, deforestation/land conversion contributes to rising levels of greenhouse gases in the atmosphere. Information on total aboveground biomass (TAGB) is
scarce for Amazonian forests. Indirect estimates based on commercial volume from forest inventory data (Brown and Lugo, 1992; Fearnside and Fearnside, 1992b), as well as direct field measurements of individual trees
have been used to predict TAGB ( Jordan; Klinge; Russell, 1983 and Higuchi et al., 1994). Estimates for TAGB in the Brazilian Amazon have ranged from 155 to 555 Mg ha-1 (Revilla Cardinas et al., 1982 and Brown).

Differences in estimates of TAGB arise in part from the methods used to measure it, as well as from the heterogeneity of the forests. Early studies involved destructive sampling to develop predictive models for tree
biomass estimations based on combinations of tree diameter at breast height (dbh), specific gravity (sg), and height (h) (Jordan; Klinge; Russell, 1983 and Higuchi et al., 1994). The models for individual tree biomass were
then applied to measurements from trees. The application of destructive and field measurements is limited by the time and cost associated with collecting field data over a large area of tropical forests. To reduce the
dependence on destructive or direct field measurements, commercial volumes derived from forest inventories have been used to estimate total tree biomass at large scales ( Brown; Brown and Brown, 1997). Based on a
compilation of results from nine studies for which direct measurements of biomass were made, as well the indirect estimates of Brown and Lugo (1992), Fearnside (1992b) estimated the average TAGB for the Brazilian
Legal Amazon at 335 Mg ha-1. In another set of studies, Kauffman et al. (1995) and Guild et al. (1998) quantified TAGB for six slashed primary forests. Their biomass estimates ranged from 293 to 436 Mg ha-1, with a
mean of 362 Mg ha-1. Although the tierra firme forest sampled by Jordan and Uhl (1978) was considered to be of low stature and biomass (335 Mg ha-1) for the Amazon, their estimate was almost 100 Mg ha-1 more than
the mean biomass for Amazonia (227 Mg ha-1) that Brown and Lugo (1992) calculated through models based on forest inventories. However, the Brown and Lugo (1992) estimates ignored components of TAGB other
than trees 10 cm dbh. Laurance et al. (1997) calculated TAGB and biomass losses from forest inventories by assuming that all non-tree biomass and trees <10 cm dbh made up 12% of the overstory (trees >10 cm dbh).
Yet, in Amazonian forests studied by Kauffman et al. (1998), there was a significant negative relationship between overstory and understory biomass. Nevertheless, TAGB is often estimated by assuming a constant
proportion between overstory trees and other biomass components ( Fearnside, 1992a).
 
 

Study Sites
 

TAGB was estimated by measuring all organic materials above mineral soil. TAGB was divided into "tree" (broad-leaved trees) and "non-tree" (other components, predominantly palms) components based on structural
and ecological significance and practicality of measurement.

Tree diameter was measured at 1.37 m above the ground (dbh) or immediately above the tree buttress or stilt roots when present. Trees were separated into seven diameter classes based on dbh (<10, 10¯30, 30¯50,
50¯70, 70¯100, 100¯200 and >200 cm dbh). Palms were sampled separately from broadleaf trees. We divided them into three categories (basal palms with no trunks, <10 and 10 cm dbh). Vines and lianas were placed in
two size classes (<10 and 10 cm dbh). Other components included small dicots (plants <1.37 m in height), litter/rootmat (forest floor), standing dead trees and palms, and dead and downed coarse woody debris (CWD).
We divided CWD into two categories: 2.5¯7.6 and 7.6 cm diameter (Kauffman and Kauffman). The forest floor component was composed of litter, small wood debris (<2.5 cm diameter), and rootmat. Rootmat contained a
large amount of decomposing organic matter, as well as live roots, and was not as well developed as that in the Venezuelan Amazon reported by Kauffman et al. (1988).
 
 

Individual equations for each forest component were used for calculating the biomass (Table 2 and Table 3). Biomass of trees 5 cm dbh was calculated from equations based on dbh given by Higuchi et al. (1998) for
Amazonian trees. Biomass of trees <5 cm dbh was calculated from equations based on dbh given by Hughes et al. (1999). Tree height was estimated from a regression equation with tree diameter as the independent
variable. Data for the tree height model were collected from 129 trees in Rondônia. Diameter at breast height of the trees ranged from 1.5 to 238 cm.

 
Steps involved with the methodology
 
 
 

Biomass of CWD was calculated by using the methods of Van Wagner (1968). Transects to measure mass of CWD 7.6 cm in diameter were 15 m long. Pieces of CWD that were 2.5¯7.6 cm in diameter were measured
along the central 5 m of the 15-m transect. Coarse woody debris was further separated into tree (dicot) wood or palm wood components. The 7.6-cm diameter class was also separated into sound or rotten classes
following the methods of Kauffman et al. (1988) and Brown (1971).

One hundred samples for each size class of CWD were collected in forests near Jamari, Rondônia, to obtain an average wood density (sg=0.41 g cm-3 for CWD 2.5¯7.6 cm, 0.49 g cm-3 for sound wood, 0.34 g cm-3 for
rotten wood, and 0.33 g cm-3 for palm wood 7.6 cm diameter). For the 2.5¯7.6 cm diameter classes, the diameter and angle off the horizontal of 65 individual pieces along a 100 m transect were measured to calculate the
quadratic mean diameter and wood particle tilt (Brown and Roussopoulos, 1974). Thereafter, we only counted pieces that intersected the line, and we used density, quadratic mean diameter, and wood particle tilt variables
to calculate biomass.

To calculate forest floor biomass, each sample was initially weighed in the field. Sub-samples were then oven-dried to determine the ratio of wet-to-dry weight. This ratio was then applied to the entire sample to convert
from wet-to-dry weight.

To estimate biomass of basal leaf palms, the number of leaves of each individual palm encountered in the 2 m×10 m plot was counted and multiplied by a mean weight per leaf derived from a random sample of 30 basal
leaves that had been oven-dried and weighed. Three equations were necessary to ascertain biomass of palms: biomass of Attlea sp. 1.78 m high was calculated with the model developed by Anderson (1983); biomass of
other palm species 10 cm dbh was estimated with the model of Frangi and Lugo (1985); and biomass of palms <10 cm dbh was calculated by using a model developed specifically for this study.

Vine biomass estimates were calculated with the model given by Putz (1983). All seedlings (<1.37 m height) were counted in each of the 16 (1 m×1 m) plots per site. Seedling biomass was based on sub-sample of 50
randomly collected oven-dried seedlings from which an average weight per seedling had been determined.

Biomass of standing dead trees <10 cm dbh was calculated from an equation developed by Hughes et al. (1999). Biomass of standing dead trees 10 cm dbh was estimated by first calculating volume, then multiplying
volume by the mean value of specific gravity of sound dead wood. Standing dead palm biomass was estimated from an equation developed for this study for palms <10 cm dbh and by multiplying volume by specific gravity
(0.327 g cm-3) for palms 10 cm dbh.
 

Here is the table of the avorementioned equations and methods
 

 Results

Total aboveground biomass

The mean TAGB for the 20 forest sites was 341 Mg ha-1 and ranged from 287 to 534 Mg ha-1 (Table 1). Mean TAGB of open forest (n=8) was 313 Mg ha-1 and ranged from 288 to 346 Mg ha-1. For dense forests (n=7),
TAGB ranged from 298 to 534 Mg ha-1, with a mean of 377 Mg ha-1. TAGB differed between open and dense forests at the P=0.11 level of probability. For ecotone forests (n=4), TAGB ranged from 298 to 422 Mg ha-1
and averaged 350 Mg ha-1.

Trees 10 cm dbh composed a mean of 78% of the TAGB for all plots combined. Biomass of trees 10 cm dbh differed between open and dense forests at the P=0.13 level of probability, averaging 238±8 and
307±33 Mg ha-1 for the two forest types, respectively. Mean biomass for all trees <50 cm dbh was similar in all three forest types
 
 

Tree density (trees 10 cm dbh) within forest types averaged 429 individuals per hectare (range, 291¯527 individuals per hectare) in open forest and 377 individuals per hectare (range, 223¯487 individuals per hectare) in
ecotone forests, while dense forests averaged 450 individuals per hectare (with a narrower range, 402¯533 individuals per hectare Table 5). Average density of small trees (<10 cm dbh) was highest in open forest and
differed by 25% between open forest (~7500 ha-1), dense forest (~5800 ha-1), and ecotone forests (~4900 ha-1). However, plots within a given forest type varied widely (~2000¯9000 ha-1 in each forest type;
 

So generally the results resembled the others, yet this study had specifics on this one genus, palms.
 

This study provides information on the TAGB and structure of forests that are located in the arc of deforestation described by Fearnside (1990) and are representative of those associated with deforestation in this region.
TAGB and structure was measured on 20 different forest stands and included components not often measured in forest inventories. Estimates of tree density, BA, QSD, and biomass of non-tree components are often used
to formulate TAGB models based on forest inventories ( Brown and Lugo, 1992). The results from this study explain some discrepancies between estimates of TAGB derived from field studies and those from forest
inventories. For example, forest inventories rarely measure palms and trees <10 cm dbh, yet these components comprised an average of 12% of TAGB. The RADAMBRASIL forest inventory was limited to quantification of
trees >31.8 cm dbh. In this study, live trees 30 cm dbh only comprised 56% of TAGB (range, 47¯72%). The non-tree, dead components and smaller wood particles of TAGB are important because they most likely will
completely burn during slash fires following deforestation and, therefore, contribute significantly to CO2 and trace gas flux to the atmosphere (Kauffman and Guild). Underestimating these components could lead to errors in
estimating C pools of standing forests as well as emissions arising from forest conversion. The results from this study, by clarifying forest structure, could be useful to improve TAGB estimates based on other data sets.