TRUNK BIOMASS ESTIMATION BY DIFFERENT METHODS IN A SUBTROPICAL FOREST

The aim of the work was to estimate trunk biomass in a Deciduous Seasonal Forest by using direct and indirect methods. The data were obtained from seven 12 x 12 m plots, bringing to a total of 64 trees. Trunk biomass was determined by direct method and indirect methods which consisted of: 1) Real Volume (RV) of the trunk multiplied by the Weighted Average (WA) and Arithmetic Average (AA) of the specific basic mass of the wood species found, resulting in Real Volume for Weighted Average (RVWA) and Real Volume for Arithmetic Average (RVAA), and 2) by the Estimated Volume (EV) multiplied by WA and by AA, resulting in Estimated Volume for Weighted Average (EVWA) and Estimated Volume for Arithmetic Average (EVAA). The trunk biomass of the three plots determined by the direct method was 11,451 Kg. The indirect method that had the most similar to the real trunk biomass (TB) value result was the EVAA (13,142 Kg), and the one that showed the largest difference was the RVWA, which was estimated at 20,061 Kg. The t-test showed significant difference of trunk biomass for indirect methods that used the RV; and the methods that used the EV did not differ statistically.


INTRODUCTION
According to Ballantyne et al. (2012), climate changes are strongly influenced by emissions of greenhouse gases (GHG), for example, carbon dioxide (CO 2 ).Estimations present that deforestation and degradation of native forests contribute with approximately 20% of the total CO 2 emissions (ANGELSEN, 2009).However, if we consider the diversity of forest ecosystems, the quantification of carbon stocks and the potential of emissions represent one of the greatest challenges within a scenario of environmental changes.
In 2005, a discussion about projects of Reduced Emissions from Deforestation (RED) had started.These projects are currently a priority on the agenda of international negotiations and have evolved to Reduced Emissions from Deforestation and Forest Degradation (REDD).They represent an easy, inexpensive and quick way to reduce GHG emissions (ANGELSEN, 2008).REDD projects are substantiated by concession of financial incentives to preserve forests and maintain carbon stocks in forest ecosystems (TURNER et al., 2009).Despite FLORESTA, Curitiba, PR, v. 47, n. 4, p. 553 -560, out/dez.. 2017. Balbinot, R. et al. ISSN eletrônico 1982-4688 DOI: 10.5380/rf.v47i4.54574 the demand for projects and the need to quantify above-soil biomass accurately, there is still no consensus about the best methodology (direct and/or indirect) to estimate biomass in native forests (GATTO, 2011).
In the direct method approach (destructive), trees are felled and their components are separated and weighted.This method is time-consuming and costly, but it provides accurate information about biomass (LI; XIAO, 2007).By contrast, Brown et al. (1989) claimed that estimations generated from direct methods are unreliable due to the fact that they are based on few tendentious and small selected plots.Fearnside (1991) contested this criticism and observed that methods based on forest inventories (indirect) may be less biased, but the estimations fall short of the values obtained by direct methods.
The indirect method (non-destructive) estimates by means of allometric equations or biomass expansion factors, which are generated from data obtained via the direct method.Non-destructive estimations use variables of forest inventory such as Diameter at Breast Height (DBH), total height, and wood specific gravity (SG) without felling the trees (GATTO, 2011).Also, they are combined with regression models (BREUGEL et al., 2011).In such cases, biomass estimation is much cheaper and faster, but it tends to be less precise.
Knowledge of the average form factor of trees is indispensable for volume-based estimations.It presents variations depending on species, DBH, age and site.By these means, form factor is the third most important variable, after diameter and height, in the determination of volume.Moreover, the basic wood density variable must also be included (FEARNSIDE, 1997).
In short, methods for biomass calculation in native forests still arouse controversy and produce mixed estimations.The working hypothesis is that different methods of biomass estimation generate significantly diverse results, even when the same database is used.In this context, the present study aims to measure trunk biomass of trees in a Deciduous Seasonal Forest fragment by using different methods in order to elucidate a few aspects of this controversy.

MATERIALS AND METHODS
The study was conducted in two areas: the first one is located in the municipality of Frederico Westphalen -RS (27º23'40"S and 53º26'10"W at 530 meters above sea level) and the second, in the municipality of Iraí -RS (27º13'35" S and 53º18'59" W at 240 meters above sea level) in the South Region of Brazil.This area represents a typical fragment of the Deciduous Seasonal Forest of primary succession that was changed by the selective logging of high commercial value species.Climate in the region is Humid Subtropical (cfa by Köppen Classification; well distributed rainfall regime and absence of marked dry season, with average annual precipitation between 1,700 and 1,900 mm (ROSSATO, 2014)).The soil is classified as Aluminum Ferric Red Oxisol (SANTOS et al., 2013).
Three sample units of 144 m² were installed for data collection, which totaled 1008 m², and were randomly distributed.Only trees with DBH above 10 centimeters were sampled.Since it is difficult to establish a parameter that indicates the mark in which the trunk starts to be considered a branch, we considered the morphological inversion point (MIP) as the maximum trunk height in this study.
Rigorous tree scaling was conducted in the field by the Smalian method; trees were measured at DBH and at 0, 25, 50, 75 and 100% MIP.The length of each section was also measured.Total trunk volume was calculated by the equation  =  × ℎ ×  (eq.01), in which  = volume (m³);  = basal area (at DBH); ℎ = height of MIP (m);  = form factor ( = 0.783).Basic density values for wood from each species were obtained from the literature (REITZ et al., 1988;MARCHIORI, 1997a;MARCHIORI, 1997b;MARCHIORI, 2000;LORENZI, 2014a;LORENZI, 2014b;MARCHIORI, 2007;RIBEIRO, 2009;PAULA;COSTA, 2011;TRAUTENMÜLLER et al., 2014); the values were determined for species from which density was not yet known.In order to determine the basic density of wood, discs of approximately two inches thickness at DBH were collected.They were labeled and divided into two symmetrically opposed wedges containing parts of heartwood and sapwood.Wooden wedges were submerged in water, where they remained until constant weight.Basic wood density was determined for each wedge by the equation  = /, in which  (basic wood density) is given in g.cm -³,  (dry weight in an oven at 103º C) in g, and Vu (saturated volume) in cm -³.The green volume was obtained by the method of hydrostatic balance (TRAUTENMÜLLER et al, 2014).
Arithmetic average and weighted average were calculated through the volume of each tree.Trunk biomass was estimated by: 1) Real Volume (RV) x Arithmetic Average (AV) of SG; 2) Real Volume (RV) x Weighted Average (WA) of SG; 3) Estimated volume (EV) x Arithmetic average (AV); and 4) Estimated volume (EV) x Weighted average (WA).These values were compared with trunk biomass (TB) that was determined in the field through the dissect method and were taken as a basis of comparison for the indirect methods.
The biomass analyses were performed by using five treatments (biomass estimation methods) of seven replicates (each field plot was considered as a replication).This characterizes dependence among the samples.Data were submitted to Shapiro-Wilk normality test, Bartlett test and paired t-test.All tests were performed with aid of R (R Development Core Team, 2017).

RESULTS
Sixty-four trees were sampled.Their average values for calculated basic wood density were 0.644 g.cm - ³ for AV and 0.682 g.cm -³ for WA, ranging from minimum of 0.303 g.cm -3 (Erythrina falcata) to maximum of 0.905 g.cm -3 (Eugenia rostrifolia) (Table 2).The value of WA was higher than the one of AV, since the formulation of AV considers the number of species.In the case of WA, volume is the most important.Therefore, large-sized trees and trees with high SG have greater influence on WA.Shapiro-Wilk test demonstrated normally distributed samples (W = 0.950).Bartlett's test accepted the homogeneity hypothesis among the averages of the observed biomass.t-test showed no statistical difference between the weight by estimated volume and the specific mass arithmetic average (EVAA), and the weight by estimated volume and the specific mass weighted average (EVWA), when compared to TB.However, the relations of the weight by real volume and the specific mass arithmetic average (RVAA) and the weight by real volume and the specific mass weighted average (RVWA) revealed significant difference (p < 0.05).
The calculated volumes were 29.38 m³ for RV and 20.42 m 3 for EV (eq.01).The calculated weight for MIP was 20.06 and 18.91 Mg for RVWA and RVAA, and 13.94 and 13.14 Mg for EVWA and EVAA, respectively.TB determined in the field was 11.45 Mg.Values for the summation of the three plots are presented on table 3. Sample units (SU); real volume and basic specific mass weighted average (RVWA); real volume and basic specific mass arithmetic average (RVAA); estimated volume and specific mass weighted average (EVWA); estimated volume e specific mass arithmetic average (EVAA); trunk biomass (TB); Coefficient of variation (CV); Coefficient of linear determination (R 2 ).*Means followed by the same letters do not differ significantly (5%).

DISCUSSION
Values of SG calculated for the species from the Deciduous Seasonal Forest (both for AV, 0.649 g.cm -³, and for WA, 0.682 g.cm -³) are close to the values found by other authors for subtropical and tropical forests (Table 3).However, a wide variation in the values of basic wood density is found.For example, when studying 90 species in tropical forests, Muller-Landau (2004) found values between 0.12 and 1.05 g.cm -3 for different sites located in the same type of forest.The results (Table 4) show that estimations that use basic wood density as one of the variables might lead to error in the calculations of biomass, above-soil biomass and carbon.Table 4. Arithmetic average (AV) and weighted average (WA) for wood specific gravity in different forest ecosystems (g.cm -3 ).Tabela 4. Média aritmética (MA) e média ponderada (MP) da massa específica básica da madeira em diferentes ecossistemas florestais (g.cm -3 ).
Source Type of Forest AV (g.cm -3 ) WA (g.cm -3 ) Variations of basic wood density in forest species occur due to the differences in the anatomic structure of the wood and to the quantity of extractive substances in each individual.These alterations occur on the basis of tree age, genotype, site quality, climate, geographic location and growth rate (TREVISAN et al., 2012).In addition, trunk irregularities, such as the occurrence of emptiness and rotting associated with age difference among individuals (MULLER-LANDAU, 2004) and senescent trees, cause a variation in the SG as well as in the real volume of the trunk.
As for trunk volume, EV (eq 01) underestimates the wood volume of the trunk.This difference can be partly explained by the variation in trunk forms of trees depending on species, site, age, genetic aspects, sanity, etc.According to Machado and Figueiredo Filho (2003), studies about trunk forms were directed at trees that could be considered as regulars, i.e., that could be compared to defined geometric figures.However, current studies that try to estimate biomass stockpile need to consider all the trunks, regardless of species, site, sanity, genetics, age, etc.Based on that, research on biomass evaluation techniques is needed.
The results confirm the considerations of Fearnside (1991) regarding direct methods that affirm that indirect methods overestimate the real weight of trunks.There is clear disagreement on biomass estimations, which can create overestimation of 75.2% in the case of WRVWA .TB considers every variation of SG in radial, tangential and longitudinal directions and the presence or absence of hollow trunks at the initial stage of rotting.By contrast, indirect methods do not consider this variation; only volume and SG.
Methods that use the WA of SG tend to increase this error when compared to the use of AV, since WA attributes greater importance to trunks of higher volume.Trunks of higher volumes belong to older trees, which are more likely to present emptiness and rotting consequently.Accordingly, trees of large diameters must be carefully analyzed when using both direct and indirect methods, because few individuals might represent a large part of the biomass of a forest ecosystem (in this case, more than 50%).

CONCLUSION
 The direct method is the most effective way to quantify trunk biomass because it assimilates all the variations. Indirect methods that use estimated volume did not differ statistically for trunk biomass; and indirect methods that use real volume presented significant differences for trunk biomass.

Table 1 .
Form Factors for commercial height in different types of native forests of Atlantic forest.Tabela 1. Fator de forma para altura comercial em diferentes tipologias florestais nativas do bioma Mata Atlântica.

Table 2 .
Basic wood density (g.cm -³) for the evaluated species.Tabela 2. Massa específica básica (g.cm -³) para as espécies estudadas.specific gravity was not found in the literature; thus the average value of the other species was used.

Table 3 .
Results for trunk biomass of each plot, treatment and statistical analysis by the t test.Tabela 3. Resultados da biomassa do tronco de cada parcela, tratamento e análise estatística através do teste t.