Study area

The study was carried out in the private plot of La Cabaña farm, located between 15° 56’ 13.54” and 15° 55’ 12.85” N, and 96° 24’ 09.13” and 96° 23 51.83” W, on a surface of 300 ha found in the physiographic region of the Sierra Sur, in Oaxaca, municipality of Pluma Hidalgo. The climate is semi warm humid with abundant summer rains (ACm), temperature of 18 to 26 °C, and precipitation of 1500 to 3500 mm. The geomorphology of the zone is complex low sierra with Phaozem soil type (^{INEGI, 2010}). The vegetation corresponds to a sub-evergreen tropical forest where the tree stratum is made up primarily by the following species: A. latifolia Sw., C. dentata DC., I. punctata Willd., Astronium graveolens Jacq., Pterocarpus acapulcensis Rose, Inga paterno Harms, Homalium trichostemon S.F. Blake, Saurauia serrata D.C., Cecropia obtusifolia Bertol., Diphysa robinioides Benth., Lonchocarpus guatemalensis Benth., Calycophyllum candidissimum (Vahl) DC., Cupania glabra Sw. and Bursera simaruba (L.) Sarg.

The species selected for this study were A. latifolia, C. dentata and I. punctata, and they are part of the natural forest. However, these species are not considered ecologically important in México (^{Pompa-García et al., 2017b}), despite their abundance in the natural vegetation and their use in traditional systems of coffee production in the Sierra Sur of Oaxaca, since they are species that provide shade and are multipurpose (^{Ventura-Aquino et al., 2008}). Likewise, the wood of A. latifolia and C. dentata is used for constructions of houses, rural works, and handles of agricultural tools (^{Pennington and Sarukhán, 2005}). Inga punctata provides some secondary products such as fodder, fruits and fuel to agroforestry producers.

Obtaining samples

For all the species, three diametrical categories were selected: CD I (less than 10 cm), CD II (between 10 and 20 cm) and CD III (more than 20 and less than 35 cm), and three individuals were elected from each category. The tree crown was divided into three sections: low, medium, high (^{Zhang et al., 2009}) and in each of them, a branch was selected randomly from which a disk (approximately 5 cm thick) was obtained, and from the same an approximate sample of 250 g of foliage. From the stem, a disk was obtained at a height of 1.30 m from the ground level. All the samples were placed in paper bags with their corresponding labels and were transported to the Laboratory of Soil Fertility and Environmental Chemistry in Colegio de Postgraduados, where they were dried in a stove (Nyle model L500) at 70 °C until constant weight as indicated by the procedures described by ^{Etchevers et al. (2005)} for determination of CC in plant samples. Then, from the material of branches and leaves collected in each section of the crown, a subsample of 50 g was taken to form a single compound sample from each component of the tree, and from the trunk slice a subsample was also obtained that included sapwood, heartwood and bark in such a way that it included the same proportion. All the subsamples were cut into small fractions (splinters) and ground with knifes and scissors. In order to achieve homogeneity in the material, the samples were pulverized until they passed through the mesh 40 in a General Electric Mod. 5KH39QN5525 mill, and were deposited in envelopes labeled for their determination.

Measuring carbon

Before the carbon analysis, the compound subsamples were dried at 70 °C for 24 h to take out the moisture absorbed from the environment. After, around 30 mg were placed in a Shimadzu TOC 5000-A automatic analyzer of total carbon, to measure the CC by the dry combustion technique at 900 °C and measuring the CO₂ released through infrared spectrometry (^{Etchevers et al., 2005}). A total of 81 compound subsamples were analyzed, corresponding to 27 individuals per species.

Statistical analysis

A three-factor ANOVA was carried out with the data to evaluate the effects of species, diametric category, and structural component in the CC and the interactions corresponding to the species. For the species factor three levels were considered (A. latifolia, C. dentata and I. punctata), the same as the diametrical category factor (CD I, CD II, CD III), and the structural component factor (stem, branches, leaves), grouping the data into a completely random design with factorial arrangement 3X3X3 with three repetitions. The Tukey test (p≤0.05) was used for the factors or interactions that presented significant differences. The assumptions of normality and homogeneity of variances were validated for the model through the Shapiro-Wilk and Levene test (^{Montgomery, 2012}). In addition, the differences between the average CC from each species and the generic value of standard biomass conversion (50 %) (^{Basuki et al., 2009}; Houghton, 2005; ^{Thomas and Martin, 2012}) were evaluated statistically with a t-student test (p≤0.01). These analyses were performed with the R software version 3.3.3 (^{R Development Core Team, 2017}) and with the agricolae package (^{De Mendiburu, 2016}).

Carbon concentration per species

The results from the laboratory showed that the CC in the three species was found in an interval of 44.13-48.70 %, with a general mean of 46.20 %. The CC was significantly different between species (F _{2, 54}=14.03, p≤0.0001). The species with the highest average CC was C. dentata with 46.71 %, and then A. latifolia (46.02 %), and I. punctata (45.86 %) (Table 1). The results from the t-test were for A. latifolia (t=-18.32, p≤0.001), C. dentata (t=-17.32, p≤0.001) and I. punctata (t=-27.23, p≤0.001), indicating that the average CC in the three species of tropical trees was statistically lower than the 50 % established as a conversion factor with regard to the biomass.

The variations in the CC in the three species shown in our study are similar to the ones found by ^{Elias and Potvin (2003)} for 32 tropical species from Panama (44.4 to 49.4 %). ^{Gayoso and Guerra (2005)} measured the CC in 16 species of native forests from Chile, and reported a variation of 34.9-48.3 % with an average of 43.7 %, while ^{Yeboah et al. (2014)} reported the CC in an interval of 45.8- 49.8 % in plantations of 19 species of tropical trees of 7 to 12 years in two ecozones of Ghana. Figueroa et al. (2005) determined a higher CC for the species of Inga (51.16 %) and of 51.19 % for six species from the mesophyll forest in the Sierra Norte of Oaxaca, México. In addition, ^{Becker et al. (2012)} point out that the variation of CC was 47.5-50 % in 23 tropical species in Eastern Africa, with a mean of 50 %. Hernández-Vázquez et al. (2012) reported an average CC of 40.8 % for the species Inga jinicuil Schltdl. in coffee agroforestry systems in San Juan Tepanzacoalco, Ixtlán, Oaxaca, value lower than that of I. punctata in our study. ^{Pompa-García et al. (2017a)} found that the CC in 82 species of trees in tropical and subtropical zones of México was 42.8 to 44.3 %, values lower than those found in our study.

The differences in the CC in the studies and in the three species analyzed in our study were attributed primarily to the physical and chemical properties of the tree tissues (^{Thomas and Malczewski, 2007}). The variation in the CC between species of tropical trees is caused by the different proportions of cellulose and lignin (^{Martin and Thomas, 2011}). In addition, in most of the studies the volatile compounds of C are unknown in the total concentration of C, which may cause a significant under estimation in the content of total C (Martin et al., 2013). Likewise, the effects of the ecological attributes of the species, such as tolerance to shade, influence importantly the variation of CC in the tree tissues (^{Gao et al., 2016}). The trees with high content of lignin tend to present high levels of CC; thus, the species of conifers have a higher CC in the wood (50.8±0.7 %) compared to the species of tropical angiosperms (47.1±0.4 %) (^{Thomas and Malczewski, 2007}). In the same way, tree species of temperate/boreal zones present a higher variation of CC (43.4-55.6 %) than the species from tropical zones (41-51.6 %) (^{Thomas and Martin, 2012}).

Carbon concentration by component and diametric category

The average values of CC in the structural components of the tree in the three species were 45.77 %, 46.04 % and 46.77 % in the stem, the branches and the leaves, respectively (Table 2), while the average CC was 45.84 % in CD I, 46.12 % in CD II and 46.63 in CD III (Table 3). The results from the variance analysis indicated that the CC was significantly different between structural components of the tree (F _2,18=18.74, p≤0.001) and between diametrical categories (F _2,18=6.79, p≤0.001) in I. punctata. Likewise, in C. dentata the CC showed significant differences between the structural components (F _2,18=25.20, p≤0.001) and between the diametric classes (F _2,18=8.79, p≤0.001). Meanwhile, in A. latifolia only the diametric classes presented significant differences (F _2,18=6.98, p≤0.01). The interaction structural component by diametric class for the three species did not show significant differences (p˃0.05).

In A. latifolia, the CC was 46.40 % in the branches, 46.10 % in the leaves and 45.56 % in the trunk, although without significant differences (p˃0.05). Instead, the CC from leaves (47.71 %) of C. dentata was significantly higher than that from the trunk (46.11 %) and branches (45.62 %) (p≤0.01). A similar behavior was observed in I. punctata: the CC was significantly higher than the leaves with 46.51 %, compared to the branches (45.62 %) and the trunk (45.46 %) (p≤0.01). The trees with diametric categories above 20 cm (CD III) had higher values (p≤0.05) in the CC than the CDI and CDII in the three species (Table 3). The lowest CC was found in the stem of the CDI of I. punctata with 44.82 %, and the highest in the leaves of CD III of C. dentata with 47.72 %. All the values of CC in the structural components and in the diametric categories were lower than 50 % in the three species (Figure 1).

Figure 1 Average carbon concentration (CC) in the trunk, the branches and the leaves in the different diametric categories (DC) for the three species of tropical trees. The data are means ± standard deviation (n=3). Different letters in the bars indicate statistically significant differences (Tukey; p≤0.05). 

The results obtained in CC for I. punctata had a similar performance to the study carried out by ^{Pompa-García et al. (2017a)} for Inga vera Willd. in the tropical forests of México, with a higher CC in the leaves (49 %) and 46 % of CC in the trunk; however, the CC in the branches (44 %) was lower than the one found in our study. ^{Pompa-García et al. (2017)} determined a higher CC in the trunk (48 %) and branches (47 %) in Croton draco Schltdl. & Cham., compared to the ones found in A. latifolia, and a lower CC in the leaves (39 %), although both species belong to the same family (Euphorbiaceae).

The three species showed the highest CC in the leaves, then in the branches and finally in the stem (Table 2). This variation is because in the leaves there is a higher accumulation of carbohydrates and the highest rates of volatile compounds (^{Watson and Casper, 1984}; ^{Yerena-Yamallel et al., 2012}). The wood in the branches has a higher concentration of lignin and extracts from the stem wood, which results in a higher CC (^{Bert and Danjon, 2006}), while the stem contains higher relative concentrations of non-structural carbohydrates (starch and simple sugars) (^{Hoch et al., 2003}).

The trees in the higher diametric categories (CD III) showed a higher CC in the three tropical species; these patterns of changes in the increase of the CC dependent on the size of the tree were observed in other species of tropical trees, in Dominican Republic and Panama (^{Martin et al., 2013}; ^{Martin and Thomas, 2013}). In the process of tree growth, the content of lignin decreases, and is compensated by the increase in other compounds of C in the wood, such as volatile compounds and, as consequence, a higher CC (^{Martin et al., 2013}). In addition, the increase of CC due to the size of the tree can be because of a higher formation of tension wood in the larger trees in response to the increase of crown load in individuals of larger size (^{Du and Yamamoto, 2007}).

The results from our study contribute key information about the inter-intra specific variation of CC and should be taken into account when quantifying the capture and storage of C in the forest. In addition, the incorporation of CC values in different structural tree components and diametric categories in allometric equations of biomass for the three tropical species (^{Aquino et al., 2015}) are an important tool to improve scenarios of global C emissions (^{Gao et al., 2016}).