Study area

The study was conducted in a livestock production system in the Villaflores municipality, Chiapas (Figure 1). The dominant climate of the region is warm sub-humid, with rains in summer and average humidity. The sampling sites are located between 16°23'57.63" N and 93°11'50.91" W, at 847 m, 1 200 - 1 500 mm of rainfall in a five month - rainy period. The predominant type of soil is Leptosol. The main type of vegetation in the region is the sub-deciduous tropical forest, with a greater abundance of species such as Quercus rugosa Née, Acacia collinsii Saff., Bursera simaruba (L.) Sarg., Enterolobium cyclocarpus (Jacq.) Griseb. and Guazuma ulmifolia Lam.

Source: Own elaboration with maps of ^{INEGI (2012)} and ^{Google Earth (2015)}.

The sites from F1 to F6 represent forest areas and from P7 to P12, pasture areas.

Leyenda = Legend; Forestal = Forestal; Potrero = Paddocks

Figure 1 Location of the sampling points within the Villaflores municipality, Chiapas, Mexico. 

The study area has a total area of 62.5 ha, of which the forest area covers 39.5 % (24.54 ha), 40 % to paddocks areas and the rest has other uses such as seasonal agriculture, fruit orchards and urban infrastructure (buildings). In Figure 1, the distribution of carbon sampling plots in forest areas and paddocks is shown.

Establishment of plots and quantification of live biomass

The samples were collected in the forest and grazing areas of the Los Flamboyanes cattle ranch. Field trips were made in the forest areas of the ranch through which twelve plots were defined in total, six of forest that did not have road crossings or human disturbance, and six other plots of pastures with or without trees. Subsequently, another systematic data capture was made, adapted to the methodology of the Inventario Nacional Forestal y de Suelo (National Forest and Soil Inventory) (INFyS) (^{Conafor, 2005}).

Rectangular plots of 40 × 25 m (1 000 m²) were established and with a minimum distance of 200 m between them. Under the principle of nested plots, all trees were recorded with a DBH (1.30 m from ground level) of at least 2.5 cm on a surface area of 500 m², and trees with a DBH of 10 cm onwards on 1 000 m².

The total height of the trees was measured with a graduated bamboo stick and the measurements were corroborated with the inclinometer. To determine the dominance of the main species identified in the sampling area, the sum of the basimetric area in m² of each of the species per hectare was calculated.

The following formula was used for the basimetric area:

Where:

AB = Basimetric area of the tree

ϖ = Constant (pi)

DAP = DBH

The biomass area of the trees was quantified by the equation proposed by ^{Cairns et al. (2003)} with the wood density correction of each species. In order to evaluate the fit with the data and make the comparison in the amount of biomass between two equations, the equation of ^{Chave et al. (2015)}.

Where:

BA = Aerial biomass

DAP = DBH

AT = Total height of the tree

P = Wood density

Pm = Wood density of the species used for the model

For the calculation of root biomass, the following formula (^{Cairns et al., 1997}) was used:

Where:

BR = Root biomass

BA = Aerial biomass

In order to quantify the biomass of the associated vegetation (herbs, shrubs and trees <2.5 cm DBH), four random samples were taken in a quadrant of 1 × 1 m in each plot. The samples were dried in the oven (RIOSSA HCF-125) at 70 ^oC for up to 72 hours to obtain the dry weight of the biomass. Once the above was achieved, carbon was quantified, by means of the carbon fraction in the biomass (47 %) and, subsequently, extrapolated to Mg C ha^-1 (^{Aryal et al., 2014}; ^{Fonseca et al., 2011}; ^{Pompa-García and Sigala, 2017}). To calculate Mg of C ha^-1 with the carbon data per tree, the following formula was used:

Mulch and dead wood

Four samples were taken, completely at random, with a 50 × 50 cm quadrant in each plot. The mulch samples were collected separately in three states of decomposition: fresh (Oi), fragmented and in the process of decomposition (Oe) and decomposed or humus (Oa). The samples were processed in the laboratory of the School of Agronomic Sciences Campus V, of the UNACH, where the dry weight of the samples g m^-2 was determined. Then the amount of carbon (Mg ha^-1) was obtained by multiplying it by the carbon fraction according to its state of decomposition: Oi (39 %), Oe (36 %), Oa (27 %), proposed by ^{Orihuela-Belmonte et al. (2013)}.

The dead wood was measured in four 25 m transects in each plot, where the diameters of 2.5 cm onwards were taken in the first 10 m and 5 cm onwards in the remaining 15 m. They were classified into three decomposition states - strong, intermediate and decomposed - since the density of wood varies according to its state of decomposition. The wood volume of each of them was calculated with the following formula:

Where:

V = Dead wood volume (m³ ha^-1)

L = Length of the sampling line (m)

d_i = Dead wood diameters at the intersection (cm)

The volume of wood was multiplied by the density of wood according to its state of decomposition: strong (78 %), intermediate (62 %) and decomposed (27 %) to quantify the dry weight of wood (^{Eaton and Lawrence, 2006}). The carbon content of dead wood biomass was calculated with the carbon fraction (47 %) proposed by ^{Fonseca et al. (2011)}.

Organic soil carbon

Four soil samples were taken at random to a depth of 15 cm in each plot, with a nucleator (AMS) of 5 cm of internal diameter to quantify the apparent density of soil and another four samples for chemical analysis of soil. The samples for bulk density were processed in the laboratory, where the dry weight was obtained (105 °C, 72 h), and then with the cylinder volume data, the apparent soil density (g cm^-3) was calculated. The content of organic matter, pH and electrical conductivity of the soil were analyzed in the laboratory of the Facultad de Ciencias Agronómicas Campus V, de la Universidad Autónoma de Chiapas (School of Agronomic Sciences Campus V, of the Autonomous University of Chiapas). With the data of carbon content (%), bulk density (g cm^-3) and coarse fraction (% of rocks and roots), the storage of organic carbon in soil (Mg C ha^-1) was calculated, with the following equation:

Where:

CC = Carbon content in soil samples (%)

DA = Apparent soil density (Mg m^-3 or g cm^-3), fitted for coarse fractions

PS = Depth of sampled soil (m)

Data analysis

In order to calculate the abundance of each tree species, the number of individuals registered within the sampling area was used and then extrapolated to individuals per hectare by two diametric classes (<10 cm DAP and ≥10 cm DAP). The species with the highest relative abundance (%) were considered as more abundant species. The dominance of forest species was determined considering the sum of the basal areas per hectare (m² ha^-1) and their relative contribution in total basal area. The comparison of carbon content in each reservoir between forest area and paddock area was analyzed by using the one-way ANOVA (p <0.05). Subsequently, the Tukey test was used (p <0.05) if necessary. Basic statistics were determined as mean, standard error and frequencies for each of the reservoirs and systems. The non-linear regression analysis was performed to compare the goodness of fit of two allometric equations (^{Cairns et al., 2003}; ^{Chave et al., 2015}) in their ability to estimate live biomass with tree DBH data. The analyzes were carried out with the statistical package Statistica 7.0 (^{Statsoft, 2007}).

Tree diversity of the forest areas

Abundance of trees

In the forest plots 46 different tree species were found, the main 20 species with the highest abundance per hectare are shown in Figure 2. Quercus rugosa Née, Acacia collinsii Saff., Guazuma ulmifolia Lam., Bursera simaruba (L.) Sarg. and Daphnopsis americana (Mill.) J. R. Johnst. were the most abundant in the study area.

Abundancia (árboles ha^-1)= Abundance (trees ha^-1)

Figure 2 Abundance of the main forest species identified. 

Dominance of trees

It can be seen in Figure 3 that the species with the highest dominance was Quercus rugosa, with a total AB of 11.07 m² ha^-1, followed by Enterolobium cyclocarpum (Jacq.) Griseb. with 1.62 m² ha^-1. The species of least dominance were Diphysa robinoides Benth. & Oerst., Albizia nipoides (Benth.) Burkart and Coccoloba cozumelensis Hemsl., with a basimetric area equal to or less than 0.01 m² per hectare.

Área basal = Basal area

Figure 3 Dominance of the main forest species identified in the forest areas. 

The role of the remaining vegetation in fragmented landscapes is appreciated not only for its capacity for carbon accumulation and mitigation of climate change, but also for its contribution to natural regeneration and conservation of biodiversity (^{Chazdon, 2008}, ^{Martínez-Sánchez et al., 2015}; ^{Mora et al., 2018}). Fragments of forest areas (about 40 % of the total area of the ranch) are home to more than 40 species of forest trees with different levels of abundance and dominance within the arboreal community.

Of the tree species recorded in the study area, few are the most abundant, which may be due to species such as Quercus rugosa, Guazuma ulmifolia and Bursera simaruba, which are the most numerous in surrounding patches of the landscape, produce fruits and seeds, which are food for wild animal species, which scatter seeds easily throughout the forest (^{Chazdon et al., 2010}). In the same way, these species can adapt to forest ecosystems with low and medium levels of human disturbance.

In a review of 161 studies conducted in Mexican territory, ^{Pompa-García et al. (2017a)} demonstrated the importance of forest ecosystems in the conservation of diversity and of ecologically important forest species. The authors concluded that the most focused ecosystem is the low deciduous forest and the predominant species correspond to the genera Quercus, Pinus, Acacia and Bursera, which are of great importance for the resilience effects of disturbed ecosystems.

When comparing these systems with other forest systems, it was found that the values of both the number of species (46) and the total number of trees per hectare (1 081) were higher than the results of ^{Basáñez et al. (2008)} with 30 species and 810 individuals ha^-1, in total, in the middle sub-perennial forest. ^{Taylor (2016)} records a greater number of species (95 in total), but fewer individuals (303 ha^-1) in pine-oak forest, than those recorded in this work. In forests of Quercus rugosa, the values obtained by ^{Hernández (2015)}, of 983 individuals ha^-1 of which 820 individuals had a DAP of 0-10 cm, are similar to those consigned here (1 081 individuals ha^-1, of 663 with DBH <10 cm), in sub-deciduous forests; this suggests that the forest under study the existence of natural regeneration processes.

The dominance of the main species identified in this work, expressed in the basimetric area, is lower than the values of ^{Basáñez et al. (2008)} in Papantla, Veracruz, in the middle subcaducifolia forest, with 76.77 m² ha^-1. In Los Altos de Chiapas, ^{Taylor (2016)} calculated 49 m² ha^-1, in pine-oak forests. In San José, State of Mexico, ^{Hernández (2015)} recorded 12.31 m² ha^-1 in Quercus rugosa forests.

Carbon reservoirs

The highest accumulation of carbon is concentrated in the living aerial biomass of the trees with 46.71 ± 9.00 Mg C ha^-1 (average ± standard error) for the forest area and 0.30 ± 0.13 Mg C ha^-1 in the paddock area (Table 1). There was a significant difference between the forest area and the pasture area in aerial biomass and root biomass (P = 0.00012). The associated vegetation collected more carbon in the pasture area 1.14 ± 0.09 Mg C ha^-1 than in the forest area 0.33 ± 0.05 Mg C ha^-1, with a significant difference (P = 0.0000) between the two systems.

The proper selection of allometric equations to quantify the living biomass of the trees can reduce the uncertainty in the estimation of carbon storage. In a comparative analysis, it turned out that the amount of carbon quantified with the equation of ^{Cairns et al. (2003)} was 28.2 % lower than with the equation of ^{Chave et al. (2015)}. The first one was developed with the forest species of the medium sub-evergreen tropical forest of the Yucatán Peninsula, while the second one using data collected from several regions of the tropics. In a non-linear regression analysis with the data of the study of interest, Chave's equation had greater goodness of fit (R² = 0.96) than that of Cairns (R² = 0.78), which supposes its best applicability (Figure 4).

Biomasa = Biomass; DAP = DBH

Figure 4 Non-linear regressions between DBH and live biomass estimated by two allometric equations. 

The amount of carbon stored in the mulch of forest areas varied according to the state of decomposition (Figure 5). A greater amount of carbon stored in the horizon Oa (decomposed material) was recorded compared to the horizons Oi (fresh) and Oe (partially decomposed). There is no reservoir of mulch in the paddock area.

Carbono en mantillo = Carbon in mulch; Horizonte = Horizon; Oi Fresco = Oi Fresh; Oe Intermedio = Oe Intermediate

Different letters above the bars indicate the significant difference (p <0.05, Tukey).

Figure 5 Carbon accumulation (Mg C ha^-1) in different mulch horizons in the forest areas. 

Organic carbon and soil properties

In the organic carbon content of the soil between forest areas and paddocks (p = 0.268, one-way ANOVA) no significant difference was found. The organic carbon accumulated in forest land was 72.57 Mg C ha^-1 with a standard error of 2.89, very similar to the carbon content in paddock soil 74.37 Mg C ha^-1 with a standard error of 3.30. When comparing the two systems, forestry and paddocks in apparent density, pH, electrical conductivity and percentage carbon content, no significant differences were observed (Table 2).

A comparison of the soils of paddocks, good and eroded, there was no significant difference in carbon content, 73.81 Mg C ha^-1, figure similar to that of the degraded paddock, 74.92 Mg C ha^-1, apparent density and carbon in percent (P> 0.05, one-way ANOVA). The pH presented a significant difference (P = 0.030), of 5.56 in the soils of good paddocks and 4.87 in the eroded ones (Table 3). In a similary way, there is a significant difference in electrical conductivity (P = 0.042), since it was lower in good paddocks (100.01 mS m^-1) than in eroded paddocks (139.29 mS m^-1). The higher values of electrical conductivity in the latter could be attributed to the high concentrations of salts.

When comparing the total carbon stored between the two different study areas, the forest area registered a greater amount of carbon (144.45 Mg C ha^-1) than the pasture area (75.95 Mg C ha^-1). Organic soil carbon (SOC) was the main reservoir in the two study systems (Table 4). In forest systems, COS contributes 50.2 % of total carbon and in pasture areas, with 97.9 % (Table 4).

Separating certain conservation areas within a livestock production unit is not a common practice in the Mexican southeast. However, some producers, consciously or unconsciously, maintain a forest area that accumulates atmospheric carbon and contributes to reducing the net emissions of greenhouse gases (CO₂ equivalent) from their ranches. The accumulation of carbon in biomass and soil within these production units also significantly supports the control of avoided deforestation emissions (^{García-Oliva et al., 2006}; ^{De Jong et al., 2010}; ^{Soto-Pinto et al., 2010}; ^{Nahed-Toral et al., 2013}). In this sense, the ranch forest area of this study helps in an ecological way to the storage of carbon in living area biomass, underground living biomass, dead wood, leaf litter and organic soil carbon.

As expected, the amount of carbon stored in reservoirs such as live tree biomass, dead wood and litter were higher in forest areas than in pasture areas. However, the hypothesis that organic soil carbon in the first areas is higher than in the second was rejected because no significant differences were found.

This can be attributed to the following factors: i) the organic carbon in the soil is a more stable reservoir and can be affected slowly by the change in land use. These patterns have also been observed in other studies in south-southeast Mexico (^{Orihuela-Belmonte et al., 2013}; ^{Aryal et al., 2015}, ^{Villanueva-López et al., 2016}; ^{Aryal et al., 2017}); ii) the contribution of organic matter by the replacement of roots of pasture areas can contribute in a significant way to maintain the organic carbon of the soil in pastures. This is consistent with the present result that the carbon accumulated in the associated vegetation (pastures, shrubs and trees <2.5 cm of DBH) was greater in pasture areas than in forest areas. However, studies on the production, turnover and decomposition of roots between forest areas and pastures are required to understand well the organic carbon dynamics of soil in these ecosystems. In addition, pasture management in terms of: grazing rotation, number of animals per unit area, addition of organic matter and synthetic fertilizers can explain the variations and trends of organic carbon accumulation in pasture areas, which were not considered in this study.

The average values of carbon stored in the arboreal component of this research are lower than those calculated by ^{Hernández (2015)} who recorded 109 Mg C ha^-1 in tree biomass. As for the biomass components of roots and associated vegetation, the same author calculated 19.88 and 1.44 Mg C ha^-1, which are very similar to those of this work. ^{Ibrahim et al. (2006)} reported an average of 90.78 and 4.83 Mg C ha^-1 of carbon in live aerial biomass in secondary forests and pastures in monoculture respectively in Esparza, Costa Rica. ^{Aguilar (2007)} obtained 1.99 Mg C ha^-1 in aerial biomass in the paddocks of monoculture systems in Marqués de Comilla, Chiapas; similar to those in this work.

In the biomass component of the associated vegetation, higher carbon content was found in the paddock system than in the forestry system, possibly because they receive greater solar radiation and due to little competition among the members of the herbaceous stratum in the second. This difference can be a function of the density of trees, type of species, age of the ecosystem and number of strata present (^{Aguilar, 2007}). In addition, it has recently been reported that the carbon stored in living biomass varies not only from one system to another, but also from one species to another, to which we should add the ecological conditions derived from the difference in carbon concentrations in the tissues (^{Pompa-García et al., 2017b)}.

The results of carbon stored in dead wood in this investigation (3.49 Mg C ha^-1) were lower than ^{Bolaños et al. (2017)} and ^{Aryal et al. (2014)} who calculated 16 and 35.87 Mg C ha^-1, respectively, for pine forests and secondary sub-evergreen forests. The value obtained from carbon in mulch of the present research (4.89 Mg C ha^-1) in sub-deciduous forests is lower than those of ^{Bolaños et al. (2017)} of 11 Mg C ha^-1. ^{Hernández (2015)} recorded 17.09 Mg C ha^-1 for mulch in Quercus rugosa forests.

The organic carbon data of the soil in the work described here are similar to those calculated by ^{Ruíz (2002)}, who reported an average of 84.3 Mg C ha^-1 at a sampling depth of 20 cm, in paddocks without trees. ^{Bolaños et al. (2017)} obtained 68 Mg C ha^-1 at a sampling depth of 30 cm, similar to those of this investigation. ^{Céspedes et al. (2012)} determined 29.05 Mg C ha^-1 at a sampling depth of 10 cm and 23.58 Mg C ha^-1 at 20 cm in pastures, which is lower than the current figures (74.37 Mg C ha^-1 at 15 cm depth of sampling). ^{García-Olivia et al. (2006)} calculated 3.6 % of C in dry forest soils of Chamela, Jalisco, a percentage close to the data gathered here. The variation of these C contents in the soil may be attributed to the management of grazing, type of soil, physiological state of the grass and the content of nutrients (^{Aguilar, 2007}).

The total carbon obtained in the present investigation in sub-deciduous forests and paddocks (144.45 and 75.95 Mg C ha^-1) is lower than that reported by ^{Aryal et al. (2014)} which is 231.67 Mg C ha^-1 in secondary forests of the Yucatán Peninsula. ^{Hernández (2015)} reached 302.50 M g C ha^-1 of total carbon in the temperate forest of the State of Mexico. ^{Bolaños et al. (2017)} calculated 376 Mg C ha^-1 of total carbon in pine forests. ^{Ruiz et al (2004)} reported a total value of carbon stored in native grasses in Matíguas, Nicaragua of 151.4 Mg C ha^-1, similar to that of this work. ^{Aguilar (2007)} determined 64.62 Mg C ha^-1 in monoculture pastures in Chiapas, Mexico, which is below the results generated here.

The ranch where this research was carried out includes 24.7 ha (40 % of the total area) of forest areas, in which each hectare stores an average of 144.45 Mg C ha^-1; therefore, this extension of felled forest to be used for pasture releases 257.76 Mg of CO₂ ha^-1. If it were decided to fell 24.65 ha of forest to plant pastures and give them the same use, it would provide 6 353.95 Mg CO₂ of GHG emissions.

By means of a GHG estimate with energy equivalences, ^{Molina (2014)} quantified 93.15 Mg CO2eq. of emissions during six months during the dry season in the study ranch. Based on this contribution, 23 % of these emissions can be reduced through agroecological technologies such as the use of biodigesters, composts and silvopastoral systems. It is expected that the removal of carbon by forestry and agroforestry activities will contribute significantly to the mitigation of the remaining emissions.