Field information

For field data collection, an adaptation to the design of sampling sites proposed by ^{Brown et al. (1981)} was used. To do this, two permanent plots of 1 hectare each (100 x 100 m) were delimited, within each of these 25 subplots of 400 m² (20 x 20 m) were drawn, in the center a site as shown in Figure 2 was installed, leaving a total of 25 locations per plot systematically distributed at 20 meters each. With these, a surface necromass inventory was carried out, including the dead wood material (MLC) and the organic layer of soil (COS) that covers the fallen leaves layer (HO) and fermentation (HF). Additional to necromass, this design also allows to take a sample of vegetation which is called lower tree than 7.5 cm in diameter at the height of 0.30 m from the floor (d_0.3 <7.5 cm) (Figure 2).

Figure 2 Escheme of sampling site involves necromass and lower tree. 

Sampling dead wood material (MLC). A planar intersection technique (^{Brown, 1974}; ^{Brown et al., 1981}; ^{Van Wagner, 1982}, ¹⁹⁶⁸) was used for this case; which uses a classification of woody parts, based on its thickness and time lag, i.e. how long it takes a woody part in gaining or losing 2/3 of its moisture to a change in the environment. Based on this classification, the woody parts are grouped in diameter categories as: MLC parts 1 hour lag (<0.6 cm in diameter), MLC 10 h (0.6-2.5 cm), MLC 100 h (2.6-7.5 cm) and MLC 1 000 h (> 7.5 cm), the latter also classified between firm and decomposed because wood density is different (^{Brown et al., 1981}).

The planar intersections technique consists to locate a sampling straight line over the forest ground and count all the MLC parts that are through it. In this study 50 sample lines were established, 25 in each plot, each line has a total length of 15 m; MLC parts were recorded at different lengths depending on their time lag, the 1 and 10 h were measured in 2 meters, the 100 h were recorded in 4 meters and fuels of 1 000 h were counted along the sampling line (15 m), also, diameter was measured and determined whether firm (1 000F) or decomposed (1 000D). HO and HF layers were sampled in squares of 30 x 30 cm. Four squares distributed in a systematic way (2 x 2 m) were established, at each site, selecting the one with higher load and fallen leaves and fermentation layer are collected, and then transported to the laboratory to determine their dry weight. The dry weight was obtained with the use of an oven at a temperature of 90 °C until constant weight was reached.

Vegetation sampling. For this study the information from low tree was used, defined between 7.5 and 1 cm in diameter at height of 0.30 m (d_0.3) in order to assess species composition in the lower stratum of the stands and how it relates to necromass loads. To measure the latter, circular sites with 2 m radius as proposed by ^{Brown et al. (1981)} were established, where species, diameter at 0.30 m, height, and crown diameter of each individual were recorded (Figure 2).

Information analysis

The statistical analysis and graphics presented in this paper were performed using free software license R v3.0.1 (^{R Development Core Team, 2011}) and the routines were created under the software R-Studio v0.9 (^{RStudio, 2014}).

Necromass loads. Is expressed in megagrams per hectare (Mg ha^-1) and different formulas are used for the estimation by component. MLC is calculated through the formulas described by ^{Van Wagner (1982)} and ^{Morfin-Rios et al. (2012)}. To determine the loads of fallen leaves, extrapolation to hectare from the 30x30 cm square is used. The formulas to calculate MLC are described below:

Where: C= fuel load (Mg ha^-1); k = constant (1.234); DC= squared diameter; DCP= average squared diameter in cm (for 1, 10 and 100 h); GE= specific density or gravity (g/cm²); c= slope correction factor, and L= transect length (m).

Vegetation analysis

To observe the possible relationships between necromass loads and species composition was performed an analysis of importance values thereof. It took into account environmental density indicators (De), dominance (Do) and frequency (Fr) per species. The density is calculated through the extrapolation of the number of trees on site to the number of trees per hectare (Nha^-1). Dominance, as the density is calculated through the extrapolation of the basal area values (G) per site to hectare, per species. Finally, frequency is obtained by counting the number of sites where species appear (^{Dombois and Ellenberg, 1974}).

Ecological importance values for each species are calculated using De, Do and Fr data through their relative values by species, as shown in Table 1.

Table 1 Formulas used to determine the importance values of species. 

Species abundance analysis. To analyze species abundance, density data described above, and species abundance curves were used. The species abundance patterns can help as support to determine the status of ecosystems and these are usually described through abundance curves, where the species are arranged graphically, for its abundance (number of individuals present) from high to low. Depending on curve shape, one can determine if the ecosystem shows dominance of a particular species, or if the abundance is distributed evenly, i.e. species have similar number of individuals (^{Magurran, 1988}; ^{Verberk, 2011}).

In this study, species abundance curve was fit to known mathematical models. Currently there are many models used to describe species diversity in a community, however, this paper only discusses three of the best known: Pareto model, Poison model from the lognormal series and Neutral model from Alonso and Mckane (^{Prado et al., 2015}).

To select the best model the akaike information criterion (AIC) was used, according to which compares the selected models taking into account their fit and complexity; when models are compared using this method, the selection of the best model is based on the lower value in AIC. Moreover, the delta AIC criterion (dAIC), which, when is less than 2 indicate that compared models explain similarly trend data, i.e. there are no differences between each other, while values greater than 2 show a difference between these models. To determine the goodness of fit of the models the x²was used as recommended by ^{Magurran (1988)}.

Comparison between plots. The nonparametric test Mann- Whitney-Wilcoxon (^{Zar, 2010}) was used for statistical comparison between plots. Particularly compared necromass data (MLC, HO and HF) between plots and as data do not come from a normal distribution (KS- no significant) proceeded to perform the range sum test from Wilcoxon (nonparametric alternative for t test) for two independent samples.

As for the comparison of species between plots, density is used as parameter to compare statistically. The analysis was performed by comparing ratios with function "prop test" in R (^{Crawley, 2007}).

Necromass loads

For P1 a total of 36.6 Mg ha^-1 of biomass, comprised by MLC 1 h (0.96 Mg ha^-1), 10 h (6.81 Mg ha^-1), 100 h (3.93 Mg ha^-1), 1 000 F with 2.75 Mg ha^-1 and 1 000 D with 1.33 Mg ha^-1 further layers HO and HF with 8.52 and 12.25 Mg ha^-1 respectively. P2 recorded a total of 49.6 Mg ha^-1 of necromass. MLC 1 h recorded 0.74 Mg ha^-1, 10 h (5.05 Mg ha^-1), 100 h (46.54 Mg ha^-1), 1 000 F (1.99 Mg ha^-1), 1 000 D (3.39 Mg ha^-1), HO (8.64 Mg ha^-1) and HF with 23.26 Mg ha^-1 (Figure 3). These MLC loads can be considered in an intermediate range compared with data reported in the study conducted by ^{Xelhuantzi-Carmona et al. (2012)}, showing a maximum of 23 Mg ha^-1 in undisturbed forest jungle. However, ^{Morfin-Rios et al. (2007)} reported close to 80 Mg ha^-1 for montane forests.

Figure 3 Distribution of fuel load (Mg ha^-1) by component and plot, in the study area. 

For plots case only recorded statistically significant differences in the fermentation layer (p< 0.001). The comparison between other variables showed no such differences; however, this result supports the hypothesis that the higher superficial necromass load is in the stand that has no traces of fire (P2), because superficial fires help to reduce fuel loads (^{Agee, 1993}; ^{Rodriguez-Trejo, 1996}; ^{Jardel-Pelaez, 2010}).

Species composition

Regarding species composition, overall 12 families, 18 genera and 20 species were recorded. The most represented family was Fabaceae with 5 genera and 6 species, followed by the Rosaceae with 2 genera and 2 species, Juglandaceae and Rutaceae recorded only one genus but with 2 species, other families had one genus and one species (Table 2).

Table 2 Species from low tree, recorded in the study area, the "x" indicates presence in P1 and P2. 

Density and species abundance evaluation. The most abundant genus was Quercus with 2292 individuals in two hectares (Table 3 and 4); however, in plot 1 was where it recorded higher density and this difference was statistically significant (p< 0.001). Followed in density by Rhus virens with 2037 adding both plots, although it has most of its individuals in P2 (p< 0.001).

Table 3 Ecological importance values of species on plot 1 (P1). 

Donde: Nha-1= número de árboles por hectárea; G (m²)= área basal total en metros cuadrados por hectárea; IVI= índice de valores de importancia.

Table 4 Ecological importance values of species in plot 2 (P2). 

Donde: Nha-1= número de árboles por hectárea; G (m²)= área basal total en metros cuadrados por hectárea; IVI= índice de valores de importancia.

Amelanchier denticulata is the third species in terms of number of individuals with 1623 and the plot with the highest values was P1 (p< 0.001). Noteworthy, that Pinus pseudostrobus, species with adaptations to fire and intolerant to shade (^{Rodríguez-Trejo and Fulé, 2003}), showed statistically significant differences between plots (P1= 923, P2= 255, p< 0.001).

In addition, the species with fewer individuals were Acacia farnesiana, Acaciella angustissima, Amyris madrensis and Carya ovata with 32 individuals each in total (Tables 3 and 4).

Species abundance curves. Known models were fitted and for both plots and the best was the neutral model from Alonso and Mckane (Figure 4). For P1 the model that follows it, is Fisher (AIC= 96.41; dAIC= 0.52; x² = 25.7; p< 0.05), and finally the Poison model (AIC= 100.58; dAIC= 4.67; x²= 61.7; p< 0.001). As for P2, a similar response had to model fit. After the neutral model, follows Fisher model (AIC= 103.2; dAIC= 0.41, x²= 9.19; p= 0.81) and Poison model (AIC= 104.5; dAIC= 1.74; x²= 8.64; p= 0.85).

Figure 4 Species abundance model for both plots. 

In this analysis can be noted that both plots fit to neutral model; although the curves have similar distribution (Figure 4); that is, few species are the most abundant and regularly decreases as the number of species increases, it is important to consider that the most abundant species are different between plots (Tables 3 and 4).

Importance value index (IVI)

The most important species show a different behavior between plots, as seen in Tables 3 and 4. For plot 1, species with higher IVI values were Quercus sp., followed by Amelanchier denticulata and Pinus pseudostrobus, these three species represents 57% of total importance values. On the other hand, plot 2 has other species as the most important, Rhus virens, Quercus sp., Cercis canadensis and represents 58% of the importance values. A case worth mentioning is P. pseudostrobus, this species shows different importance values between P1 and P2, indicating a more active regeneration in the plot that has less necromass load (P1), on the other hand, species R. virens recorded high importance values in the plot with the highest necromass load (P2).

This could indicate that these species are directly related with frequency of fires. In the case of P. pseudostrobus could be said that they benefit from forest fires, and this also has to do with the fact that is a species intolerant to shade and with some features that make it resistant to forest fires (^{Rodríguez-Trejo and Fulé, 2003}). On the contrary, R. virens may be assumed that is related with the absence of forest fires, since the absence of fire on prolonged periods allows the accumulation of surface necromass, specifically the fermentation layer. This is particularly important, since a modification to frequency of fires could lead to changes in species composition of understory and one of the factors that have been proven, affects the frequency of forest fires, is climate change, through the alteration of fire regime.

^{Himmelsbach et al., (2012)} has worked in the area with the most important species to determine their tolerance to drought, the authors reported that P. pseudostrobus supports more short but high-intensity drought while other species, such as Quercus, support long low intensity drought. Which could have consequences on species composition in these ecosystems, as changes in drought regime, as demonstrated in the study by ^{Yocom et al., (2010)} could stress, specifically, plants making them more susceptible to natural disturbances such as forest fires.