Study area

The study included the forest regions Bosque La Primavera, Sierra de Quila and Sierra de Tapalpa of Jalisco, Mexico (Table 1), characterized as temperate forests with sub-humid climate and summer rainfall (^{Comisión Nacional del Agua [CONAGUA], 2020}). The regions differ in slope, elevation and exposure, but have similar type of soil (euric regosol) and coniferous vegetation, with high fire occurrence (^{Huerta & Ibarra, 2014}).

Previous studies in these regions indicate that the frequency of forest fires has increased in recent years, mainly in Bosque La Primavera and Sierra de Tapalpa with 165 fires∙year^-1 and, to a lesser extent, in Sierra de Quila with 10 fires∙year^-1 (^{Flores-Garnica, Flores-Rodríguez, Lomelí-Zavala, Ruíz-Guzmán, & García-Bernal, 2019}). A heterogeneous mosaic of patches with levels of wildfire severity has been generated due to the high frequency of fires, and topographic, climatic, and vegetation variation. Sites with low severity show very little burn damage; those with moderate severity have burned trees in the tree stratum and small fractions of crowns or understory; and in sites with extreme severity, the tree, shrub and herbaceous stratum are completely burned (^{Lloret, 2004}). Based on these severity levels, sites with extreme fire and moderate fire were selected in the three study regions, and intact sites that had not been damaged by fire in recent decades were used as control.

Experimental design

The study was based on a factorial experimental design composed of two factors: (a) regions (Bosque La Primavera and Sierra de Quila and Sierra de Tapalpa) and (b) severity levels (no fire, moderate fire and extreme fire); the combination of these gave a total of nine treatments. Within each of the severities per region, three sites were randomly selected, which were considered as replicates, resulting in 27 sampling sites. The sampling unit was circular (400 m²), where all individuals larger than 7.5 cm diameter at breast height (at 1.3 m above ground level) were evaluated. The following data were recorded: taxonomic identity (genus, species and common name), and basal area and total height using a diameter tape (Forestry Suppliers Inc. 283d) and a laser hypsometer (Forestry Pro Nikon 8381), respectively.

Data recording and analysis

Each region was characterized by its composition, horizontal structure, diameter class, diversity and similarity of tree species. Composition was obtained by collecting botanical material from the field of all tree species in sampling sites. Identification was done by taxonomic keys, while the validity of scientific names was corroborated on ^{The Plant List (2013)} platform. The horizontal structure of the species was described by abundance, defined as the number of trees; dominance according to the basal area; and frequency was determined by presence at sampling sites (^{Alanís-Rodríguez et al., 2020}). The data obtained were used to calculate the importance value index (IVI), which acquires percentage values.

Abundance was calculated according to the number of trees using the following formulas:

where,

A _i  = absolute abundance (individuals∙ha^-1)

AR _i = relative abundance of species i in relation to total abundance (%)

N _i  = number of individuals of species i

S = sampling area (ha).

Relative dominance was calculated according to the basal area with the following formulas:

where,

D _i = absolute dominance (m²∙ha^-1)

DR _i = relative dominance of species i in relation to dominance (%)

Ab _i = basal area of species i (m²)

S = area (ha).

Relative frequency was obtained with the following formulas:

where,

F _i = absolute frequency

FR _i = relative frequency of species i in relation to the sum of frequencies (%)

f _i = number of sites where species i is present

NS = total number of sampling sites.

The importance value index (IVI) was estimated as:

where,

AR _i = relative abundance of species i in relation to total abundance

DR _i = relative dominance of species i in relation to total dominance

FR _i = relative frequency of species i in relation to total frequency.

Histograms of diametric distributions were created according to the number of trees recorded for the three study regions; equations and R² were calculated for each graph. The analyses were performed using the statistical program PAST version 3.2 (^{Hammer, Harper, & Ryan, 2001}).

Diversity and richness indices were determined using the Shannon's index (H'), which estimates habitat heterogeneity by number of species present and relative abundance; Margalef's species richness (Dmg), which evaluates the biodiversity of a community based on numerical distribution of individuals, combining the number of species and number of individuals; and Simpson's diversity (D), which determines whether the community is composed of very abundant species (^{Magurran, 2004}). These indices were calculated using the following equations:

where,

S = number of species present

Pi = number of individuals of species i

N = total number of individuals

P _i = proportion of species i in the community (ni/N)

ni = number of individuals of species i

To determine statistically significant differences among treatments, a two-factor analysis of variance (ANOVA) was performed between fire severity levels and study regions. Data were checked for compliance with normality assumptions using the Shapiro-Wilk test. Significant differences were corroborated with the Tukey tests (α = 0.05) using ^{GraphPad Prism (2019)} version 8.2.1.

The similarity in the composition of the three regions was determined based on the abundance of tree species using a ^{Bray-Curtis (1957}) dendrogram, which reflects the percentage of similarity or dissimilarity of the regions. The analyses were performed using PAST version 3.2 (^{Hammer et al., 2001}).

Composition

A total of 573 trees belonging to 12 species distributed in six families were recorded, which are shown in Table 2. According to Table 3, the Pinaceae and Fagaceae families were the most represented (86.2 %) of the tree composition in the three regions, while the remaining 13.7 % corresponded to the species Arctostaphylos pungens Kunth, Bursera penicillata (Sessé & Moc. ex DC.) Engl., Crataegus mexicana Moc. & Sessé ex DC. and Fraxinus uhdei (Wenz.) Lingelsh. The tree composition of the three study regions corresponds to temperate forests typical of mountainous areas of Mexico (^{De León, García, Andrade, & Ruíz, 2013}; ^{Graciano-Ávila et al., 2020}). The highest number of individuals in the fire regions/levels belong to the genera Pinus and Quercus in the sites with moderate and extreme fire (54 to 89 individuals). This coincides with ^{Juárez-Martínez and Rodríguez-Trejo (2003}), who mention that burned areas show greater natural regeneration than non-fired areas, because fire causes the death of trees and, consequently, facilitate the opening of the canopy and favors the establishment of pioneer species. Authors such as ^{Graciano-Ávila et al. (2017)} point out that the abundance of Pinus and Quercus is due to their wide distribution in mountain ranges, but also to their high diversity, as they are genera, which host most of the species of their respective families (Pinus 50 and Quercus 161). Although the abundance of these genera is mainly due to climatic affinity, it may also be due to irregular forest management, which leads to conglomeration of small groups of species at the local level, due to the extraction of other species of greater economic value (^{López-Hernández et al., 2017}).

Horizontal structure

Tables 4, 5 and 6 include the structural parameters of species in the three study forest regions. The Pinaceae family had the highest abundance of individuals regardless of fire severity levels. This agrees with that reported by ^{Hernández-Salas et al. (2018}), who found higher abundance of species of the genus Pinus in temperate forests of Chihuahua. In Bosque La Primavera, only the Fagaceae and Pinaceae families were present; specially Q. magnoliifolia (39 individuals, 65 %) and P. douglasiana (39 individuals, 54.93 %) were the most abundant. In Sierra de Quila, Q. obtusata was the most abundant with 42.5 %, while in Sierra de Tapalpa, P. devoniana was the most abundant (66.2 %).

In moderate severity sites, the most dominant species were P. douglasiana (49.3 to 53.9 %) and P. devoniana (88.4 %), while Q. magnoliifolia (51.1 %) and P. devoniana (88.2 %) were dominant in extreme severity sites (Tables 4, 5 and 6). The dominance values ranged from 0.2 to 50 m²∙ha^-1 in sites of moderate and extreme severity; this is the result of recurrent forest fires that have led to changes in growth, but also in slope and orientation, as well as in the topoforms (valleys and hills) where they develop (^{Martínez-Antúnez et al., 2013}).

Authors such as ^{Guzmán (2009}) mention that an IVI greater than 50 % represents the ecological dominance of a taxon. In this study, IVI values in La Primavera and Quila were less than 50 %, which would indicate greater species competition for resources and less dominance of post-fire taxa; however, Sierra de Tapalpa had the highest IVI values for the Pinaceae family, with P. devoniana (60 to 70 %) being dominant in sites with and no fire. Despite the high IVI of this species, the value was lower than that reported (80 %) by ^{Alanís-Rodríguez et al. (2011}) in sites affected by fire involving pine species.

Diameter class

Figure 1 shows that diameter class distribution had an inverted J trend, i.e., more individuals in smaller diameter class. The Bosque La Primavera recorded a greater number of individuals in 20 cm (n = 59) and 30 cm (n = 56) diameter class in moderate fire severities. In the case of Sierra de Quila, the areas with moderate and extreme severity had a greater number of individuals (n = 21 and 22), only in the 20 cm diameter class. In contrast, in the Sierra de Tapalpa, the 10 cm diameter class had the highest number of individuals (n = 40) in the extreme fire severity level. In general, the dominance of individuals in regrowth or juveniles with diameters of 10 to 30 cm was greater in sites with moderate and extreme severity, which agrees with ^{Rodríguez, Mata, Moya, and Guzmán (2003}) and ^{Quintero-Gradilla et al. (2019}) for temperate forests affected by fire in the Sierra de Monterrey and Sierra de Manantlán, Jalisco. The results indicate that the post-fire forests are growing with active regeneration, since more than 50 % of the individuals were grouped in three diameter classes, which would guarantee the survival of the ecosystem by having a reserve of young trees (^{Manzanilla-Quijada et al., 2020}).

Figure 1 Diameter class of tree species recorded in the study regions (Bosque La Primavera [A], Sierra de Quila [B] and Sierra de Tapalpa [C]) with three levels of forest fire severity (no fire, moderate and extreme). Bars indicate the mean ± standard error of three replicates (sites). Dotted lines show the trend of an inverted J; that is, more individuals in smaller diameter class. 

Diversity and richness indices

The specific richness in this study was 12 species, a lower result than that reported by ^{Mora-Donjuán et al. (2013}), who recorded 20 species in post-fire forests. Unlike the above-mentioned study, the present study evaluated contrasting fire severities, so species mortality increases according to a higher severity (^{Neris et al., 2016}).

ANOVA indicated that the model was significant; however, the Tukey's test results showed that seven of the nine possible interactions were not statistically different (F = 165.4, P > 0.05). Severity level and region are analyzed separately in Figure 2. Diversity and richness indices were significantly different among severity levels (F = 106.6, P ≤ 0.001) (Figure 2A). Shannon diversity was higher (F = 106.6, P ≤ 0.001) for the site with moderate level (H' = 1.83 ± 1.73). Diversity was higher than that reported by ^{Alanís et al. (2008)} and ^{García-García, Alanís-Rodríguez, Aguirre-Calderón, Treviño-Garza, and Graciano-Ávila (2020}) with values of H'= 1.6 and H'= 0.64 to 1.6, respectively, in post-fire sites. Margalef's richness index was also significantly different (F = 2.08, P ≤ 0.001) with a D _Mg = 2.01 ± 1.88 for moderate sites; however, it is lower than the D _Mg = 3.16 value reported by ^{Graciano-Ávila, Alanís-Rodríguez, Aguirre-Calderón, Rubio-Camacho, and González-Tagle (2018)}. Regarding the Simpson's index, the D value = 2.02 ± 1.88 for the site with moderate severity (F = 2.08, P ≤ 0.001) was higher than that reported by ^{Quintero-Gradilla et al. (2019}) in their study on post-fire stand chronosequence of 8, 28, and 60 years with values of D = 0.09.

When analyzing fire levels of evaluated regions, differences were found in diversity and richness indices in Sierra de Quila (F = 33.26, P ≤ 0.001; Figure 2C) and in Sierra de Tapalpa (F = 9.97, P ≤ 0.02; Figure 2D), but not in Bosque La Primavera (P ≥ 0.05) (Figure 2B). The increase in diversity and richness in sites with moderate fires is related to the intermediate disturbance hypothesis, where there are favorable conditions for species and their biotic interactions (competition), which promotes greater diversity, due to the generation of ecological niches that avoid the competitive exclusion of species (^{Huston, 2014}). In addition, physicochemical properties of the soil, understory vegetation and canopy cover are not substantially modified for moderate fire severities, which favors the survival and development of species (^{Cadena-Zamudio, Flores-Garnica, Flores-Rodríguez, & Lomelí-Zavala, 2020}).

Finally, diversity indices between study regions (Figure 2E) were similar (P ≥ 0.05). Shannon's diversity indices H' =1.82 ± 1.72 and Margalef richness D _Mg = 1.57 ± 1.45 in Sierra de Quila were higher than those reported by ^{Buendía et al. (2019)} of H' = 0.83 to 1.78 and ^{Hernández-Salas et al. (2013}) with D _Mg = 0.90 to 1.04 for temperate forest in Mexico.

Figure 2 Diversity and richness indices per forest fire severity (A); fire severity levels in Bosque La Primavera (B), Sierra de Quila (C), Sierra de Tapalpa (D); and in the study regions (E). In each index, different letters denote means (± standard error) significantly different according to the Tukey's test (P < 0.05) among severity level or among regions. 

Similarity coefficient

The grouping analysis (Bray-Curtis similarity coefficient) showed the formation of an isolated entity and a well-defined group (Figure 3). The first corresponded to Sierra de Tapalpa with 17 %; on the other hand, the group formed by Bosque La Primavera and Sierra de Quila had 33 % similarity, sharing the species Q. obtusata, Q. resinosa and P. douglasiana, which corresponds to an oak-pine plant community. The results of this study agree with ^{Domínguez-Gómez et al. (2018}), who found similarity greater than 30 % in their evaluation sites. The similarity of Bosque La Primavera and Sierra de Quila are influenced by the number of individuals of taxa with higher composition, mainly in sites with moderate and extreme fire (Table 3), but also by physiographic characteristics such as altitudinal affinity (1 500 to 2 000 m) (^{Santiago-Pérez, Ayón, Rosas-Espinoza, Rodríguez, & Toledo, 2014}). The low similarity of the Tapalpa region may be due to the characteristics of the area such as altitude, presence of hills and slopes (Table 1).

Figure 3 Bray-Curtis similarity coefficient of three forest regions in Jalisco, Mexico.