Small-fragment, high turnover: soil microenvironment fluctuation effect on tree diversity in a Neotropical montane oak forest

Sampling site

The study was performed in the Medio Monte Natural Protected Area (San Bartolo Tutotepec, Hidalgo state, Mexico; 20°24′51.50″N, 98°15′32.07″W; 1,800–1,944 m.a.s.l.; Fig. 1A). The study forest is the largest continuous relict-NMoF in eastern Mexico (ca. 151.63 ha; SEMARNAT & Secretaría del Medio Ambiente y Recursos Naturales, 2010) with rugged terrain with steep slopes >40°. The climate is temperate (Cwb sensu Peel, Finlayson & Mcmahon (2007)) with three distinct seasons throughout the year: dry cool October-January, dry warm early February-May, and humid June-September. Average annual temperature 12.7−14 °C (Rodríguez-Ramírez et al., 2021), the total annual precipitation is 1,200–2,015 mm, and relative moisture is 60–85% (Miranda & Sharp, 1950). The buried soil is covered by younger deposits (Andosol-Humic; IUSS Working Group, 2022). The study forest displays pH values of 4–6 (Peters, 1995).

Study transects and tree inventory

During 2019 from January to December, we established and geo-referenced (with a GPS; Garmin eTrex^® 10; Garmin Ltd., Olathe, KS, USA) four permanent transects (as far as possible from anthropic activities and roads) with a length of 100 m and ∼300 m apart in the study forest (Fig. 1B). To identify the structural importance of tree species, at each transect five sample plots (20 × 20 m) were set up and 20 m apart to assure data independence (Roberts-Pichette & Gillespie, 1999). Conspicuous tree species with a diameter at breast height (DBH) ≥ 1.5 cm were collected, identified and quantified every three months (Rodríguez-Ramírez, Sánchez-González & Ángeles Pérez, 2018). Tree species were identified using specific taxonomic keys (Rzedowski, 2015) and the vascular plant species nomenclature was updated at the Tropicos website (Missouri Botanical Garden; http://www.tropicos.org). The specimens were deposited in the FES–Iztacala–UNAM herbarium, Mexico (IZTA; https://www.zaragoza.unam.mx/herbario/).

Microenvironmental variables

We measured five microenvironmental variables tri-monthly in each plot between 8:00 and 17:00 to normalize the daily fluctuation (Table 1): soil moisture, soil temperature, soil pH, depth litterfall, and light incidence (see Rodríguez-Ramírez, Sánchez-González & Ángeles Pérez, 2018 for more details). Four measurements were taken at the four cardinal points (e.g., north, south, east, and west) of each plot. The values were averaged for statistical analysis.

Analyses vegetation

We assessed the Importance Value Index (IVI; express the dominance and biological success of specific species; Susilowati et al., 2020) of each tree species and performed rank-abundance curves to identify the structure roll of each tree species among transects, estimated by summing relative frequency, density and dominance (Naidu & Kumar, 2016):

IVI = ($\frac{1}{2}$) relative BA + relative D.

We calculated basal area (BA) using the formula BA = πr², where π = 3.1416 and r = (0.5) (DBH) is the radius of the tree; and relative density (D) was determined using the number of individuals of each species (Mostacedo & Fredericksen, 2000). To characterize the tree structure and composition of each transect, we developed semi-realistic Richards profile diagrams using the software Adobe Illustrator v. 23.0.5 (http://www.adobe.com).

Diversity analyses

To ensure that our tree survey effort was enough to record a significant sample for the inventory, we used sample coverage estimates suggested by Chao & Jost (2012), ranging from 0 (minimum) to 100% (maximum). Tree diversity was estimated using q = 0 (species richness) and q = 1 (species diversity) (Hill, 1973). We performed this estimation among transects. The α-diversity, for a given diversity order q, is given by: ${\displaystyle q_D=\left(\begin{array}{cc}\sum _{i=1}^s\hfill & p_i^q\hfill \end{array}\right)1/\left(1-q\right)}$

where S is the number of species, pi is the comparative abundance of species i, and q is the order number of diversity. We performed rarefaction curves to standardize samples that differ in terms of individual size or plot size (Chao, KH & Hsieh, 2016). They were compared with 95% confidence intervals. We performed the analysis using the iNEXT R-library (Chao et al., 2014; https://chao.shinyapps.io/iNEXTOnline/).

We partitioned β-diversity according to the procedure of Podani & Schmera (2011), which is based on the approach of Baselga (2010). According to this method, total dissimilarity (βcc) is 1-minus Jaccard similarity coefficient. βcc is divided into two components: turnover (β.3, replacement of one species by another or loss of species) and nestedness (βrich, differences in species richness). The β-diversity was assessed using the IVI of each tree species (Naidu & Kumar, 2016). These analyses were performed in R-software v. 2.4.1 (R Core Team, 2018), using the R-script of Carvalho, Cardoso & Gomes (2012).

Linking microenvironmental variables with turnover tree diversity

We performed a principal component analysis (PCA) using tree abundance data to identify the most strongly discriminating soil microenvironmental variables (Anderson et al., 2011; Kent, 2011). All microenvironmental variables were first transformed to log10 after adding a constant because each variable showed different value types (Kindt & Coe, 2005). We performed a redundancy analysis (RDA; Borcard, Gillet & Legendre, 2011) to analyze the relationship between turnover tree diversity (IVI data) and microenvironmental variables among four transects. We intend the technique to reveal major trends in the variation of a multidimensional dataset in a reduced space of selected, linearly independent dimensions (Legendre & Legendre, 1998). We achieved the analyses with the R-software and vegan package (Oksanen et al., 2016).

Based on the IVI tree species rank-abundance curves results, we performed a quasi-Poisson generalized linear model (GLM; Ver Hoef & Boveng, 2007) to determine specific microenvironmental variables that could influence on turnover of tree species with high IVI values; the tree species as dependent variables, and microenvironmental variables as independent factors. Specific predictor microenvironmental variables were log or square root transformed to address non-normality (Shapiro-Wilk test; Hanusz, Tarasinska & Zielinski, 2014). In all cases, we selected the best model fitted for each microenvironmental variable; we used the Adjusted Akaike’s information criteria (AICc) selecting the best model with the minimum AICc value (Hurvich & Tsai, 1989). Statistically significant variables are P < 0.05 and P < 0.01 (Borcard, Gillet & Legendre, 2011). We achieved all GLM analyses through using R-software with the glm2-function in stats package. Besides, response curves of six tree species with high IVI values against specific soil microenvironmental variables were performed with CANOCO v. 5.0 (Šmilauer & Lepš, 2014).

Floristic composition

A total of 22 tree species were identified in the four transects (Table 2). The vegetation profiles allowed to detect dissimilarity floristic species structure and composition among transects (Fig. 2). The average tree diameter (DBH) in the transect T1 was 89.8 cm; T2 = 114.5 cm; T3 = 93.9 cm; and T4 = 97.9 cm (Fig. 2). Overall, the Importance Value Index (IVI) showed ranges from 0.333 (Prunus serotina Ehrh.) to 0.005 (Conostegia arborea Steud.) (Table 2, Fig. 3). By each transect the rank-curve showed different vegetation composition and structure dominated by: T1, Mexican beech (IVI = 0.252); for T2 and T3, Mexican Clethra (IVI = 0.192, 0.313 respectively); and T4, Wild black cherry (IVI = 0.333), (Fig. 3). The uncommon tree species were: for T1, Sweetgum (Liquidambar styraciflua L.; IVI = 0.018); for T2 and T4, Capulín (IVI = 0.005); and T3, Quebracho (Quercus insignis M. Martens & Galeotti; IVI = 0.017), (Fig. 3).

Tree diversity

We obtained tree diversity completeness >90% in each transect (Table 3). The tree α-diversity values (q = 0, tree species richness; and q = 1, ecological diversity) and rarefaction curves were not significantly different among transects (Table 3, Fig. 4). The T4 showed the highest richness (17 species) and the lowest richness values were registered in T1 and T3 (12 species), (Fig. 4A). The highest tree ecological diversity was found in T2 (10.18) and the lowest value was detected in T4 (7.3), (Fig. 4B).

β-diversity

Overall, the total average tree β-diversity (βcc) among transects was ∼0.55; turnover (β.3) ∼0.36 and nestedness (βrich) ∼0.18. Considering all pairs of comparisons of all sampled transects, partitioning total β-diversity, we found that between T1-T3 and T2-T4 were explained by turnover (β.3 = 0.70 and 0.54 respectively) and null nestedness value for both pairs (βrich = 0.00). Similarly, the relationship between T1-T4 was explained by turnover (β.3 = 0.40) and low nestedness (βrich = 0.25). The pairs that showed similar values of turnover and nestedness were between T1-T2 and T2-T3 (β.3 = 0.22; βrich = 0.27). The lowest turnover and nestedness values were between T3-T4 (β.3 = 0.11; βrich = 0.29), (Fig. 5). Each transect showed unique tree species composition such as Mexican beech (Fagus mexicana, T1), Sweetgum (Liquidambar styraciflua, T1), Pezma (Cyathea fulva (M. Martens & Galeotti) Fée, T2), Palo de agua (Perrottetia ovata Hemsl., T4), Wild avocado (Persea americana, T2), Palo verde (Turpinia insignis (Kunth) Tul., T4) and Querbrachos (Quercus delgadoana, T2; Q. meavei, T4; and Q, trinitatis, T4).

Local microenvironmental variables effect on tree diversity

Microenvironmental values were not differing significantly among transects (Fig. 6); notwithstanding, we got the most contrasting light incidence values where T1 recorded ∼98% and T4 ∼77% (Fig. 6D); and we recorded the opposite tendency in pH values among transects (from 6.3 to 7; Fig. 6E).

The first four PCA axes explained 80% of the microenvironmental variation. Moreover, the global permutation test of redundancy analysis ordination (RDA) displayed that the relationship between tree species and microenvironmental variables (i.e., soil moisture, soil temperature, litterfall depth, light incidence, and pH) were significant (P < 0.01). The first two axes explained 95% of the variance in tree species and between tree species and soil microenvironmental relationship (Table 4; Fig. 7). Soil moisture influences tree composition in the T1, soil temperature on tree composition in the T4, and light incidence on tree composition in the T3 (Fig. 7). Quasi-Poisson generalized linear model analysis (GLM) showed that soil microenvironmental variables significantly influenced specific tree species; Mexican beech (Fagus mexicana), white Quebracho (Quercus delgadoana), Pezma (Cyathea fulva) and Aguacatillo (Beilschmiedia mexicana) were influenced by soil moisture and soil temperature; while Mountain magnolia (Magnolia schiedeana) and Pezma (Dicksonia sellowiana var. arachneosa (Sodiro)) were influenced by the light incidence (Table 4; Fig. 8).