Response of tree diversity and community composition to forest use intensity along a tropical elevational gradient

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2019 The Authors. Applied Vegetation Science published by John Wiley & Sons Ltd on behalf of International Association for Vegetation Science. 1Biodiversity, Macroecology and Biogeography, University of Goettingen, Göttingen, Germany 2Centro de Investigaciones Tropicales, Universidad Veracruzana, Xalapa, Veracruz, Mexico 3Instituto de Ecología A.C., Xalapa, Veracruz, Mexico 4German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany 5Institute of Biology, Leipzig University, Leipzig, Germany 6Centre of Biodiversity and Sustainable Land Use (CBL), University of Goettingen, Göttingen, Germany


| INTRODUC TI ON
Tropical mountains are characterized by steep gradients in climate and other environmental conditions that lead to rapid changes in diversity and species composition with elevation. Globally, tropical mountains are centers of plant diversity and endemism (Barthlott, Mutke, Rafiqpoor, Kier, & Kreft, 2005;Kier et al., 2009) resulting from the high environmental heterogeneity that affects both ecological and evolutionary processes (Antonelli et al., 2018;Stein, Gerstner, & Kreft, 2014). Tropical elevational gradients are considered natural laboratories where drivers of diversity patterns and ecosystem functions can be studied over short geographical distances (Körner et al., 2017;Sanders & Rahbek, 2012). However, tropical mountain ecosystems are also highly vulnerable to land use (Kidane, Stahlmann, & Beierkuhnlein, 2012;Malhi et al., 2010) and climate change (Cuesta et al., 2017), yet our understanding of how anthropogenic change may affect plant diversity and community composition along elevational gradients remains limited (Peters et al., 2019).
Explanations for the elevational gradients in tree diversity have focused principally on temperature and precipitation, soil nutrient concentrations, the mixing of biotas, spatial constraints associated with area and mid-domain effects, and -to a lesser extent -anthropogenic disturbances (Homeier et al., 2010;Peters et al., 2019;Rana, Gross, & Price, 2019;Slik et al., 2019;Toledo-Garibaldi & Williams-Linera, 2005;Zhang, Xu, & Li, 2013). The effect of land use type and intensity on species richness and composition might change along elevational gradients (McCain & Grytnes, 2010), as it might be amplified or weakened by climate changing with elevation (Peters et al., 2019). For example, if the impact of land use on diversity is higher at lower elevations than at higher elevations, species richness patterns could shift from monotonic to hump-shaped. Conversely, if the impact of land use on diversity is stronger at mid-elevations than at lower ones, species richness patterns could change from humpshaped to monotonic.
Forest use intensity, from here on in defined as the conversion of (near-) natural, complex structured forest ecosystems to simplified, managed ecosystems with more frequent resource use or extraction (Nepstad, Uhl, Pereira, da Silva, & da Silva, 1996;Tscharntke, Klein, Kruess, Steffan-Dewenter, & Thies, 2013;Vitousek, Mooney, Lubchenco, & Melillo, 1997), may also influence the composition of forest communities by altering environmental conditions. In tropical forest ecosystems, highly intensive forest uses, such as cattle grazing and agroforestry, increase light availability and air temperature, reduce air humidity and soil moisture and have negative effects on propagule dispersion (Holl, 1999;Lebrija-Trejos, Pérez-García, Meave, Poorter, & Bongers, 2011). High forest use intensity may shift tree species composition to forests dominated by species better adapted to tolerate such conditions, e.g. fast-growing and light-demanding pioneer species, whereas more shade-tolerant late-successional species are often unable to persist (Craven, Hall, Berlyn, Ashton, & Breugel, 2015;Lohbeck et al., 2013). As a result of high-intensity forest use, light conditions increase and favor the establishment of early successional tree species. Therefore, floristic composition -and to a lesser extent species richness -of young secondary and degraded forests usually differs markedly from that of old-growth tropical forests (Gossner et al., 2013;Rozendaal et al., 2019). Yet, how such changes in species composition are mediated by climatic changes along elevational gradients is largely unknown (but see Peters et al., 2019).
Human impact on tropical mountain forests, such as logging and deforestation for agriculture, has transformed large parts of these ecosystems into human-dominated forested landscapes (Laurance, Sayer, & Cassman, 2014). Globally, the direction and magnitude of changes in species richness depend strongly on the kind, intensity, severity, incidence, and timing of disturbances (Barlow et al., 2018;Foley et al., 2005;Gibson et al., 2011;Newbold et al., 2015). Yet, we know little about the impacts of forest use intensity on tree diversity and composition along environmental gradients, particularly elevational gradients in tropical forests. While it is likely that these impacts will shift because tree communities differ in their resilience to similar forest uses (Crouzeilles et al., 2016), there is no a priori expectation whether the impacts will be stronger, weaker, or similar with changes in elevation.
Here, we examined how the interaction of elevation and forest use intensity affects tree diversity and community composition along an elevation gradient from sea level to treeline within a global biodiversity hotspot in central Veracruz, Mexico. Specifically, we asked: (a) how do tree diversity and community composition vary with elevation;and (b) how do the effects of forest use intensity on tree diversity and community composition change within elevational sites? We hypothesized that tree diversity monotonically decreases with increasing elevation (Aiba & Kitayama, 1999;Homeier et al., 2010;Slik et al., 2019;Toledo-Garibaldi & Williams-Linera, 2005), that high forest use intensity at lower elevations may shift species richness from monotonic to a hump-shaped pattern (McCain & Grytnes, 2010) and expected a consistently negative effect of forest use intensity on tree diversity and associated shifts in tree species composition (Gibson et al., 2011;Newbold et al., 2015).

| Study area
Our study was conducted along an elevational gradient, from sea level close to the Gulf of Mexico (19.5894 N,  tropical-dry at lower elevations, to temperate-humid at mid-elevations and cold-dry at high elevations (Gómez-Díaz et al., 2017;Soto Esparza & Giddings Berger, 2011). Temperature decreases with elevation, with mean annual temperature ranging from 26°C near sea level to 9°C at the highest site. Mean annual precipitation varies from 1,222 mm at low elevations, 2,952 mm at mid-elevations and 708 mm at high elevations ( Table 1).
The study area is located in the transition zone between two biogeographic regions, the Neotropical and Nearctic, in the Mesoamerican biodiversity hotspot (Morrone, 2006;Myers, Mittermeier, Mittermeier, Fonseca, & Kent, 2000). Biogeographically, the upper part of the elevational gradient falls into the convergence zone between the Trans-Mexican Volcanic Belt and the Sierra Madre Oriental (Rodríguez, Morales-Barrera, Layer, & González-Mercado, 2010

| Study design and data collection
The study was conducted at eight sites along the elevational gradient, separated by about 500 m in elevation ( Figure 1). Hereafter, we refer to every elevation as 0, 500, 1,000, 1,500, 2,000, 2,500, 3,000, and 3,500 m. At each site, we established 15 plots of 20 m × 20 m, including five plots each in old-growth, degraded, and secondary forests. In total, 120 non-permanent forest plots (4.8 ha) were inventoried. Forest use intensity was defined following Leuschner, Wiens, Harteveld, Hertel, and Tjitrosemito (2006), Carvajal-Hernández and Krömer (2015)  after clearcutting, sometimes with cattle grazing, with small diameter trees, classified as high forest use intensity. In each plot, we measured and identified all trees with a diameter at breast height (DHB) ≥5 cm (Homeier et al., 2010;Toledo-Garibaldi & Williams-Linera, 2005

| Analysis of tree diversity
We estimated species diversity as species richness (Hill number q = 0), Shannon diversity (q = 1), and Simpson diversity (q = 2) in terms of effective species numbers (Jost, 2006). These diversity indices give increasing weight to species abundances; while species richness gives equal weight to common and rare species, Shannon and Simpson diversities emphasize the contributions of common and abundant species, respectively. For each diversity index, we estimated species accumulation curves using sample-based rarefaction and extrapolation (Chao et al., 2014), pooling data by forest use intensity for each elevation. We used rarefaction and extrapolation because the number of individuals may vary systematically with forest use intensity, which may bias estimates of species diversity in plots with more individuals. As we observed similar patterns for Shannon and Simpson diversities, we only present and discuss results for species richness. At the plot level, we estimated species diversity using a fixed sample coverage of 95% with the 'iNeXT' package (Hsieh, Ma, & Chao, 2016) to permit unbiased comparisons of species diversity across forest use intensities and elevations. To meet model assumptions of normality, we natural-log-transformed all diversity indices.
At the plot level, we examined the effect of forest use intensity at each elevation on species richness and Shannon and Simpson diversities by a nested analysis of variance (ANOVA) using the R function aov, where forest use intensity and elevation were treated as categorical variables. We performed post-hoc comparisons using a Tukey's Honest Significant Differences test with the packages 'car' (Fox & Weisberg, 2011) and 'multcomp' (Hothorn, Bretz, & Westfall, 2008).
We examined changes in tree community composition among elevational sites and levels of forest use intensity using non-metric multidimensional scaling (NMDS) with both incidence-based Jaccard dissimilarity and abundance-based Bray-Curtis dissimilarity (adjustment noshare = 0.1; 999 permutations) using the packages ecodist (Goslee & Urban, 2007)

| Changes in tree diversity along the elevation gradient
We recorded a total of 4,555 individual trees belonging to 217 species distributed among 80 families and 154 genera (Appendix S1).
The most diverse families were Fagaceae (15 species Alnus. Across all elevations, 18% of all species exclusively occurred in old-growth forest followed by 16% in secondary forest, and 9% in degraded forest. At most elevations (500, 1,000, 2,000, 3,000 and 3,500 m) species accumulation curves overlapped across forest use intensities, revealing that species pools of degraded and secondary forests were of similar size (indicated by overlapping 95% confidence intervals) as those of old-growth forests (Figure 2). In contrast, at 0 and 1,500 m species accumulation curves showed significant differences in species richness between forest use intensities (95% confidence intervals did not overlap). Species accumulation curves at 2,500 m showed that the tree species richness in degraded forests was higher than the species richness of old-growth forests (Figure 2). For most levels of forest use intensity, species accumulation curves did not reach an asymptote at elevations between 0 and 2,500 m (except for secondary forests at 0 m). At 3,000 and 3,500 m, species accumulation curves for all levels  Leopold (1950); MAT, mean annual temperature (°C); MAP, mean annual precipitation (mm/a). Climate data were obtained from National Meteorological Service of México (SMN, 2019) for 1951-2010.

MONGE-GONZÁLEZ Et aL.
of forest use intensity reached an asymptote, indicating that these forests have been fully sampled (Figure 2). Pairwise comparisons among forest use intensity levels within elevations showed significant differences in tree species richness between old-growth and secondary forests at 0, 500, 1,500 and 3,500 m (p-value < 0.05; Appendix S3; Figure 3). While these differences were associated with higher species richness in oldgrowth forests than in secondary forests at 0, 500 and 1,500 m, the inverse pattern was observed at 3,500 m, i.e., higher species richness in secondary forests than in old-growth forests. Similarly, we found significant differences in tree species richness between degraded and secondary forests at 1,500, 2,500 and 3,500 m (pvalue < 0.05; Appendix S3), with higher tree species richness in degraded than in secondary forests at 1,500 and 2,500 m but the inverse trend at 3,500 m.

| Effects of forest use intensity on local tree species diversity along the elevation gradient
In old-growth forests, tree species richness along the elevational gradient was best described as a low-plateau pattern, where species richness was highest from 0 to 2000 m, after which it decreased monotonically ( Figure 3). Tree species richness in degraded and secondary forests showed a bimodal pattern with peaks at 1,000 and 2,000 m, declining towards both ends of the elevational gradient ( Figure 3). Similar patterns were observed for Shannon and Simpson diversities (Appendices S8, S9).

| Tree community composition among gradients of forest use intensity and elevation
We found that tree community composition varied significantly   Figure 4; Appendices S6, S7, S10). Tree community composition varied between degraded and secondary forests in most elevations except at 2,000 and 3,000 m based on Jaccard dissimilarity and at 500, 2,000, and 3,000 m based on Bray-Curtis dissimilarity (Appendices S6, S7, S10).

| D ISCUSS I ON
Our results revealed that the interaction between elevation and forest use intensity affected tree diversity, i.e. species richness, Shannon and Simpson diversity as well as community composition.
Importantly, we found that the effects of forest use intensity on tree diversity were not consistent along the elevational gradient, with tree diversity decreasing significantly in secondary compared with old-growth forests at only three elevations (0, 500 and 1,500 m).
Together, our results suggest that the direction and magnitude of the effects of anthropogenic forest disturbance on tree diversity in tropical forests are context -dependent and will be difficult to generalize more broadly.

| Forest use intensity affects tree diversity along the elevational gradient
Forest use intensity affected tree diversity, and we found signifi- Old-growth and degraded forests exhibited similar plot-level species diversity at all elevations, a pattern that is consistent with previous studies (Rutten et al., 2015;Zhang et al.,2013) and that suggests that high local-scale diversity can be maintained under moderate levels of disturbance. Furthermore, these results suggest that degraded forests may act as reservoirs of native tree diversity and, thus, play an important role in the conservation of diverse tropical forests Rozendaal et al., 2015). A potential explanation of the observed pattern is that low or medium forest use intensities or the creation of gaps in the forest canopy may create new habitats that favor the establishment of fast-growing and light-demanding tree species (Ramıŕez-Marcial et al., 2001;Zhang et al., 2013) or may facilitate the growth of advanced regeneration of shade-tolerant species into larger size classes (Brokaw, 1985;Denslow, 1987). However, forests subjected to frequent disturbances may be vulnerable to biological invasions in the future (Alpert, Bone, & Holzapfel, 2000), although we did not detect the presence of any non-native woody species in our inventory. On the other hand, we found that secondary forests had lower tree diversity compared to old-growth forests at half of the elevational sites. This suggests that high forest use intensity reduces diversity and that tree communities, particularly old-growth forests, need more time to recover in species composition (Crouzeilles et al., 2016;Gossner et al., 2016;Peters et al., 2019;Rozendaal et al., 2015). Yet the effects of both low and high forest use intensity, and the subsequent trajectory of recovery, may depend on intrinsic biotic and abiotic conditions within each elevation, e.g. differences in dispersal limitation, distance from adjacent forest, and propagule sources (Holl, 1999;Martínez-Garza & González-Montagut 1999;van Breugel et al., 2013).
The interactive effects of elevation and forest use intensity resulted in contrasting tree diversity patterns along the elevational gradient. The low-plateau elevational gradient for tree diversity in old-growth forests gradually changed into a bimodal pattern for degraded and especially secondary forests. These results are broadly in line with a recent multi-taxon study from Mount Kilimanjaro in Tanzania that showed interactive effects of climate and land use change on diversity trends (Peters et al., 2019).
Specifically, in our study, the low plateau pattern in old-growth forest was driven by similar values in species richness between sea level and 2,000 m, above which it decreased strongly towards the treeline. Such an elevational pattern in tree diversity has been described before, although it appears to be uncommon (Jankowski et al., 2013;Rana et al., 2010). The high species diversity between sea level and 2,000 m may be linked to climatic conditions; for instance, tree diversity usually increases with temperature and precipitation (Homeier et al., 2010;Toledo-Garibaldi & Williams-Linera, 2005). In this regard, it is interesting that the highest species richness observed in our study occurred in the warm but comparatively dry lowlands (Portillo-Quintero & Sánchez-Azofeifa,2010). However, this observation is in line with previous studies concerning trees along elevational gradients showing that temperature is the primary climatic predictor of tree diversity and stronger than precipitation (Sharma, Behera, Das, & Panda, 2019;Toledo-Garibaldi & Williams-Linera, 2005). We attribute the sharp decrease in tree diversity above 2,000 m mainly to low minimum temperatures and the frequent occurrence of frost (-3°C absolute minimum temperature in winter; Pereyra, Palma, & G., & Zitacuaro C, I., 1992;Toledo-Garibaldi & Williams-Linera, 2005; C.I. Carvajal-Hernández, unpublished data). These thermal conditions represent strong biophysical constraints that likely limit the occurrence of tropical tree species (Veintimilla et al., 2019;Zanne et al., 2014), which is consistent with patterns reported from the Himalayas by Bhattarai and Vetaas (2006) and Rana et al. (2010).
Interestingly, this decrease in tree diversity above 2,000 m was also observed for degraded and secondary forests, suggesting a strong role of ecological factors associated with elevation.

| Floristic composition along the elevational gradient and forest use intensity
We found marked differences in tree composition related to forest use intensity at most elevations, especially between oldgrowth and secondary forests. This suggests that, at most elevations, high-intensity forest use strongly affected community composition, which supports results from previous studies in Neotropical forests (Norden, Chazdon, Chao, Jiang, & Vílchez-Alvarado, 2009;Dent et al., 2013 DeWalt, & Denslow, 3). This observed shift in forest composition may be explained by the fact that the time needed to recover species composition may be longer than that for species richness (Rozendaal et al., 2015), as strong environmental filtering in tropical secondary forests limits the diversity of adaptive trait combinations (Lebrija-Trejos et al., 2011). For instance, the abiotic conditions in secondary forests typically favor light-demanding, fast-growing species over shadetolerant, slow-growing species, which are better adapted to abiotic conditions found in old-growth forests (Bazzaz & Pickett, 1980;Crouzeilles et al., 2016;Ewel, 1980;Finegan, 1984;Gómez-Pompa & Vásquez-Yanes, 1974;Guariguata & Ostertag, 2001;Swaine & Whitmore, 1988). Other factors may similarly contribute to the differentiation in species composition of old-growth and secondary forests, such as previous forest use type and proximity of seed sources (Guariguata & Ostertag, 2001;Muñiz-Castro, Williams-Linera, & Benayas, 2006;Rozendaal et al., 2015;Zhang et al., 2013). In contrast, the composition of tree communities did not vary significantly between old-growth and degraded forests within five or six of the eight studied elevations (based on incidence or abundance-based dissimilarities, respectively). This suggests that moderate forest use intensity in most degraded tropical forests did not appreciably alter abiotic conditions, as many common species (probably shade-tolerant, slow-growing MONGE-GONZÁLEZ Et aL. ones) that also occurred in old-growth forests were able to persist in degraded ones. Our finding is in line with that of a previous study on herbaceous species along the same elevational gradient, which also reported a similar floristic composition of old-growth and degraded forests (Gómez-Díaz et al., 2017). It is important to note that our study did not consider the possible impacts of forest use intensity on the tree seedling community, which may capture the impacts of disturbances more readily than the mature tree community (Alvarez-Aquino et al., 2004;Ramírez-Marcial, 2003). While forest disturbances of even moderate intensity may result in extinction debts in the long term (Moreno-Mateos et al., 2017), the similar floristic composition of old-growth and degraded forests at most elevations supports the idea that degraded forests may act as important reservoirs of biodiversity in human-modified tropical landscapes.

| CON CLUS IONS
We found that forest use intensity significantly altered tree species diversity and composition, and that this effect was modified by elevation. Our results provide evidence that, even in human-dominated tropical landscapes, degraded and secondary forests may safeguard considerable levels of tree diversity. Due to their greater similarity to old-growth forests, degraded forests may act as reservoirs for biodiversity conservation and restoration. In conclusion, the interactive effects between land use and elevation render predictions across elevations difficult and highlight the value of examining how forest use intensity may alter diversity patterns along elevational gradients in tropical forests.

AUTH O R CO NTR I B UTI O N S
MLM-G, HK and TK conceived the study; MLM-G, AH-S and GC-C collected data; MLM-G, DC and NG-R analysed the data; MLM-G wrote the paper with contributions from HK, DC, NG-R, VG-J and TK; all authors discussed the results and commented on the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Primary data are available in Appendix S1 of Supporting Information.