Taxonomic, Phylogenetic, and Functional Diversity of Ferns at Three Di ﬀ erently Disturbed Sites in Longnan County, China

: Human disturbancesare greatly threateningto the biodiversityof vascular plants. Compared to seed plants, the diversity patterns of ferns have been poorly studied along disturbance gradients, including aspects of their taxonomic, phylogenetic, and functional diversity. Longnan County, a biodiversity hotspot in the subtropical zone in South China, was selected to obtain a more thorough picture of the fern–disturbance relationship, in particular, the taxonomic, phylogenetic, and functional diversity of ferns at di ﬀ erent levels of disturbance. In 90 sample plots of 5 × 5 m 2 along roadsides at three sites, we recorded a total of 20 families, 50 genera, and 99 species of ferns, as well as 9759 individual ferns. The sample coverage curve indicated that the sampling e ﬀ ort was su ﬃ cient for biodiversity analysis. In general, the taxonomic, phylogenetic, and functional diversity measured by Hill numbers of order q = 0–3 indicated that the fern diversity in Longnan County was largely inﬂuenced by the level of human disturbance, which supports the ‘increasing disturbance hypothesis’. Many functional traits of ferns at the most disturbed site were adaptive to the disturbance. There were also some indicators of fern species responding to the di ﬀ erent disturbance levels. Hence, ferns may be considered as a good indicator group for environmental stress.


Study Sites
Longnan County (24.91 • N, 114.79 • E) is located in South China, with a subtropical monsoon climate and with an elevational range of 190-1430 m a.s.l. We selected three study sites ( Figure 1; Table 1) to carry out fern investigations. The distance between the three sites is approximately 65 km (Jiulianshan National Nature Reserve (JLS) to Leigongshan Family Farm (LGS)), 37 km (JLS to Anjishan Provincial Forest Park (AJS)), and 33 km (AJS to LGS). The annual rainfalls of the three sites fall in the isohyet of 1500-1600 mm, while their mean temperatures in January and July fall in the isotherm of 8 • C and 28 • C, respectively [84]. According to the maps of the U.S. Geological Survey, the generalized geologic ages of all three sites are Paleozoic/Precambrian (PzpCm) [85]. Detailed descriptions of Longnan County and the three sites were provided in our previous studies [86][87][88][89].
Diversity 2020, 12, x FOR PEER REVIEW  3 of 16 to a more thorough picture of the fern-disturbance relationship and (2) assess the Hill number-based taxonomic, phylogenetic, and functional diversity of ferns at different levels of disturbance.

Study Sites
Longnan County (24.91°N, 114.79°E) is located in South China, with a subtropical monsoon climate and with an elevational range of 190-1430 m a.s.l. We selected three study sites ( Figure 1; Table 1) to carry out fern investigations. The distance between the three sites is approximately 65 km (Jiulianshan National Nature Reserve (JLS) to Leigongshan Family Farm (LGS)), 37 km (JLS to Anjishan Provincial Forest Park (AJS)), and 33 km (AJS to LGS). The annual rainfalls of the three sites fall in the isohyet of 1500-1600 mm, while their mean temperatures in January and July fall in the isotherm of 8 °C and 28 °C , respectively [84]. According to the maps of the U.S. Geological Survey, the generalized geologic ages of all three sites are Paleozoic/Precambrian (PzpCm) [85]. Detailed descriptions of Longnan County and the three sites were provided in our previous studies [86][87][88][89].  The locations of three study sites in Longnan County, China. The imagery data of Longnan County were obtained from Map World (http://www.tianditu.gov.cn) and produced with QGIS 3.8 [90].

Sampling Techniques
We chose typical roads/trails across the dominant vegetation type of each site for fern investigation. At each study site, we placed 30 sampling plots in suitable habitats at the roadsides. There were a total of 90 plots in our study. Each plot was 5 m × 5 m, as in some previous local-scale practices [48,51,52,91,92], which is suitable for roadside sampling, especially for the managed forests and orchards at LGS. The distances between every two plots were all above 600 m. We covered large areas and thus encountered more fern species at each site. For the analyses, we treated all 30 plots at each site as one assemblage.
We identified and counted all the ferns in each plot. The undetermined fern species were marked and taken back to the laboratory for further identification using keys. As lycophytes are not ferns [93], we excluded them from our analyses. Fern species names were verified with the R package 'plantlist' [94].

Data Analyses
We used the number of individuals for every fern species in an assemblage to calculate the following diversity indices: (1) Based on the assemblage datasets (Table S1), taxon richness were computed with Microsoft Excel 2016.
(2) At present, we do not have the molecular sequences of all fern species and could not build a species-resolution phylogenetic tree. Therefore, we generated a genus-resolution phylogenetic tree ( Figure S1, File S1) with the R package 'V.PhyloMaker' using the 74,533-species mega-tree GBOTB.extended.tre, which covers all families in extant vascular plants [95]. The mega-tree is updated from two recent mega-trees [96,97], with the function to bind the undetermined species to their close relatives in the phylogenetic tree [95]. Phylogenetic diversity indices were calculated with the R package 'picante' [98].
(3) We extracted the fern traits (Table S2) from publications and websites. We took into account 17 traits, including the maximum lamina length, maximum lamina width, petiole length, frond heterophylly, lamina texture, leaf arrangement, rhizome type, rhizome position, lamina dissection, lamina shape, sori shape, sori position, indusia, reproduction type, phenology, habit and scales/hair density [12,70,71], to calculate the functional distances between species and hence the functional diversity of the three assemblages. For the quantitative traits, we used the measured values directly; for the presence/absence traits, we assigned 1-0 values; and we quantified qualitative traits with biological/ecological ranks (Table S2). The summary statistics, including the mean, median, minimum, maximum, and standard deviation of each trait value, are listed in Table S2. The functional diversity indices were obtained with the R package 'FD' [99].
(4) Taxonomic, phylogenetic, and functional diversity through Hill numbers (q) are unified standardization methods to quantify the diversity of different communities [73][74][75][76][77][78]. The equivalent diversity indices linked to Hill numbers of order q = 0, 1, 2 are listed in Table 2. In every case, we performed these analyses with the R package 'hillR' [100].  [73,74,101,102] phylogenetic diversity # the effective total phylogenetic distance between species [73,103] functional diversity $ the effective total functional distance between species [104][105][106] @ p i = the relative abundance of the ith species; S = the number of species (i.e., species richness); H = Shannon's entropy; λ = Simpson's dominance index [73,74,101,102]. # T = the age of the root node of the ultrametric phylogenetic tree; B T = the set of all branches in the time interval [−T, 0]; L i = the length of the ith branch; a i = the total abundance descended from the ith branch; FPD = Faith's phylogenetic diversity; A = Allen's phylogenetic entropy; Q = Rao's quadratic entropy; the corresponding equations and the equivalent diversity indices for taxonomic classification and nonultrametric phylogenetic tree can refer to [73,103]. $ d ij = the functional distance between the species pair (i, j); p i p j = join probability for the species pair (i, j); S = the number of species; Q = Rao's quadratic entropy; FAD = Functional attribute diversity; GS = the weighted Gini-Simpson's index [101,[104][105][106].
We performed indicator species analysis based on the number of individuals for all fern species in the 90 sample plots. Indicator species for different disturbance levels were obtained using the R package 'indicspecies' [107].

Sample Coverage
Overall, we recorded a total of 4 subclasses, 9 orders, 20 families, 50 genera, and 99 species of ferns, as well as 9759 individual ferns (Table S1; Figure S1). The sample coverage curves of the ferns at three sites indicated that all sampling efforts were sufficient as they began to level off at the 26th, 19th, and 15th plot (i.e., 650 m 2 , 475 m 2 , and 375 m 2 ) for JLS, AJS, and LGS, respectively ( Figure 2).
The taxonomic diversity indicators all showed the same arrangement from higher to lower values at the three study sites: JLS > AJS > LGS, except in the abundance of individuals, which was greater at LGS (Table 3). This trend was consistent along with the increasing q values of Hill numbers for all species ( 0 TD or species richness S), common species ( 1 TD or e H ), and dominant species ( 2 TD or 1/ λ ) (Figure 3a). The same pattern was found for phylogenetic diversity, including Faith's phylogenetic diversity (linked to 0 PD) and Rao's quadratic entropy (linked to 1 PD) (Figure 3b) and for functional diversity, including functional attribute diversity (FAD or 0 FD) and the weighted Gini-Simpson index (linked to 2 FD) (Figure 3c).
The trends of the Hill number-based taxonomic, phylogenetic, and functional diversity all leveled off, and the differences between three sites became smaller (Figure 3) when the weighting of the relative abundance of each species (pi) increased with the increase of order q ( Table 2). The phylogenetic diversity among the three sites became similar when q > 1, one possible reason might be that our phylogenetic tree was at the genus resolution rather than species resolution ( Figure S1).  A sampling area of 600 m 2 is enough for human-made forests, while 900 m 2 is suggested for natural forests in the tropical regions of Malaysia and Singapore [110]. Our study sites were located in the subtropical regions, so the minimum sampling areas were smaller than those in the tropical regions. Our results also agreed that the minimum sampling areas should increase from economic to natural forests [110]. The total sampling area of 750 m 2 (i.e., 30 plots) should be enough to show the relative biodiversity values at different disturbance levels.
The taxonomic diversity indicators all showed the same arrangement from higher to lower values at the three study sites: JLS > AJS > LGS, except in the abundance of individuals, which was greater at LGS (Table 3). This trend was consistent along with the increasing q values of Hill numbers for all species ( 0 TD or species richness S), common species ( 1 TD or e H ), and dominant species ( 2 TD or 1/λ) (Figure 3a). The same pattern was found for phylogenetic diversity, including Faith's phylogenetic diversity (linked to 0 PD) and Rao's quadratic entropy (linked to 1 PD) (Figure 3b) and for functional diversity, including functional attribute diversity (FAD or 0 FD) and the weighted Gini-Simpson index (linked to 2 FD) (Figure 3c).  Hence, all three diversity components (taxonomic, phylogenetic, and functional) of fern assemblages studied here showed the same pattern of decreasing diversity with the increase of disturbance degrees, which is in accordance with the 'increasing disturbance hypothesis' [8,55,57,65,111]. Our sampling was conducted along roadsides, which are disturbed habitats in themselves. Thus, we are careful to interpret them in direct relation to disturbances as such, as they The trends of the Hill number-based taxonomic, phylogenetic, and functional diversity all leveled off, and the differences between three sites became smaller (Figure 3) when the weighting of the relative abundance of each species (p i ) increased with the increase of order q ( Table 2). The phylogenetic diversity among the three sites became similar when q > 1, one possible reason might be that our phylogenetic tree was at the genus resolution rather than species resolution ( Figure S1).
Hence, all three diversity components (taxonomic, phylogenetic, and functional) of fern assemblages studied here showed the same pattern of decreasing diversity with the increase of disturbance degrees, which is in accordance with the 'increasing disturbance hypothesis' [8,55,57,65,111]. Our sampling was conducted along roadsides, which are disturbed habitats in themselves. Thus, we are careful to interpret them in direct relation to disturbances as such, as they likely combine the effects of different degrees of disturbances at the regional level (study sites) and at the local scale (individual roadsides). A previous study in Mexico found that the functional diversity of ferns and lycophytes decreased with disturbance in species-rich habitats, but not so in species-poor habitats [12]. As our study sites were located close to each other and have similar ecological conditions, we may assume that they originally shared a similar fern flora; thus, we cannot assess the effect observed in Mexico. Clearly, the relationship of fern diversity to habitat disturbance is complex and requires further case studies.
Previous studies have found inconsistent trends among the taxonomic, phylogenetic, and functional diversity along disturbance gradients [112][113][114][115]. However, in our study, all three diversity elements consistently indicated that the biodiversity was the highest along the roadsides within the natural forests at JLS but at the lowest in the severely disturbed LGS.
Many functional traits of ferns showed differences from the species of the less altered site to those of the most disturbed site (Table 4). We found that ferns in the more strongly disturbed sites tended to have longer, narrower, and thinner laminae, and more commonly had monomorphic fronds, with sori in one row on either side of the veins, and false indusia. They also had less vegetative reproduction. At the most disturbed site of LGS, the petioles were longest, the rhizomes were all below ground, long creeping rhizomes were less common, the fronds were more strongly divided, and the scales or hairs became denser. It is tempting to interpret these trends as adaptive directly; however, it should be borne in mind that many of them show strong phylogenetic patterns, being present in some clades and not in others. Accordingly, certain traits may be overrepresented in some habitats, not because they themselves are adaptive, but rather because some other trait of a group of ferns is adaptive. Bearing these limitations in mind, some patterns deserve closer consideration. Disturbed sites have higher solar radiation and lower air humidity, which is suitable for drought-tolerant ferns but not for humidity-dependent ones [12]. Several of the functional traits recovered as being overrepresented in the disturbed site in our study were previously suggested to be adaptations to disturbance. Thus, in dry and hot environments, plants have narrower and pinnate leaves to allow for heat dissipation [116], and ferns show more coriaceous leaves, succulent rhizomes, denser leaf scales, and higher cell wall elasticity [15]. Rhizome type is related to space/light competition [117], frond heterophylly to spore dispersal facilitation, high lamina thickness to frost/drought adaptations, high scale and hair density to heat/water/herbivory protection, hydathodes to water transport facilitation, indusia to spore protection from water washing, and buds to wet environments [71].

Indicator Fern Species for Different Disturbance Levels
There were seven indicator fern species for the roadside in the strongly disturbed study sites, four for the site with an intermediate disturbance level, seven for the site with a low disturbance level, and one for the latter two sites combined (Table 5). Many fern groups, such as Gleicheniaceae, Dennstaedtiaceae, Pteridaceae, and Schizeaceae, include species that are clearly adapted to open, sunny habitats with poorly developed soils and that quickly colonize disturbed sites. Many tropical landslides are quickly colonized by Gleicheniaceae [118]. Among the seven high-disturbance indicators, previous studies found that Dicranopteris pedata and Blechnum orientale dominate the early successional stage on landslide trails [119]; Odontosoria chinensis frequently colonizes disturbed places [120]; Lygodium microphyllum has the ability to invade hurricane-disturbed areas [121]; Cyclosorus parasiticus adapts to disturbed habitats [122]; Christella dentata is common along roadsides but can also dominate in undisturbed, lightly disturbed, and moderately disturbed forests [123], which makes it a poor disturbance indicator; Pteris semipinnata is a dominant understory species in the economic needle-leaf forest of Pinus massoniana [124]. Our results are thus in accordance with the previous observations and point to a certain generality of the results obtained by us.
Among the seven low-disturbance indicators and one intermediate + low disturbance indicator: Ctenitis subglandulosa is a dominant species in the herbaceous layer in a natural evergreen forest [125]; the tree fern Cyathea hancockii is a nationally protected plant in China; and Angiopteris fokiensis and Osmunda vachellii are provincially protected plants in Jiangxi. Clearly, much less is known about these species than about the species of the highly disturbed sites, which may partly reflect that the disturbance-adapted species typically have large ranges and are common.

Conclusions
In general, the taxonomic, phylogenetic, and functional diversity of ferns in Longnan County has been greatly influenced by human disturbances, which supports the 'increasing disturbance hypothesis'. Many functional traits of ferns have adapted accordingly to the increasing disturbance degree. We also identified some indicator fern species corresponding to different disturbance levels. Hence, ferns are an effective indicator group for environmental stress. Compared to the traditional diversity indices, Hill number-based diversity profiles provided a continuous and thorough picture of fern diversity patterns among different disturbance levels. However, the relationship of fern diversity to habitat disturbance is complex and requires further case studies. In the future, more sites should be included, and the disturbance levels should be quantified. Standard sampling protocols, molecular-based species phylogeny, and unified functional traits should also be adopted to obtain meaningful comparisons with previous publications and to obtain global-scale patterns of fern diversity.