Variation of functional traits within species

For those epiphyte species that have 20 or more measured individuals, the intraspecific variation of functional traits was analyzed using individual-based linear model. There were four species with sufficient number of measured individuals, including Lemmaphyllum microphyllum (n = 35), Goniophlebium formosanum (n = 31), Hoya

carnosa (n = 23) and Aeschynanthus acuminatus (n = 22). Same as the procedure

mentioned in previous section, six functional traits were first transformed if necessary and then were used as response variables in linear model using R 3.5.0 (R Core Team, 2018). However, with such a small sample size, it is not proper to use multiple explanatory variables. Hence, only height was used as an explanatory variable. It is difficult to use this analysis to make strong conclusions, while the result may still tell some information.

doi: 10.6342/NTU201900489

Results

1. Variation in species composition

Species composition of the 130 plots not below bird’s-nest ferns varied between

vertical zones based on the result of DCA (Figure 22). The plots within the same zone were closer to each other than to plots in different zones. Besides, plots in zone 1 to zone 6 were roughly distributed from right to left along first DCA axis. This indicates that species composition of plots within the same vertical zone was more similar, and species composition changed gradually from lower zones to higher ones.

Those species located near the left end of DCA space, including Pyrrosia lingua (Pyli), Pholidota cantonensis (Phca) and Davallia trichomanoides (Datr), prefered growing on higher places (Figure 23). On the other hand, species preferring growing on lower places were close to the right end of DCA space, including Crepidomanes auriculatum (Crau) and Abrodictyum obscurum (Abob) (Figure 23).

doi: 10.6342/NTU201900489

Figure 22. DCA analysis on plots not below bird’s-nest ferns (showing plots)

Dots with six different colors represent plots located in different vertical zones. Plots in the same zone were close to each other, while the distances between plots in different zones were greater. Besides, plots in zone 1 to plots in zone 6 were roughly distributed from right to left along first DCA axis, indicating that species composition changed gradually from lower zones to higher ones.

doi: 10.6342/NTU201900489

Figure 23. DCA analysis on plots not below bird’s-nest ferns (showing species) Species located close to the left end of first DCA axis were those species preferred growing in higher zones. In contrast, species located in the right preferred growing in lower zones. The complete scientific names of these species abbreviations are shown in Appendix 2.

Four positional variables were used in CCA separately to calculate proportion of variation they can explain alone. The results are shown in Table 2. Height explained most variation of species composition (3.99%), and result of permutation test was significant (p < 0.001). Angle difference for sunlight (ADS) and angle difference for water (ADW) also explained significant variation if tested alone (p = 0.015 & 0.023).

doi: 10.6342/NTU201900489

Table 2. Positional variables and the proportion of variation they explained alone in CCA Positional variables Proportion of

variation explained

p-value of permutation test

Height 3.99% < 0.001***

ADS 1.17% 0.015*

ADW 1.09% 0.023*

Inclination angle 0.85% 0.277

When conducting forward selection, height was included into model first and followed by ADS, while ADW and inclination angle were not included. Hence, there were 2 variables, height and ADS, in the optimal CCA model and they explained 5.15% of variation together (p < 0.001) (Figure 24). Height was basically along first CCA axis, and ADS was along second CCA axis, indicating that the effects of height and ADS were nearly independent (Figure 24).

doi: 10.6342/NTU201900489

Figure 24. Optimal CCA model using two explanatory variables

Optimal CCA model was determined by a forward selection procedure using permutation test. There were two explanatory variables, height and ADS (angle difference for sunlight), in optimal CCA model, and they explained 5.15% of variation of species composition together. The directions of these two variables in CCA space were nearly perpendicular to each other, indicating that the influences of them on species composition were nearly independent.

2. Distribution patterns of different species

There were great variations in the height distribution of the 23 abundant species (Figure 25), indicating that these species used different niches. The synaptic table showing the significant Φ coefficient of each abundant epiphyte species in each vertical zone are shown below (Table 3). There were some unique representative species in zone 1 and

doi: 10.6342/NTU201900489

zone 4. Representative species in zone 1 included Crepidomanes auriculatum (Crau) and Microsorum brachylepis (Mibr), while representative species in zone 4 included Vittaria anguste-elongata (Vian), Goniophlebium formosanum (Gofo) and Loxogramme salicifolia (Losa). Those representative species in zone 5 were all also representative in zone 6, including Davallia trichomanoides (Datr), Hoya carnosa (Hoca) and Dischidia formosana (Difo). Relative frequency of all abundant epiphyte species in all vertical zones is shown in Appendix 4.

Relative frequency of four chosen species is shown below using bar plots (Figure 26). These four species were representative in zone 1 (Microsorum brachylepis), representative in zone 4 (Vittaria anguste-elongata), representative in zone 5 and zone 6 (Davallia trichomanoides) and not representative for all zones (Lemmaphyllum microphyllum), respectively. The distribution patterns of these four species were rather

different (Figure 26).

doi: 10.6342/NTU201900489

Figure 25. Height distribution of the 23 abundant epiphyte species

The box and whisker of each species was drawn based on the height of the plots containing that species. The lower and upper limits of the box were the first and the third quartiles. The black line inside the box indicated the median, and the species were sorted based on the medians. The whisker extended to the lowest or highest data point that still within 1.5 interquartile range (IQR) from limits of the box. Data points outside this range were viewed as outliers and shown as white dots. The height distribution of different species showed large differences, indicating niche differentiation. Complete scientific names of these species abbreviations are listed in Appendix 2.

doi: 10.6342/NTU201900489

Table 3. The synoptic table showing representative species in each vertical zone

The Φ coefficient of all abundant epiphyte species in all vertical zones were calculated, but only those Φ coefficient that is significant based on Fisher’s exact test and higher than 0.2 are shown in the synoptic table. Species with those significant and high enough Φ coefficient are viewed as representative species in those zones. Species in the table were sorted using Φ coefficient. Species representative in only 1 zone are listed first, followed by species representative in more than 1 zone, and those species not

representative in all zones are not listed. The complete scientific names of these species abbreviations are shown in Appendix 2.

Species Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6

doi: 10.6342/NTU201900489

Figure 26. Relative frequency in each vertical zone for four species

Relative frequency is the proportion of plots containing that species in a certain vertical zone. The four species here were chosen based on the result of Φ coefficient analysis.

Microsorum brachylepis was a representative species in zone 1. Vittaria anguste-elongata was a representative species in zone 4. Davallia trichomanoides was a representative species in both zone 5 and zone 6, while Lemmaphyllum microphyllum was not representative in all zones. The distribution patterns of these four species were rather different.

Because angle difference for sunlight (ADS) was correlated with species composition based on CCA, a box plot was made to show the range of ADS each abundant epiphyte species distributed (Figure 27). Compared to height, the ADS distribution of different species did not show large differences. Most of the species were distributed with

doi: 10.6342/NTU201900489

large range of ADS. Only few species had limited distribution, such as Liparis nakaharae (Lina) and Microsorum brachylepis (Mibr).

Figure 27. ADS distribution of the 23 abundant epiphyte species

The box and whisker of each species was drawn based on the angle difference for sunlight (ADS) of the plots containing that species. The lower and upper limits of the box were the first and the third quartiles. The black line inside the box indicated the median, and species were sorted based on the median. The whisker extended to the lowest or highest data point that still within 1.5 interquartile range (IQR) from limits of the box. Data points outside this range were viewed as outliers and shown as white dots. ADS distribution of different species did not show large differences. The complete scientific names of these species abbreviations are shown in Appendix 2.

doi: 10.6342/NTU201900489

3. Variation of functional traits of all epiphytes

Based on individual-based linear model (first approach), the relationships between all the six functional traits and all positional variables are summarized in Figure 28, and more details are shown in Appendix 5. For simple-leaved species, five functional traits had significant relationships with height. SLA (p < 0.001) and Chl_mass (p < 0.001) were negatively correlated with height, while LT (p = 0.009), LWC_area (p = 0.012) and Chl_area (p = 0.045) were positively correlated with height (Figure 28). LDMC only showed marginal significant relationship (p = 0.082) with height. The relationships between the six functional traits of simple-leaved epiphytes and height are shown more clearly using scatterplots and regression curves in Figure 29. On the other hand, for compound-leaved species, only LWC_area (p = 0.002) and LDMC (p = 0.041) showed significant relationships with height (Figure 28).

For positional variables except for height, only LT of compound-leaved epiphytes had a significant positive relationship with ADW (p = 0.006) (Figure 28). However, there were some marginal significant results. LT (p = 0.098), Chl_area (p = 0.094) and Chl_mass (p = 0.062) of simple-leaved epiphytes and Chl_area (p = 0.095) of compound-leaved epiphytes were marginal significantly correlated with inclination angle. Besides, LT of compound-leaved species was marginal significantly correlated with ADS (p = 0.064).

doi: 10.6342/NTU201900489

Figure 28. Relationships between functional traits and all positional variables based on individual-based linear model

Red colors in the figure indicate positive relationships between functional traits and positional variables, while blue colors indicate negative relationships. The densities of colors reflect the degree of significance based on partial F-tests in linear models. Those results with p-value between 0.05 and 0.1 were marginally significant. There were five traits of simple-leaved species and two traits of compound-leaved species were significantly correlated with height. LT of compound-leaved species was significantly correlated with ADW. Abbreviations: LT = leaf thickness; SLA = specific leaf area;

LDMC = leaf dry matter content; LWC_area = leaf water content (per unit area); Chl_area

= chlorophyll content (per unit area); Chl_mass = chlorophyll content (per unit mass);

ADS = angle difference for sunlight; ADW = angle difference for water.

doi: 10.6342/NTU201900489

Figure 29. Scatterplots showing relationships between individual-based functional traits and height

The red curve in each plot shows the predicted values based on the optimal linear model.

Dash line was used if the relationship was marginally significant. Because trait values were transformed using square root or natural logarithm when conducting linear model, the predicted values were retransformed to original trait units and shown in plots. If there were more than one explanatory variables in the optimal linear model, values of other variables were set at mean values when calculating predicted values. The partial R² of

doi: 10.6342/NTU201900489

height in the optimal linear model are shown. This value indicates the proportion of variation that height explained after excluding those variation already explained by other variables in the optimal linear model. The number of star signs (*) besides partial R² indicates the degree of significance based on partial F-test (*: p < 0.05; **: p < 0.01; ***:

p < 0.001). Abbreviations: LT = leaf thickness; SLA = specific leaf area; LDMC = leaf dry matter content; LWC_area = leaf water content (per unit area); Chl_area = chlorophyll content (per unit area); Chl_mass = chlorophyll content (per unit mass).

The results of plot-based community weighted mean (CWM) analysis (second approach) were summarized in Figure 30, and more details are shown in Appendix 6.

Height was the only positional variable that had significant relationships with functional traits based on CWM analysis. LT (p_max = 0.019) and LWC_area (p_max = 0.021) were significantly and positively correlated with height, while SLA ( p_max = 0.004) and Chl_mass (p_max = 0.018) were significantly and negatively correlated with height. The relationships between CWM functional traits and height are shown more clearly using scatterplots (Figure 31). For those traits having significant relationships with height based on CWM analysis, the relationships were consistent to the results of individual-based linear model.

doi: 10.6342/NTU201900489

Figure 30. Relationships between all positional variables and functional traits based on plot-based CWM analysis

Red colors in the figure indicate positive relationships between community weighted mean (CWM) functional traits and positional variables, while blue colors indicate negative relationships. The densities of colors reflect the degree of significance based on p_max permutation tests. Height was the only positional variable that had significant relationships with functional traits. LT and LWC_area were significantly and positively correlated with height, while SLA and Chl_mass were significantly and negatively correlated with height. Abbreviations: LT = leaf thickness; SLA = specific leaf area;

LDMC = leaf dry matter content; LWC_area = leaf water content (per unit area); Chl_area

= chlorophyll content (per unit area); Chl_mass = chlorophyll content (per unit mass);

ADS = angle difference for sunlight; ADW = angle difference for water.

doi: 10.6342/NTU201900489

Figure 31. Scatterplots showing relationships between CWM trait values and height The relationships between community weighted mean (CWM) trait values and height were examined using linear models and tested by p_max permutation tests. The R² value indicates the proportion of variation that height explained. The p_max value is a reliable index to test the significance of linear model in CWM analysis. The number of star signs (*) besides p_max indicates the degree of significance (*: p_max < 0.05; **: p_max < 0.01;

***: p_max < 0.001). For those traits with significant (p_max < 0.05) relationships with height, the regression line were shown in the figure. Abbreviations: LT = leaf thickness;

doi: 10.6342/NTU201900489

SLA = specific leaf area; LDMC = leaf dry matter content; LWC_area = leaf water content (per unit area); Chl_area = chlorophyll content (per unit area); Chl_mass = chlorophyll content (per unit mass).

4. Variation of functional traits within species

The intraspecific variation of functional traits were analyzed for four species with sufficient measured individuals. There was no significant relationship between functional traits and height for all of these four species. However, for marginally significant results, SLA of Lemmaphyllum microphyllum is negatively correlated with height (p = 0.093), LDMC of Lemmaphyllum microphyllum is positively correlated with height (p = 0.078), and Chl_area of Hoya carnosa is negatively correlated with height (p = 0.081). The complete result of intraspecific functional trait analysis is shown in Appendix 7.

doi: 10.6342/NTU201900489

Discussion

1. Variation in species composition

The result of CCA confirms that height had significant influences on epiphyte species composition. Similar results were also reported in several previous studies (Krömer et al., 2007; Zotz, 2007; Parra et al., 2009). This result may come from close relationships between height and several environmental variables. Solar radiation, temperature fluctuation and wind speed typically increase with height while air humidity decrease in a forest (Richards, 1996; Wagner et al., 2013). Vertical environmental gradients of temperature fluctuation and air humidity in the study site were also confirmed.

These environmental variables are influential to physiology of epiphytes (Lambers, 2008;

Zotz, 2016), and hence may influence epiphyte species composition by direct environmental filter or indirect effect via biotic interactions (Cadotte and Tucker, 2017).

Besides, diameters, longevities and stabilities of growing substrates (trunks or branches) may influence epiphyte species composition, and they are also suggested to change with height (Cabral et al., 2015).

Angle difference for sunlight (ADS) was also suggested to have effects on epiphyte species composition based on CCA. The effects of growing aspect on vascular epiphyte assemblages were seldom discussed, while aspect has been proposed to be related to microclimatic conditions, especially light intensity (Davies-Colley et al., 2000), and has

doi: 10.6342/NTU201900489

been usually viewed as an influential factors in studies of non-vascular epiphytic lichens or bryophytes (Kantvilas and Minchin, 1989; Moe and Botnen, 2000). This may be caused by the difference of sampling scales between vascular and non-vascular epiphyte studies.

Vascular epiphytes are usually surveyed by larger sampling unit, such as vertical zones or Johansson zones (Johansson, 1974), and these sampling methods are difficult to quantify the effects of micro-scale factors like aspect. The results of this study suggest that aspect of growing site should also be considered in vascular epiphyte studies if we are interested in micro-scale variation of epiphyte species composition.

Inclination angle of growing substrate has been proposed to have effects on epiphyte species composition (Zotz, 2007), while it was not significant in CCA in this study. The non-significant result of this study should not be simply interpreted as denying the importance of inclination angle. This study primarily focused on effects of height on epiphyte species composition and inclination angle of all the sampling plots were within a range from -20° to 20° except one plot (-30°). The non-significant result was possibly come from this small sampling range of inclination angle. I suggest that influences of inclination angle on epiphyte species composition should be further investigated using a more proper sampling scheme.

doi: 10.6342/NTU201900489

2. Distribution patterns of different species

The result of Φ coefficient analysis confirms that height distribution of different epiphyte species showed differentiation, consistent with many previous studies (Johansson, 1974; Krömer et al., 2007; Zotz, 2007; Parra et al., 2009). There were two unique representative species in zone 1—Crepidomanes auriculatum (Hymenophyllaceae) and Microsorum brachylepis (Polypodiaceae). Fern species in Hymenophyllaceae have been suggested to be sensitive to air humidity variation, and typically prefer to grow in very humid environment (Parra et al., 2009). Microsorum brachylepis also grows in humid understory based on previous observations (郭城孟，2008). There were up to three unique representative species in zone 4, suggesting that environmental conditions in zone 4 are different from the others. It was observed that height of the lowest branch of most sampled trees are within zone 4, and these branches may influence nearby microenvironment by blocking sunlight or accumulating organic matter, hence having unique species.

On the other hand, there was no unique representative species in zone 2, 3, 5 and 6.

Zone 2 and zone 3 possibly served as transition zones for species turnover, so did not have unique representative species. All of the species representative in zone 5 were also representative in zone 6, suggesting that environmental conditions of these two zones were very similar. Besides, the species representative in zone 5 and 6 usually had some

doi: 10.6342/NTU201900489

drought-tolerant features, like scaled rhizomes of Davallia trichomanoides (Davalliaceae) and succulent leaves of Hoya carnosa (Apocynaceae) and Dischidia formosana (Apocynaceae).

In an epiphyte study in Bolivia, Krömer et al. (2007) classified epiphyte species into different ecological types including habitat generalists, trunk specialists, canopy specialists and hemiepiphytes. Among them, trunk specialists were similar to the representative species in lower zones in my study, and canopy specialists were similar to the representative species in higher zones. Krömer et al. (2007) also reported that most of trunk specialists were pteridophytes while most of canopy specialists were orchids.

However, the results of my studies were somewhat different. Most of representative species in zone 1 to zone 4 were pteridophytes, and representative species in zone 5 and zone 6 were still pteridophytes and some species in Apocynaceae while there was no any representative orchid species. Compared with Neotropical forests, relative species richness and relative abundance of orchids to other epiphytes were much lower in my study site, and the canopy environment was occupied by species in Apocynaceae instead of orchids. The sparseness of orchids in my study site may come from the lower temperature in winter. Epiphytic orchids are suggested to be less tolerant to cold weather relative to other epiphytes (Zotz and Hietz, 2001), and their species richness and

doi: 10.6342/NTU201900489

abundance hence declined more dramatically from tropical to temperate regions (徐嘉君，

2007).

In this study, epiphytes were surveyed using vertical zones which were defined by absolute height, while many other studies used different sampling schemes. For example, Johansson (1974) divided each sampled host tree into 5 zones based on tree structure.

Zone I and zone II represented lower part and upper part of tree trunk, while zone III to zone V represented inner to outer tree crown respectively (Figure 32). These zones are called Johansson zones and are frequently used in epiphyte studies (e.g., Krömer et.al., 2007). I adopted vertical zoning scheme rather than Johansson’s zoning scheme mainly because this study was aimed to analyze the effects of vertical environmental gradients,

在文檔中福山地區維管束附生植物之物種組成及功能性狀在垂直方向的變化趨勢 (頁 67-107)