Study plots
We chose two human-modified forests in Meinong District in Kaohsiung City, and one reference plot in the Nanjenshan Reserve in Pingtung County (Fig. 1). Both human-modified forests experienced negligible human disturbances since 1992. The reference primary forest, Nanjenshan Reserve, has been designated as a nature reserve since 1982.
The mean temperature in Meinong District is 24.0 °C (Central Weather Bureau 2014). The rainy season occurs from May to September. During this period, about 89% of the annual rainfall accumulated in the area. The lowland human-modified forests surrounding Meinong (295–650 m a.s.l.) are ecological buffer zones between human development and montane primary forests in the Central Mountain Range. In addition, mountains surrounding Meinong District shelter the human-disturbed forests from typhoon disturbances. Thus, exotic tropical timber species, such as those in the family of Dipterocarpaceae, were able to reach their natural heights of more than 30 m.
The first human-modified forest is an exotic species-poor plantation, Yellow Butterfly Valley (YBV; 120°35′33″E, 22°55′58″N). The area of YBV is 2805 ha (Chen 2013). YBV is a plantation where approximately 6 non-native timber and fruit tree species have been widely planted. Thus, the forest is relatively poor in its richness of non-native tree species. The name Yellow Butterfly Valley (YBV) is adopted by the local people referring to the yellow butterflies (Catopsilia pomona) which are attracted by the widely planted exotic tree, Senna siamea (Chen 2013). Based on our interviews with local farmers and researchers, YBV has experienced two intensive tree planting events. The first intensive period (from 1936 to 1945) was during the Japanese occupation. Two non-native timber trees, namely Senna siamea, and Tectona grandis, were extensively planted. The second intensive planting period was from 1949 to 1975, when the Taiwan Forestry Bureau enforced the rent afforestation policy. Fruit trees, including Mangifera indica, Litchi chinensis, Euphoria longana, and Bambusa spp., were commonly planted for afforestation and non-timber forest products. Based on land use policy amendment (Huang 2002) and local interviewees, logging activities eased around the late 1970s, reflecting the alteration of local economic activity. In 1992, some parts of YBV had additional plantations of Mangifera indica in order to receive governmental compensation for the Meinong reservoir plan (Meinung People’s Association 1994; Tsai et al. 2013). These Mangifera trees were not managed after planting because the reservoir construction plan was stopped by local people (Meinung People’s Association 1994). Thereafter, YBV has become a natural sightseeing attraction, especially for its yellow butterflies, with negligible human logging disturbances.
The other human-modified forest is an exotic species-rich arboretum, Shuanghsi Tropical Botanical Garden (STBG; 120°36′3″E, 22°56′10″N). STBG is adjacent to YBV and its area is 7.56 ha (Yang et al. 2010). STBG was originally known as the Meinong Twin Creek Arboretum which has high richness in exotic tree species. In 1935, 270 tree species were planted for the purpose of cultivating tropical non-native forest timber species (Yang et al. 2010). Chang (1970) found that majority of the originally planted species were not able to adapt to the environment in STBG, and by 1968 only 97 non-native species survived (Chang 1970). In 2008, the recorded non-native species in STBG was even lower, with only 75 non-native tree species (Yang et al. 2010). In 2008, the dominant (based on basal areas) non-native species in STBG included Swietenia macrophylla, Spathodea nilotica and Terminalia calamnsanai (Yang et al. 2010). Concurrently, more than 46 naturally recruited native tree species have been recorded, such as Schefflera octophylla, Ficus irisana, and Machilus zuihoensis (Yang et al. 2010). STBG was officially opened to public visitation in 1987 (Chuang 2013), and since then it has been jointly managed as a recreation park by the Taiwan Forestry Bureau and the local community (Chuang 2013).
The reference plot is within the Nanjenshan Reserve, namely Nanjenshan Plot I (NPI; 120°50′51″E, 22°04′54″N) (Fig. 1). Its elevation ranges from 224 to 275 m (Chao et al. 2010a). NPI was established in 1993 (2.1 ha in size) for the purpose of long-term ecological research (Chao et al. 2010a). The mean annual temperature is 22.7 °C and the mean annual rainfall is 3252 mm, without a dry season (Chao et al. 2010b). Tree census of NPI has been conducted every 7–9 years since its establishment (Chao et al. 2010a). The forest in NPI is a tropical lowland rainforest which was classified as a Ficus-Machilus zone (Hsieh et al. 2000). A more recent study categorizes the NPI forest as a tropical foothill evergreen broad-leaved forest, which is in a Dysoxylum–Machilus zone (Li et al. 2013). The major difference is that the Ficus–Machilus zone describes the overall lowland forests in Taiwan (Su 1984), whereas the Dysoxylum–Machilus zone is only the lowland forests confined to southern Taiwan (to the south of 23.2°N) (Li et al. 2013).
The reference primary forest is the least disturbed lowland tropical primary forest in Taiwan (Editorial Committee of the Flora of Taiwan 1994–2003), and it has long-term census records (Chao et al. 2010a). However, the reference forest is not a perfect control because it is located approximately 130 km away from the studied human-modified forests, and it has no dry season. Nevertheless, we adopted NPI as a reference forest for the following reasons: (1) There was no lowland primary forest in Meinong (Weng 2013). (2) Based on the vegetation classification scheme in Li et al. (2013), the potential tropical foothill evergreen broad-leaved forest in Taiwan is to the south of 23.2°N. Both Meinong District (22.9°N) and the Nanjenshan Reserve (22.1°N) are within the range. (3) All the three plots are relatively sheltered from the impacts of typhoon [a crucial disturbance type for vegetation in Taiwan (Lin et al. 2010)]. Therefore, although the reference plot was not perfect, it is able to give some insights into the forest structure and composition of potential primary forests. We did not intend to imply that the Nanjenshan forests would be the final successional forest for the two human-modified forests. Rather, the reference plot was used to provide a basal line for the comparison of diversity and regeneration status between the two human-modified forests.
Quadrat sampling
For each of the human-modified forest, 4 transects were set up in July 2013. The distance between transects was at least 90 m and each transect had 4 quadrats. Therefore, a total of 16 quadrats (each size 10 m × 5 m) were sampled in each of the study plots (Loo 2015). In order to minimize environmental variations, we placed transects with aspects of approximately 300° and slope angles <40°. The quadrats were set up along a human accessibility distance gradient (from trail/river side into forest interior) at 10 m intervals. Some quadrats (n = 5 out of 32) did not follow the 10 m interval rule as we attempted to avoid steep slopes (>40°), hill top, and forest light gaps. These heterogenetic microenvironments were not considered in our study (Loo 2015).
In the Nanjenshan primary forest, we selected 16 quadrats (at 10–15 m intervals), measuring 10 m × 5 m. These quadrats were confined to those located in aspects at approximately 300° and slope angles <40° in the Nanjenshan Plot I (Loo 2015).
Data collection
We conducted tree censuses of the human-modified forests in July and August 2013. We used tree census data of NPI conducted in 2008. Vegetation data, including tree species, height, and DBH (diameter at breast height), were collected for evaluating forest structure, diversity, and composition. All trees ≥1 cm DBH in the selected quadrats were measured. Plant identification and nomenclature were based on Flora of Taiwan (Editorial Committee of the Flora of Taiwan 1994–2003) with three exceptions. (1) Radermachera sinica was misspelled as Radermachia sinica in Li (1998). Our study used its correct name according to International Plant Names Index (2005). (2) Glochidion ovalifolium was not recorded in Flora of Taiwan. We followed the name of this species in Lu and Hsu (2003) and Hsu et al. (2006). (3) The familial classification of the recorded species were updated with APG IV system (The Angiosperm Phylogeny Group 2016).
Data analysis
Forest structure
We classified tree heights into four classes: class 1: ≤5 m; class 2: >5 and ≤ 10 m; class 3: >10 and ≤20 m; class 4: >20 m. We also classified DBH measurements into three size classes: class 1: ≥1 and <10 cm; class 2: ≥10 and <20 cm; class 3: ≥20 cm. We assumed that trees in the size class 1 (DBH ≥1 and <10 cm) had naturally regenerated within 20 years, approximately after the major historical human disturbance. We assumed that trees in the size class 2 (DBH ≥10 and <20 cm) had survived for approximately 20–40 years. Also, we assumed that trees in the size class 3 (DBH ≥20 cm) were the remnant trees or planted trees that had survived for more than 40 years. We then compared the structures between the primary forest (i.e., NPI) and the human-modified forests (i.e., YBV and STBG) by the distribution frequency of height and size classes. We used the proportion of native and endemic species individuals to compare regeneration statuses in the three forests. We assumed that a higher percentage of sapling individuals belonging to species native to Taiwan would indicate a better regeneration status.
Species diversity
Species richness, effective species number, evenness, and rarefaction (Magurran 2004) were used to compare species diversity. Species richness (S) refers simply to the number of species, which is weighted by the number of rare species (Hill 1973). Effective species number (denoted by N
1
) is the exponentially transformed Shannon index (exp (Hʹ)) (Hill 1973).
$$N_{ 1} = { \exp }\left( { - \varSigma p_{i} { \ln }p_{i} } \right),$$
where ρ
i
refers to the ratio of the individuals counted for the i
th species to the total individuals in a plot. Effective species number represents a diversity index weighted by the proportion of individuals. Evenness is N
1/S which represents the equitability of the species in a community (Buzas and Hayek 1996). Due to differences in sampling efforts (e.g., the number of individuals) between communities, comparing rarefaction curves can help to account for the sampling efforts on the patterns of species diversity (Gotelli and Colwell 2001; Magurran 2004). The rarefaction curve is produced by randomly and repeatedly re-sampling a pool of N individuals or N quadrats, and then plotting the average number of species represented by 1 to N number of individuals or quadrats (Gotelli and Colwell 2001). All the indices were computed by the software PAST (PAleontological STatistics, version 3.04, Natural History Museum, University of Oslo, Norway).
Species importance values
We calculated the species importance values (IV %) as
$${\text{IV}}\% = \left( {RBA\% + RD\% + RF\% } \right)/ 3,{\text{ where}}$$
RBA% is a relative basal area, RD% is a relative density, and RF % is a relative frequency (Mueller-Dombois 1974). The equations are as follows:
$$\begin{aligned} RBA\% = \left( {BA/\varSigma BA{\text{of all the species in the plot}}} \right) \times 100\% ; \hfill \\ RD\% = \left( {D/\varSigma D{\text{of all the species in the plot}}} \right) \times 100\% ; \hfill \\ RF\% = \left( {F/\varSigma F{\text{of all the species in the plot}}} \right) \times 100\% ; \hfill \\ \end{aligned}$$
Basal area of an individual was calculated as π (DBH/2)2, and the values of all the individuals for a target species were summed and transformed into a unit area value (BA, m2 ha−1). Density (D, number of individuals ha−1) was the number of individuals per ha for a target species. Frequency (F) was the number of quadrats in which a target species was found within a plot, divided by the total number of quadrats in that plot (Mueller-Dombois 1974).
Similarity indices
We compared species similarities of the three plots using a qualitative similarity index of Sørensen and a quantitative similarity index of Motyka (Mueller-Dombois 1974). The Sørensen and Motyka indices emphasize different properties in similarity. The first emphasizes the presence or absence of a species. The latter considers the quantity of a species (e.g., density). The Sørensen similarity index S
s was calculated as
$$S_{s} = ( 2C/\left( {A + B} \right) \times 100\% ),$$
where C is the number of shared taxa found in both plots; A is the number of all taxa found in plot A; B is the number of all taxa found in plot B (Mueller-Dombois 1974). The Motyka similarity index IS
mo was calculated as
$$IS_{mo} = ( 2M_{w} /\left( {M_{A} + M_{B} } \right) \times 100\% ),$$
where M
W
is the lower stem density of shared taxa found in both plots; M
A
is the total stem density found in plot A; M
B
is the total stem density found in plot B (Mueller-Dombois 1974). The possible maximum number of the two indices is 100%, indicating that the species composition of the compared plots is exactly the same.