Study Sites

This study was conducted in the TNR (19°36′–19°41′ N, 110°58′–111°03′ E), in the northeast of Hainan Island, South China. TNR covers 44 km² at 338 m.a.s.l. and has a tropical monsoon climate, with a distinct wet season from May to October and a dry season from November to April. The mean annual temperature is 23.9°C, and the mean annual precipitation is 1,721.6 mm. The primary forest type throughout TNR was tropical evergreen monsoon forest. All of these forests were deforested prior to 1980 and degraded to grassland, shrub, or, secondary forest. After the TNR was established in 1983, deforestation was banned and forests have been well protected. The history of each plot was reconstructed based on official records from forest faculty or from interviews with the administration of TNR. Each plot was investigated to determine whether human logging or grazing occurred. Representative sites were selected using the following criteria: (a) existence of sufficiently large, spontaneously successional sites; (b) known age of vegetation; (c) general similarity of soils, topography, and primitive vegetation of the sites; and (d) no evident additional disturbance. It thus could be inferred that these study communities for all successional stages could likely develop into the same climax community.

We chose two different successional stages within TNR. One was 30-year-old shrub forest (Shrub: 19°38′26.0″ N, 111˚01′59.0″ E), and the other was 50-year-old secondary tropical evergreen monsoon forest (Forest: 19°40′11.3″ N, 110°01′06.4″ E; Figure 1). Shrub was regarded as the early stage of Forest during community succession. The litter fall in Shrub and Forest is 4.15 to 8.63 t hm² and 2.96 to 10.60 t hm², respectively. Both types of vegetation are mainly distributed between 150 and 330 m elevation.

Figure 1.

Locations of the study plots in secondary tropical evergreen monsoon forest (Panel (a)) and shrub forests (Panel (b)).

Data Collection

Nine 50 × 50 m plots were established randomly within each vegetation type (Shrub and Forest) using an electronic total station (Leica TSP1200+; Heerbrugg, Switzerland). All plots were located on southeastern slopes with inclination ranging from 30° to 45°. Each plot was divided into twenty-five 10 × 10 m subplots and one hundred 5 × 5 m quadrats for precisely measuring the tree coordinates. All free-standing trees with diameter at breast height ≥1.0 cm were mapped and were identified to species according to the nomenclature in Flora Reipublicae Popularis Sinicae (Editorial Committee of Flora of China, 2004).

Soil organic matter (OM), pH, total nitrogen (TN), total phosphorus (TP), total potassium (TK), available nitrogen, available phosphorus (AP), and available potassium (AK) were measured during the wet season. Five soil profiles in each plot were chosen randomly and sampled. For each profile, the litter or grass layer above the soil were carefully removed, and a top-soil sample (0–20 cm) was obtained (Jobbágy & Jackson, 2001). Soil samples were mixed evenly, air-dried, and then ground to a fine powder using a mill. Finally, the samples were hand-sieved and analyzed.

Soil for OM (g · kg⁻¹) analysis was ground to ∼2.00 mm. First, 0.5 g of soil were mixed with 5 ml of 1 N K₂Cr₂O₇ solution and 5 ml 98% H₂SO₄ and left for 30 min. Second, deionized water and concentrated H₃PO₄ were added, and the mixture was titrated with 0.5 mol · L⁻¹ ferrous ammonium sulfate solution until the color changed from violet-blue to green, indicating the end point of titration.

Soil for pH analysis was ground to ∼2.00 mm. First, 25 g of soil were added to deionized water and stirred. Second, pH of the suspension was measured using a pH electrode at 20°C to 25°C.

Soil for TN (g · kg⁻¹) analysis was ground to ∼2.00 mm. First, 1 g of soil was digested in 98% H₂SO₄ with K₂SO₄–CuSO₄ × 5 H₂O–Se. Second, ammonium-N from the digest was obtained by steam distillation using 0.1 mol · L⁻¹ NaOH in excess to raise the pH. Third, the distillate was collected in 2% H₃BO₃ and titrated with 0.05 M H₂SO₄ to pH 5.0. Finally, the concentration of TN was calculated according to changes in the volume of the 0.05 M H₂SO₄ solution.

Soil for TP (g · kg⁻¹) analysis was ground to ∼2.00 mm. First, 0.25 g of soil were digested with 60% HClO₄ solution. Second, the digest was diluted with ammonium-vanadomolybdate reagent, standards, and samples at 700 nm wavelength. Third, TP was calculated from the calibration curve.

Soil for TK (g · kg⁻¹) analysis was ground to ∼2.00 mm. First, 0.25 g of soil were melted using 2 g solid NaOH for 15 min at 720°C. Second, the solution was collected and dissolved in deionized water. Third, a calibration curve as plotted using a series of appropriate potassium standards, and TK was measured using a flame photometer (Shanghai Precision Instruments Co., Ltd. 6400 A).

Soil for available nitrogen (mg · kg⁻¹) analysis was ground to ∼2.00 mm. First, 2.0 g of soil were added to 0.2 g Devarda’s alloy, and the sample was digested with 2 ml 0.01 mol · L⁻¹ boric acid solution and 11.5 ml 1.8 mol · L⁻¹ NaOH. Second, the solution was diluted with 0.01 M HCl.

Soil for AP (mg · kg⁻¹) analysis was ground to ∼2.00 mm. First, 3.0 g of soil were digested with 50 ml 0.03 N NH₄F–0.025 N HCl solution. Second, the solution was then diluted with 2 ml of 0.06 M boric acid solution, deionized water, and ammonium-vanadomolybdate reagent, and 0.25% p-nitrophenol indicator was added. Third, a standard curve was arranged by reading the absorbance of blank with an ammonium-vanadomolybdate reagent, standards, and samples at 700 nm wave length. Finally, AP concentrations were calculated according to the calibration curve.

Soil for AK (mg · kg⁻¹) analysis was ground to ∼2.00 mm. First, 5.0 g of soil were digested with 50 ml 1 mol · L⁻¹ ammonium acetate solution. Second, the supernatant was collected, and the combined ammonium acetate extract was diluted with 1 N ammonium acetate solution. Third, a calibration curve was plotted by a series of suitable potassium standard, and the extract was measured with a flame photometer at 767 nm wavelength (Anderson & Ingram, 1989).

Data Analysis

The species diversity of each plot was estimated as species richness, the Shannon–Wiener index (H), and Pielou’s evenness index (E). Species richness was calculated as the number of species, while the H and E were calculated according to the following equations

Shannon–Wiener index (H):

Pielou index (E):

Dominant species were defined as the plant species with importance value more than 1.00.

Sørensen’s dissimilarity coefficient were used to estimate species turnover in each 50 m × 50 m plot, and its detail equation is given as below:

Differences in species diversity, species evenness, species turnover, and soil nutrients between two vegetation types were assessed by Wilcoxon’s rank tests. Relationships between plant diversity and soil conditions were evaluated using multiple stepwise linear regression. Selection of the best model depended on statistical methods for model comparisons, namely, on R² and the difference between the models’ Akaike Information Criterion values. The effects of soil nutrients on dominant species in each recovery stage were assessed using linear regression. Analyses were conducted using SAS (SAS Institute, Cary, NC, USA). Critical soil nutrients for different dominant species were evaluated by canonical correspondence analysis using CANOCO version 4.5 for Windows. In addition, a Monte Carlo permutation test was conducted with 499 permutations for the significant tests of variables and ordination axes.

Patterns of Species Diversity Between Shrubs and Forests

In the Shrub, we found a total of 26,926 trees of 132 species in 48 families. The dominant species were Maclurodendron oligophlebium, Hydnocarpus hainanensis, Mischocarpus sundaicus, Aporosa yunnanensis, and Ardisia crassinervosa. In the Forest, we recorded a total of 16,354 trees of 148 species in 52 families. The dominant species were Schefflera heptaphylla, Syzygium acuminatissimum, Xantolis longispinosa, and Koilodepas hainanense (Supporting information, Appendix 1).

All three measures of diversity (species richness, Shannon–Wiener index, and species evenness) were significantly different between the two vegetation types. Shrub plots had lower species diversity than Forest plots (Table 1). Species turnover, however, was not significantly different between the two vegetation types (Table 1).

Table 1.

Comparisons of Species Richness, Shannon–Wiener Index, Species Evenness, and Species Turnover Between the Shrub and Forest.

Patterns of Soil Nutrients Between Shrubs and Forests

Soil nutrients varied between Shrub and Forest: pH, total P, and total K were all significantly higher in Shrub than Forest, whereas OM and available N were significantly higher in Forest than Shrub (Table 2). Other soil nutrients, including total N, available P, and available K, did not differ between the two vegetation types.

Table 2.

Comparisons of Soil Conditions Between Shrub and Forest Sites: OM, TN, TP, TK, AN, AP, and AK.

Effects of Soil Nutrients on Plant Diversity

Species diversity was significantly correlated with OM, total N, total K, total P, and available K using multiple regression (Table 3); while species evenness, however, was significantly correlated with OM and available P in both vegetation types.

Table 3.

Multiple Regression Analyses Between Soil Nutrients and Species Richness, Shannon–Wiener Index (H), and Species Evenness (E) in Shrub and Forest: pH, OM, TN, TP, TK, AN, AP, and AK.

Effects of Soil Nutrients on Dominant Species

Canonical correspondence analysis revealed a significant correlation between soil nutrients and ordination axes (Monte Carlo permutation test, p = .002). In total, OM, soil K, P, and N contributed largest effects for dominant species in Forest and Shrub (Figure 2; Supporting Information, Appendix 2). In Shrub, the abundance of the dominant species Arytera littoralis, Canthium horridum, Litsea glutinosa, Ficus hispida, and Sapindus saponaria declined, while the abundance of Sterculia lanceolata, Cryptocarya chinensis, and Garcinia oblongifolia increased, with changes in soil nutrients (Supporting Information, Appendix 3). Similarly, in Forest, the abundance of Syzygium tephrodes, Syzygium buxifolioideum, Xantolis longispinosa, and Psychotria asiatica declined, while the abundance of Maclurodendron oligophlebium, Drypetes arcuatinervia, Camellia furfuracea, and Hydnocarpus hainanensis increased along the soil nutrient gradient (Supporting Information, Appendix 4).

Figure 2.

Canonical correspondence analysis between dominant species abundance (black species abbreviations) and soil nutrients (red nutrient abbreviations) in secondary tropical evergreen monsoon forest (Panel (a)) and shrubs (Panel (b)).

Patterns of Soil Nutrients in Forests and Shrubs

Soil nutrients are considered as one of the main factors limiting tropical forest structure, primary productivity, and other biological processes such as plant root allocation, growth, and litter production (Vitousek, Porder, Houlton, & Chadwick, 2010; Wright et al., 2011; Zhang et al., 2015). As we expected, we found differences in soil nutrients between the early- (Shrub) and mid- (Forest) successional sites. For example, soil OM was greater, but pH was lower in Forest compared with Shrub. Similar trends were also observed by Sparrius and Kooijman (2013) in inland dunes.

Total p values were significantly higher in Shrub than Forest. Soil P is an important nutrient regulating plant growth and development in forest ecosystems (Chen, Chondron, Davis, & Sherlock, 2000; Paul & Clark, 1997). Moreover, forest productivity at the late recovery stages has been found to be limited by P (Cuevas & Medina, 1988; Gehring, Denich, Kanashiro, & Vlek, 1999), particularly in tropical forest ecosystems (Asner, Townsend, & Bustamante, 1999; Porder, Vitousek, Chadwick, Chamberlain, & Hilley, 2007). A previous study has shown that as an inducible enzyme, phosphatase can enhance the demand for P during forest recovery (Chen, Chondron, & Xu, 2008), which may explain the increased available P concentrations in forests in our study.

There was no significant difference in total N between the two vegetation types, but there was in available N. Increases in nitrogen availability during forest recovery are also reported in other studies (Erickson, Keller, & Davidson, 2001; Lamb, 1980; Maloney, Garten, & Ashwood, 2008; Robertson, 1984). This increase in N may be because uptake of soil N by aboveground vegetation in shrubs is higher than forests. In Forest, however, the N uptake could be compensated by the available N from litter mineralization and by inputs from N-fixing species. In addition, as a relatively young ecosystem, shrubs generally lack atmospherically derived elements, especially N, but is rich in rock-derived elements such as P (Broadbent et al., 1999). It is reasonable to assume that in shrubs, primary production accumulation is constrained by N supply (Zhang et al., 2015). It is also expected that high total P in shrubs could partly compensate for the low N availability, which has also been observed in previous studies (Du, Pan, Li, Hu, & Wang, 2011; Huang et al., 2013).

Although K is usually abundant in the environment (Zheng & Shangguan, 2007), it can be more easily leached than N or P, and up to 70% of all studied terrestrial ecosystems are K limited (Sardans & Peñuelas, 2015). Furthermore, a significant decline of total K and nonsignificant increase in available K during forest recovery implies that processes such as translocation or nutrient uplift allow maintenance of higher available K levels in the topsoil (Table 2; Barré, Berger, & Velde, 2009). A previous study showed that soil available K concentrations can be lower than available P and available N concentrations (Sardans, Peñuelas, Prieto, & Estiarte, 2008), which contrasts with our findings. Changes in total K and available K, therefore, may be caused by drought during forest recovery (Table 2; Sardans & Peñuelas, 2007).

Correlations of Soil Nutrients With Species Diversity and Dominating Species Abundance in Forests and Shrubs

In global biomes, the broad-scale variation in patterns of species richness is often controlled by contemporaneous climate (Vázquez-Rivera & Currie, 2015) and the water–energy balance over ecological time (Kreft & Jetz, 2007). However, our study was performed at the local scale in sites with similar climate, topography (such as aspect and inclination of slope), and soil parent materials. We therefore assumed that soil nutrient availability is a critical factor influencing species distributions. Indeed, we found that species richness in both successional stages was significantly associated with nutrients such as OM, total N, and total P (Table 3). Nevertheless, OM and total N may be the main limiting resources associated with the changes in species diversity during forest recovery. Indeed, studies show that that community species compositions even within successional forests are likely influenced by the availability of soil nutrients (Guariguata & Ostertag, 2001). One possible reason why OM and total N are the main soil elements responsible for the changes in species diversity is that, as forest recovery proceeds, forest net primary productivity increases, which may cause a loss of N in soils. Higher OM and total N content can balance this loss of soil N, increasing species diversity, and dominant species abundance (Schmidt, Veldkamp, & Corre, 2015). This is especially the case for OM, which contains considerable amounts of N (Long, Yang, & Li, 2012). Thus, an increase in OM may compensate for the nonsignificant change in total N, ultimately resulting in an increase in available N during recovery, thereby increasing species diversity and dominant species abundance from early to late succession.

Dominant Plant Species and Their Responses to Soil Nutrients

Species composition of local communities is often affected by environmental gradients (Mason, de Bello, Mouillot, Pavoine, & Dray, 2013). In our local-scale study, a gradient in soil nutrient concentration played a large role in species coexistence and replacement. A previous study showed that as forest succession proceeded, the main soil nutrients limiting dominant species changed (Zhang et al., 2015). Although in our study, this phenomenon might have been caused by soil conditions in the two stands, it also reflects that the requirements of plant species for soil nutrients in different successional stages differ (Huang et al., 2013).

The two forest successional stages had different dominant species (Supporting Information, Appendixes 1, 3, and 4), which might have been caused by differential responses of dominant species to soil nutrients (Perroni-Ventura et al., 2013). In shrubs, the early-successional stage, the abundance of species such as Arytera littoralis, Litsea glutinosa, and so on, declined, while the other species increased. In contrast, in Forest, which represents a mid-successional stage, the abundance of species, for example, Syzygium buxifolioideum, Litsea glutinosa, and so on, declined, while the other species increased. These results suggest that late recovery species are more able to adapt to the changing understory environment (Huang et al., 2013). The response of dominant species to limiting soil nutrients could predict the destiny of dominant species in late-successional or climax forest. Thus, in shrubs, we would predict the following species to be dominant: Syzygium tephrodes, Maclurodendron oligophlebium, Syzygium buxifolium, Aporosa yunnanensis, Hydnocarpus hainanensis, Xantolis longispinosa, and Ardisia crassinervosa; whereas in forest, we would predict Syzygium buxifolium, Elaeocarpus sylvestris, Litsea variabilis, Dillenia turbinata, Ancistrocladus tectorius, Bridelia balansae, Hancea hookeriana, Litsea verticillata, Schefflera heptaphylla, Casearia membranacea, and Olea dioica. Decline of different dominant species in each successional stage differed (Supporting Information, Appendices 3 and 4). The species with higher decline rates may be eliminated rapidly in the future forest recovery, whereas the species with lower decline rates might persist for a longer time. This phenomenon may be caused by transient adaption to the appearance of a more heterogeneous environment in later successional stages (Bartels & Chen, 2010). However, these species will likely eventually make space for mature-phase dominant species. As forest ecosystems are highly complex, the factors influencing species coexistence and replacement during recovery may be extremely diverse (Lin et al., 2016). Thus, further studies are required to fully explore the species coexistence mechanisms in secondary forest recovery processes.