The Community Structure And Assembly Rules Of Tropical Cloud Forests In Hainan Island, South China

Tropical forests are terrestrial ecosystems harboring the most abundant species and complicated structures, and play important roles in the regional and global biodiversity conservation and ecological function maintenance. Studies on the mechanisms of species coexistence in these ecosystems have been a vitally important topic in community ecology in recent years. Although some studies have been carried out in the low altitudinal tropical forests, species assembly and community structuring in the high altitudinal tropical cloud forests are poorly understood. Compared with the lower altitude tropical forests, environmental conditions in tropical cloud forests are quite different, characterized by frequent fog, low temperature, high humidity and strong winds. Moreover, trees in these forests are typically more deformed and elfin with small sturdy leaves, few buttress,and thick cover of epiphytes. These special environmental and physiognomical features may confine a unique rule of community assembly. The small stature of trees in the tropical cloud forests make it possible to precisely measure some important functional traits (such as specific leaf area and maximum species height). Tropical cloud forests in Hainan Island are distributed at more than 1200 m altitude in Bawangling Mt., Jianfengling Mt., Wuzhishan Mt., Yinggeling Mt., etc, including tropical montane evergreen forest (TMEF) and tropical (montane) dwarf forest (TDF or TMDF). To explore the structuring and assembly rules of tropical cloud forest communities, we investigated the species diversity and environmental conditions, and measured the specific leaf area (SLA), dbh and height for 5765 individuals trees and shrubs (dbh≥1 cm) in TMEF and TMDF in Bawangling National Natural Reserve, Hainan island, South China. Then, we compared the environmetal conditions, community stucture and assembly rules for the two cloud forest types. The main results are as follows.(Ⅰ) Environmental conditions and community features of the tropical cloud forests1. Daily photosynthetically active radiation (PAR) showed a unimodal curve both in TMEF and TMDF, but PAR was significantly lower in TMEF than TMDF. From May to October, mean daily air temperature differed significantly between TMEF and TMDF and showed a unimodal curve in the two forests, with average values of (21.76±2.44)°C and (19.33±1.03)°C, respectively. Additionally, mean daily relative humidity differed significantly between TMEF and TMDF and showed an"inverse S"curve; average values were (88.44±2.90) % and (97.71±0.80) %, respectively. TMEF had higher total nitrogen, total phosphorous, available nitrogen, organic matter, pH and soil thickness, but lower total potassium and available phosphorous than TMDF. Slope, cover of exposed rock and altitude were lower in TMEF than TMDF. Principal component analysis and Pearson's correlation analysis indicated that air temperature, soil phosphorous, potassium, nitrogen and the three topographic factors were the most important predictors of distribution of these tropical cloud forests.2. The dominant species in TMEF were Syzygium araiocladum, Distylium racemosum, Cyclobalanopsis disciformis, Psychotria rubra and S. buxifolium, and the dominant species in TMDF were D. racemosum, S. buxifolium, Engelhardtia roxburghiana, P. rubra and S. lancifolia. The common dominant families for the two forest types were Lauraceae, Symplocaceae, Rubiaceae, Fagaceae and Oleaceae. The common dominant genera for the two forest types were Symplocos, Cyclobalanopsis, Lithocarpus and Beilschmiedia. The S?rensen species similarity index for the two forest types was 0.71. The mean stem density for saplings (1 cm≤dbh﹤5 cm) and small trees (5 cm≤dbh﹤10 cm) was significantly lower in TMEF than TMDF, while there were no differences in mean stem density for adult trees (10 cm≤dbh) between these two forest types. Mean dbh for small trees and adult trees were significantly higher in TMEF than TMDF, while mean dbh for saplings were significantly lower in TMEF than TMDF. Mean plant height for saplings, small trees and adult trees were significantly higher in TMEF than TMDF.(Ⅱ) Community assembly based on species diversity3. The observed species richness values, as well as the species richness values predicted by 1st order Jackknife estimator, 2nd order Jackknife estimator and bootstrap estimator, were significantly higher in TMEF than TMDF; while the individual abundance was significantly lower in TMEF than TMDF. Compared with power curve and exponential curve, logistic curve was the optimal model simulating the species-area relation for the two forest types. After Brokenstick model, Niche preemption model, Zipf model, Zipf-Mandelbrot model and Neutral theory model were used to simulate the species-abundance distribution for TMEF community and TMDF community, it was found that Zipf-Mandelbrot model was the optimal model predicting the species-abundance distribution in TMEF, while Niche preemption model and Zipf-Mandelbrot model were the optimal models predicting the species-abundance distribution in TMDF. A general linear anaysis model revealed that PAR, air temperature, soil nitrogen and phosphorus were the major environmental filters predicting the species richness and individual abundance between the two forest communities..4. Patterns of species co-occurrence in TMEF and TMDF were assessed at 5 m×5 m, 10 m×10 m, 20 m×20 m and 30 m×30 m plot sizes using four species co-occurrence indices, including the number of checkerboard species pairs, the checkerboard score of matrix, the number of species combinations and the variance ratio. All combined species, species with different abundance classes and species with different dbh classes in TMEF all showed less co-occurrence patterns than null model tests at 5 m and 10 m plot sizes, indicating that species were non-randomly segregated, and competition probably impacted on the community assembly in this forest. The species co-occurrence indices were the highest at 5 m plot size. In contrast, all combined species, species with different abundance classes and species with different dbh classes in TMDF all showed more co-occurrence patterns than null model tests at 20 m and 30 m plot sizes, indicating that species were non-randomly aggregated, and facilitation probably impacted on the community assembly in this forest. The species co-occurrence indcies were the highest at 20 m plot size.5. Patterns of species distribution in tropical cloud forests were assessed using Donnelly nearest neighbor distance index. Percentage of species in regular distribution was higher than that of random distribution and clumped distribution, while percentage of species in clumped distribution was the lowest; percentage of species in random distribution and clumped distribution decreased, while percengate of species in regular distribution increased with increasing dbh. Patterns of distribution of saplings for the common dominant species, S. buxifolium and D. racemosum in both TMEF and TMDF, showed a clumped distribution, their small tress showed a clumped or random distribution, and their adult trees showed a random distribution. A univariate O-ring function analysis revealed that saplings and small trees for D. racemosum were aggregated at less than 10 m plot size, while saplings and small trees for S. buxifolium were aggregated at more than 23 m plot size. A bivariate O-ring function analysis showed that the spatial patterns between small trees and saplings, adult trees and saplings, and adult trees and small trees for D. racemosum were positively associated at less than 5 m plot size. On the contrary, the spatial patterns between small trees and saplings, and adult trees and small trees for S. buxifolium were negatively associated.6. Based on null model tests, (41±4) % and (37±2) % heterospecific trees, (43±4) % and (38±7) % conspecific trees, and (28±2) % and (44±5) % all combined tree species in TMEF and TMDF, showed significant correlations between dbh and nearest neighbor distance, indicating that non-random processes affected on community assembly in tropical cloud forests. (23±3) % and (26±5) % heterospecific trees, (27±5) % and (19±9) % conspecific trees, (17±8) % and (27±7) % all combined tree species in TMEF and TMDF, showed significantly positive correlations between dbh and nearest neighbor distance, indicating that competition impacted on community assembly in the tropical cloud forests. However, negative correlations between dbh and nearest neighbor distance based on null model tests showed that facilitation affected community assembly in the tropical cloud forests. Both importance and intensity of competition statistically increased with decreasing forest productivity from TMEF to TMDF for all the shared species spanning these two forests. (Ⅲ) Community assembly based on functional traits7. Both abundance-weighted mean SLA and plant height were significantly higher in TMEF than TMDF, while phenotypic plasticity in SLA and height was significantly lower in TMEF than TMDF. Multiple linear regression analyses indicated that among the measured environmental factors, both air temperature and soil total phosphorus were significantly correlated with mean SLA and its plasticity, and only air temperature was correlated with mean height. But both air temperature and soil total phosphorus was significantly correlated with the plasticity index of plant height. These results indicate that air temperature and soil phosphorus acted as environmental filters on decreasing SLA and plant height, while increasing the plasticity of the functional traits from TMEF to TMDF, and thus impact on species composition and community structure in tropical cloud forests.8. SLA decreased significantly with increasing solar irradiance within the three study sites in TMDF, and differed significantly among the three sites both for within and among species comparisons. Mean plot SLA, accounting for both within and among species across the three sites, increased significantly in relation to air temperature but not local PAR and soil total phosphorus. Alpha SLA decreased significantly with increasing solar irradiance within the three sites and beta SLA differed significantly among the three sites. The strong relationship between both intra- and interspecific variation in SLA and environmental conditions strongly confirms the role of trait variation in the assembly of plant species into tropical cloud forest communities via environment filtering related to light availability and air temperature.9. Community assembly were examined based on null model tests at 5×5 m, 10×10 m, 20×20 m and 30×30 m plot sizes using a trait-based approach in TMDF. Lower SLA growing forest plants were over-represented within forest communities at 20 m and 30 m plot sizes, and taller growing forest plants were over-represented within forest communities at the four plot sizes. The correlation between effect size of the test statistic (i.e. mean trait value) of null model tests for specific leaf area and air temperature was positive, and that for maximum species height and percentage reduction of PAR was positive, revealing that there is a stress-tolerator advantage for lower specific leaf area species to be adapted to low temperature, and a size-advantage for taller species competing for light. Thus species are non-randomly assembled with respect to both specific leaf area and maximum species height in the tropical montane dwarf forest community, and that these processes are driven by both the temperature stress and biotic competition.10. In conclusion, the low air temperature and low soil phosphorus act as environmental filters on species with different functional traits, affecting their distribution and abundance in tropical cloud forests; meanwhile, both competition and facilitation processes affect the spatial patterning and proportions of individual abundance within a community. The tropical cloud forest communities are mainly assembled by niche-based non-random processes.