Forest structure and species composition
A 1 ha plot (100 x 100 m) was randomly established in each of the MSF and MTF in 1980. Within each plot, twenty circular subplots (113 m2) were randomly located. All trees ò0.1 m in DBH (diameter at breath height) in each subplot were identified and the DBHs were recorded.
Litterfall and ground litter biomass
Twenty 1 m2 baskets lined with 1 mm2 mesh fiberglass screen were randomly placed in each 1 ha plot at 1 m above ground. Litterfall was collected biweekly for 1 yr between November 1980 and October 1981. All litter samples were separated into leaves, flowers, fruit, wood, and miscellaneous materials (mostly bark), oven-dried at 70 Centigrade for 72 hrs, and weighed. Ground litter was collected from 0.25 m2 subplots in both the dry season (March, 40 subplots) and the wet season (September, 20 subplots) randomly located in each of the 1 ha plots. Each sample was separated into wood and miscellaneous categories, oven-dried at 70 Centigrade for 72 hrs, and weighed.
Leaf decomposition
A total of 140 litterbags were constructed among which 70 litterbags were filled with fresh leaves collected from the MSF and 70 from the MTF (Blair et al., 1990). These fresh leaves represented the tree species composition in natural litterfall in April at both sites. The species composition of litterfall was determined using leaffall in the 1-m2 randomly placed baskets in each plot. Fresh leaves were collected for each forest within 24 hours of senescence in May and sorted by species. Leaves of the 13 most common species from each forest, representing 82-91 percent of the total fresh leaf mass, were placed in 1- mm2 mesh fiberglass screen bags (0.2 by 0.25 m) in proportion to their biomass in litterfall (Table 1). Leaves of the remaining 25 species were put in a miscellaneous category and randomly selected to obtain a total of 10-g fresh leaf material (4.6 g mean oven-dry mass) in each bag.
For each leaf type, 35 litterbags were placed equally in 5 randomly selected subsites in the 1 ha plots where leaves were collected. In order to separate the effects of leaf chemistry on leaf decomposition from those of biotic and abiotic environmental conditions between the two forests, the other 35 litterbags for both leaf types were together placed in a 10 x 10 m plot in the MTF at five randomly selected locations. The 1 ha plots covered an area with heterogeneous geographical locations including ridges and valleys, whereas the 0.01 ha plot was located on a ridge top and upper slope position. One randomly selected litterbag was collected after 0, 7, 14, 28, 60, 120, and 300 days in the field at each site from each plot. There were 2 (leaf types) x 2 (incubation sites) x 5 (replicates) x 7 (collections) = 140 litterbags.
Small roots which had entered the litter bags were carefully removed in the laboratory. The remaining litter in each bag was oven dried at 70 Centigrate for 72 hours and weighed to determine dry mass loss. Dry litter was then ground with a Wiley mill through a 0.85 mm (20 mesh) stainless steel sieve. One gram of the ground leaf material from each litterbag was digested with H2O2 and concentrated HNO3 (Luh Huang and Schulte, 1985) before analyzing for P, K, Ca, and Mg content with an atomic absorption and emission spectrophotometer. Total C and N in leaf material were analyzed by direct combustion in a C-H-N analyzer (Carlo-Erba Model 1106).
Data analyses
Basal area, tree density, and leaffall for each species were calculated for each forest. Total litterfall rate and ground litter biomass of all species were also calculated for each forest. The analysis of variance (ANOVA) was employed to test the differences in basal area, tree density, leaffall rate, and ground litter biomass between the forests or the seasons (SAS, 1987). Leaf decay rates were calculated using a single exponential model Mt = M0e-kt, where Mt was the remaining mass of leaf materials in a litter bag at time t and M0 was its initial mass. Slope k was obtained using linear regression after taking the natural logarithms of the equation. Multivariate tests were used to examine differences in k values between the two forests and between sites (SAS, 1987). The significance level was set at a = 0.05.
Percentage of initial weight remaining was calculated using oven-dried weight. Percentage of initial element remaining in litterbags was obtained by multiplying the ratio of total element at time t (t = 0, 7, 14, 28, 60, 120, and 300 days) to that at time 0 by 100. Total element content in a litterbag was calculated as the product of element concentration and the oven- dry weight. Differences in elemental concentrations between forests and among decomposing dates were tested with ANOVA. Where significant differences were obtained by ANOVA, Bonferroni t-test (SAS Inc., 1987) was used to compare these differences in elemental concentrations. The significance levels for ANOVA and Bonferroni t-test were both set at 0.05. Plots of residuals vs. predicted values indicated that no variables significantly violated the homogeneity assumption, with the exception of K concentrations. A non-parametric analysis was employed by ranking K concentrations prior to performing ANOVA and Bonferroni t test.
