EDAPHIC FACTORS AND THE LANDSCAPE-SCALE DISTRIBUTIONS OF TROPICAL RAIN FOREST TREES

Abstract

Tropical rain forests have the highest tree diversity on earth. Nonrandom spatial distributions of these species in relation to edaphic factors could be one mechanism responsible for maintaining this diversity. We examined the prevalence of nonrandom dis- tributions of trees and palms in relation to soil type and topographic position (''edaphic biases'') over a mesoscale (573 ha) old-growth tropical rain forest (TRF) landscape at the La Selva Biological Station, Costa Rica. All trees and palms $10 cm diameter were mea- sured and identified in 1170 circular 0.01-ha plots centered on an existing 50 3 100 m grid. Topographic position was classified for each plot, and slope and aspect were measured. Soil type data were taken from a previous study (Clark et al. 1998). A total of 5127 trees and palms were identified in 267 species. Detrended Correspondence Analysis and Canon- ical Correspondence Analysis showed that highly significant edaphic gradients were present, with swamp or highly fertile soils separated from the less fertile, well-drained upland soils. Species composition remained significantly related to topographic position when soil type was controlled for. The main floristic gradients were still significant when flooded sites were excluded from the analyses. Randomization tests on a weighted preference index were used to examine the relations of individual species to soil types and, within the dominant soil type, to topographic position. Of the 132 species with N $ 5 individuals, 33 showed significant associations with soil type. Within the dominant soil type, 13 of 110 analyzable species were nonrandomly associated with one or more topographic positions. For a variety of reasons, including issues relating to sample size and adequate edaphic characterization of landscapes, we suggest that the ;30% of species shown to be edaphically biased in this study is an underestimate of the true degree of edaphically related distri- butional biases. To evaluate this hypothesis will require mesoscale vegetation sampling combined with quantitative soil analyses at the same scale in a range of tropical rain forests. If edaphic distributional biases are shown to be common, this suggests that edaphically linked processes leading to differential recruitment are similarly common.