A better understanding of the reasons for variation in tropical rain forest (TRF) structure is important for quantifying global above-ground biomass (AGBM). We used three data sets to estimate stem number, basal area, and AGBM over a 600-ha old-growth TRF landscape (La Selva, N.E. Costa Rica). We analyzed the effects of soil type, slope angle, topographic position, and different sample designs and measurement techniques on these estimates. All three data sets were for woody stems greater than or equal to 10 cm in diameter. Estimated AGBM was determined from stand-level measurements using Brown's (Brown, 1997) allometric equation for Tropical Wet: Forest trees. One data set was from three subjectively-sited 4-ha plots (the 'OTS plots'), another was based on 1170 0.01 ha plots spaced on a regular grid (the 'Vegetation map plots'), and the third was from 18 0.5 ha plots (the 'Carbono plots') sited to provide unbiased samples of three edaphic conditions: hat inceptisol old alluvial terraces; Rat ultisol hill-tops; and steep ultisol slopes. Basal area, estimated AGBM and the contributions of major life forms were similar among studies, in spite of the differences in sampling design and measurement techniques. Although the Carbono plots on Rat inceptisols had significantly larger and fewer trees than those on ultisols, AGBM did not vary over the relatively small edaphic gradient in upland areas at La Selva. On residual soils, the largest trees were on the flattest topographic positions. Slope angle per se was not correlated with basal area or AGBM within the residual soils. Errors introduced by palm and liana lift: forms, as well as hollow trees, did not significantly affect AGBM estimates. In contrast, the methods used to measure buttressed trees had a large impact. Plot sizes of 0.35-0.5 ha were sufficient to achieve coefficients of variation of <12% for basal area with only six replicates in a given edaphic type. AGBM estimates ranged from 161 to 186 Mg/ha. These low values appear to be mainly due to the Tropical Wet Forest allometry equation used. This in turn may be indicative of a real and substantially lower ratio of biomass/basal area in Tropical Wet Forest than in Tropical Moist, as previously noted by Brown (1996). Our results indicate that for upland TRF landscapes with levels of environmental variation similar to La Selva, AGBM will be relatively insensitive to soil type and topography. However, because topography and soil type had much stronger effects on stem size, stand density, and spatial heterogeneity of sterns, stand dynamics may be more sensitive than AGBM to this range of conditions. We recommend that future studies of landscape-scale forest structure employ stratified sampling designs across major environmental gradients. Unbiased sampling with replication, combined with consistent and well-documented measurement techniques, will lead to a greatly improved understanding of the magnitude of and reasons for variation in forest structure and AGBM within TRF landscapes. (C) 2000 Elsevier Science B.V, All rights reserved.
