Heterogeneity of Soils and Vegetation in an Eastern Amazonian Rain Forest: Implications for Scaling Up Biomass and Production

Transferring fine-scale ecological knowledge into an understanding of earth system processes presents a considerable challenge to ecologists. Our objective here was to identify and quantify heterogeneity of, and relationships among, vegetation and soil properties in terra firme rain forest ecosystems in eastern Amazonia and assess implications for generating regional predictions of carbon (C) exchange. Some of these properties showed considerable variation among sites; soil textures varied from 11% to 92% clay. But we did not find any significant correlations between soil characteristics (percentage clay, nitrogen N], C, organic matter) and vegetation characteristics (leaf area index LAI], foliar N concentration, basal area, biomass, stem density). We found some evidence for increased drought stress on the sandier sites: There was a significant correlation between soil texture and wood δ¹³C (but not with foliar δ¹³C); volumetric soil moisture was lower at sandier sites; and some canopy foliage had large, negative dawn water potentials (ψ_ld), indicating limited water availability in the rooting zone. However, at every site at least one foliage sample indicated full or nearly full rehydration, suggesting significant interspecific variability in drought vulnerability. There were significant differences in foliar δ¹⁵N among sites, but not in foliar % N, suggesting differences in N cycling but not in plant access to N. We used an ecophysiological model to examine the sensitivity of gross primary production (GPP) to observed inter- and intrasite variation in key driving variables—LAI, foliar N, and ψ_ld. The greatest sensitivity was to foliar N; standard errors on foliar N data translated into uncertainty in GPP predictions up to ±10% on sunny days and ±5% on cloudy days. Local variability in LAI had a minor influence on uncertainty, especially on sunny days. The largest observed reductions in ψ_ld reduced GPP by 4%–6%. If uncertainty in foliar N estimates is propagated into the model, then GPP estimates are not significantly different among sites. Our results suggest that water restrictions in the sandier sites are not enough to reduce production significantly and that texture is not the key control on plant access to N. Received 28 June 2001; accepted 13 March 2002.