Landscape-scale variation in the structure and biomass of the hill dipterocarp forest of Sumatra: Implications for carbon stock assessments

One of the first steps in estimating the potential for reducing emissions from deforestation and forest degradation (REDD) initiatives is the proper estimation of the carbon components. There are still considerable uncertainties about carbon stocks in tropical rain forest, coming essentially from poor knowledge of the quantity and spatial distribution of forest biomass at the landscape level. We evaluate the influence of site and topography on forest structure, biomass and carbon stock over a forested landscape in Sumatra. Sixteen sites were selected across a landscape of ca. 1000 km(2). and within each site, a network of small 0.1 ha plots was laid down perpendicular to the major topographical gradient. Structural parameters for trees with diameter bigger or equal to 10 cm and major life forms were recorded for each plot. The total sampled area was 70.2 ha. The above-ground biomass (ACBM) in the 16 sites ranged from 271 +/- 19 to 478 +/- 38 Mg ha(-1), with a mean of 361 +/- 7 Mg ha(-1). This value is similar to the global 'tropical wet forest' estimate of Keith et al. (2009), but 23% higher than the biome default value given in the IPCC (2006) guidelines. Plots on acid tuff were the most distinctive in having fewer and smaller trees. with a resulting low AGBM. The number of stems was low, but the average tree diameter was larger on the nutrient-rich young volcanic and esitic soils. Landscape-scale variability of forest structure and dynamics was considerable. Within sites, there was little variation of mean height or mean diameter of the trees, but significant differences in stem density, basal area and AGBM between land facets. Both topography and geology appeared to affect the dynamics and the biomass of the forest. At the same time, our findings confirm the need for caution in extrapolating data from small plots to larger spatial scales when estimating carbon stocks fora given forest type ora region. Our results for C stock values for Sumatra lowland and hill dipterocarp forests (135-240 Mg ha(-1), with a mean of 180 Mg ha(-1)) showed higher variability in C stock values at landscape level than generally reported in the literature. Based on our results, stratified sampling is recommended for biomass and carbon estimation. Within an error range of 6-8% of the ACBM, a minimum area of 4-6 ha should be sufficient to estimate biomass with satisfactory accuracy at the landscape scale. Networks of several small plots across landscape are preferable to fewer larger plots, providing that the sampling is representative of the land facets in the area. (C) 2009 Elsevier B.V. All rights reserved.