Normalization of wood density in biomass estimates of Amazon forests

Wood density is an important variable in estimates of biomass and carbon flux in tropical regions. However, the Amazon region lacks large-scale wood-density datasets that employ a sampling methodology adequate for use in estimates of biomass and carbon emissions. Normalization of the available datasets is needed to avoid bias in estimates that combine previous studies of wood density that used wood sampling at diverse positions in the bole or with various methods of density determination. This paper examines the question of whether regressions for radial variation and for variation in wood density along the bole, both developed in dense forest in central Amazonia (CA), are suitable for the open forests in southern Amazonia (SA) that are currently the target of most of Amazonia's deforestation activity. The wood density of the heartwood and density of full disks or slices (bark, sapwood and heartwood) in each tree were measured to assess the radial variation. For variation along the length of the bole, wood densities at breast height and at the top of the bole were used. Moisture content of the bole was measured in SA and compared with values reported by studies from CA in similar dense forest. Comparing regressions that predict full-disk density from heartwood density, the pattern of radial variation differs slightly and significantly between the two forest types (ANCOVA p = 0.006); the slopes have similar values but the intercepts differ. Variation along the bole in the two forest types does not differ significantly (p = 0.144), so the CA model for predicting mean bole density from the density of a slice at breast height gives an unbiased estimate of the mean bole density when applied to SA trees. In SA the mean moisture content of the bole was 0.416 ( +/- 0.068 S.D.; n = 223 trees). Moisture content of the bole had a strong inverse relationship with basic wood density (r = -0.77), which explains the lower moisture content in the trees in CA relative to SA. A much weaker inverse relationship was found between moisture content and green wood density (r = -0.292). The relationship between wood basic density and green ('fresh') density presented in this study provides an alternative means of obtaining basic wood density directly in the field when oven drying of samples is not possible. (c) 2008 Elsevier B.V. All rights reserved.