UNCERTAINTY IN THE BIOMASS OF AMAZONIAN FORESTS - AN EXAMPLE FROM RONDONIA, BRAZIL

A critical factor in estimating the contribution of tropical deforesta tion to nutrient mobilization and to CO2 build-up in the atmosphere is the amount of biomass available to bum. The biomass data for Brazil, a major site for deforestation, are few and of uncertain accuracy. Rec ent international agreements, however, require national inventories of sources and sinks for atmospheric greenhouse gases; such inventories will need better estimates of biomass and their uncertainties. To prov ide additional information on biomass uncertainty and on forest struct ure in southwestern Amazonia, a region of active deforestation, we mea sured in 1988 the diameter, bole and canopy heights of 474 trees cover ing a total of 1 ha (10 000 m(2)) in the Ecological Station of the Sam uel Hydroelectric Reservoir in Rondonia (845'S, 63 degrees 23'W). Usin g allometric equations based on destructively sampled trees, we estima ted the largest biomass component, standing alive aboveground biomass (SAAB), as 285 Mg (dry weight) ha(-1). Fallen trunks and litter were 3 0 Mg and 10 Mg ha(-1), respectively, The sum of these components, 325 Mg ha(-1), is an underestimate of the total biomass because the biomas s of roots, vines, shrubs, and small trees was not measured. Measureme nt error of SAAB is +/- 20%, +/- 57 Mg ha(-1) about the mean (95% conf idence interval), as derived by a Monte Carlo simulation. The SAAB dis tribution among trees is highly skewed: 3% of the trees contain 50% of the SAAB. For forests of similar distributions, sampling units typica lly used for biomass estimates (less than 2000 m(2)) will usually prod uce biomass estimates significantly different from those of larger uni ts. Based on subsamples of our data, sampling units of 1000 m(2) or sm aller had at least a 75% chance of being outside the confidence interv al of the global mean (228-342 Mg ha(-1)) derived from Monte Carlo sim ulation. To improve estimates of SAAB in similar forests a sampling pr ogram should focus on emergent and large canopy trees, the dominant co ntributors to biomass.