Biomass estimation in the Tapajos National Forest, Brazil - Examination ofsampling and allometric uncertainties

changes in the biomass of Amazon region forests represent an important comp onent of the global carbon cycle but the biomass of these forests remains p oorly quantified. Minimizing the error in forest biomass estimates is neces sary in order to reduce the uncertainty in future Amazon carbon budgets. We examined forest survey data for trees with a diameter at breast height (DB H) greater than 35 cm from four plots with a total area of 392 ha in the Ta pajos National Forest near Santarem, Para, Brazil (3 degrees 04'S, 54 degre es 95'W). The average frequency of trees greater than 35 cm DBH was approxi mately 55 ha(-1). Based on tree diameters, allometric relations, and publis hed relations for biomass in other compartments besides trees of DBH > 35 c m., we estimated a total biomass density of 372 Mg ha(-1). We produced a hi ghly conservative error estimate of about 50% of this value. Trees with dia meters greater than 35 cm DBH accounted for about half of the total biomass . This estimate includes all live and dead plant material above- and below- ground with the exception of soil organic matter. We propagated errors in s ampling and those associated with allometric relations and other ratios use d to estimate biomass of roots, lianas and epiphytes, and necromass. The ma jor sources of uncertainty in our estimate were found in the allometric rel ations for trees with DBH greater than 35 cm, in the estimates of biomass o f trees with DBH less than 35 cm., and in root biomass. Simulated sampling based on our full survey, suggests that we could have estimated mean biomas s per hectare for trees (DBH greater than or equal to 35 cm) to within 20% (sampling error only) with 95% confidence by sampling 21 randomly selected 0.25 ha plots in our study area. (C) 2001 Elsevier Science B.V. All rights reserved.