Methods for obtaining accurate, spatially explicit estimates of biomass density in tropical forests are required to reduce uncertainties in the global carbon cycle, and to support international climate agreements and emerging carbon markets. Three-dimensional (3-D) remote sensing techniques sensitive to the vertical structure of vegetation provide a unique opportunity for mapping and monitoring forest carbon stocks across large areas in the tropics. However, approaches to forest biomass estimation from remotely sensed structure have yet to be fully developed to deliver the required biomass accuracy. In this research, we use airborne laser scanning (ALS), space-based lidar observations (ICESat/GLAS), and detailed in situ measurements made at the Tapajós National Forest, Brazil, to advance methods of biomass estimation in tropical regions from remotely sensed structure. The overall objectives were to (1) test and refine methods for the extraction of structural information from lidar data; (2) determine the accuracy of lidar estimates of structure in relation to detailed field measurements of vertical structure; and (3) develop and validate structure-based models for optimal prediction of aboveground biomass. Because remote sensing approaches to biomass estimation begin with field estimates of biomass, field plot data collected at Tapajós were also used to gain a better understanding of the uncertainty associated with plot-level biomass estimates obtained specifically for calibration of remote sensing data. This included an evaluation of the error resulting from spatial disagreement between field and remote sensing measurements (i.e., co-location error), and the error introduced when accounting for temporal differences in data acquisition. Results show that a new approach to biomass estimation based on Fourier transforms of lidar profiles significantly improves predictions of aboveground biomass ranging from 2 to 538 Mg ha⁻¹ in primary and secondary forests. Data from two different regions in the Amazon were used to demonstrate and test this method in a range of conditions. The improvement in biomass estimation performance was consistent across sites and the approach was integrated in a multi-stage scaling strategy to biomass estimation to produce a wall-to-wall map of biomass across a large area in the Amazon.