The issue of global climate change and an increasing interest in the reduction of fossil fuel carbon dioxide emissions by using forest biomass for energy production has increased the importance of quantifying forest biomass in recent years. The official U.S. forest carbon reporting is based on the forest biomass estimates obtained from the equations, sample tree measurements, and forest area estimates of the U.S. Forest Service, Forest Inventory and Analysis (FIA). These biomass estimates differ from the estimates obtained from regional and other commonly used biomass equations and the difference is more evident in the component biomass estimates. In this dissertation, I assessed the efficiency of different sampling strategies to estimate crown biomass using data collected destructively from sampled trees. In terms of bias and root mean squared errors (RMSE), the stratified random sampling with probability proportional to branch basal diameter was better than other methods when 3 or 6 branches per tree are sampled but a systematic sampling with ratio estimation technique produced the smallest RMSE when 9 or 12 branches per tree are sampled. Total and component aboveground biomass estimates were obtained using the existing approaches and locally fitted equations based on the data collected in this study. The use of existing equations resulted in biased component biomass estimates along with higher RMSE. The locally fitted system of component biomass equations with seemingly unrelated regression (SUR) provided better estimates than existing equations. The need to use other explanatory variables in addition to the diameter at breast height (DBH) to estimate component biomass was justified by decrease in RMSE. Beta, Dirichlet, and multinomial loglinear regressions that predict proportion of biomass in each component were unbiased and produced lower RMSEs compared to the SUR methods for most of the species-component combinations. Three different methods for adjusting regional volume and component biomass equations were applied. All the adjustment methods were able to improve the performance of regional equations. Based on the leave one out cross validation, the RMSEs in cubic volume including top and stump (CVTS) and component biomass estimation were similar for the adjustments from a correction factor based on ordinary least square (OLS) regression through origin and an inverse approach. The adjustment based on OLS with intercept did not perform as well as the other two adjustment methods. When only one tree is available for calibration of regional models, we found it useful to use the tree with maximum DBH to calibrate regional CVTS and bark biomass equations and the dominant tree to calibrate bole, foliage, and branch biomass rather than to use randomly selected one tree.