Biomass represents a promising renewable energy opportunity that may provide a more sustainable alternative to the use of fossil resources by minimizing the net production of greenhouse gases. Yet, allometric models that allow the prediction of biomass, biomass carbon (C) and nitrogen (N) stocks rapidly and non-destructively have not been developed for tropical perennial C₄ grasses currently under consideration as potential feedstocks in Hawaii and other subtropical and tropical locations. The objective was to develop optimal allometric relationships and site-specific models to predict aboveground biomass, C and N content of individual stalks of one-year sugarcane (Saccharum officinarum cv. 65-7052), energycane (Saccharum officinarum x Saccharum rubustom cv. MOL-6081), Napier grass (Pennisetum purpureum cv. bana grass), and two-year sugarcane cultivated as biofuel feedstock. For all allometric equations, a simple power model provided the optimal prediction of aboveground biomass and its C and N stocks. Stalk diameter and dewlap height were the most effective predictors (R² > 90) for aboveground biomass, biomass C and N stocks. The models were tested and validated for different locations. Critical reassessment of the quality and the robustness of these models across biofuel feedstock types was provided using large dataset. Although optimal prediction required the development of site-specific allometric relationships, these equations can also be applicable for establishing aboveground reference levels of potential feedstock quantities and C content which could be used for measuring, reporting and verifying the C footprint and C budget for any biofuel cropping system.