This research concerns the investigation of an alternative measurement to directly indicate the severity of torrefaction. Composition and process data from both batch and continuous torrefaction experiments using willow, wheat straw, and cattail biomass were combined and analyzed. The mass yield, which is an indication of torrefaction severity was correlated to the net change in residual carbon concentration (Delta C), and second the total change in mass (Delta M-t) was correlated to net change in mass of carbon per 100 g of feedstock (Delta M-c). Analysis of the experimental data show a polynomial relationship between the dry mass yield (Y-m) and the change residual carbon concentration (SC). This relationship is Y-m = 5.05 Delta C-2 - 3.96 Delta C + 0.98 (R-2 = 0.89). The uncertainty in this correlation is +/-7.3% (w/w). The relation between total change in mass (AMt) and change in carbon mass (AMc) meanwhile was found to fit a linear model by AMc = 0.36ANIt + 1.04 with a coefficient of determination of 0.96. Each of these models was then validated by introducing experimental data from numerous published materials focused on biomass torrefaction. That data included bench and pilot scale experiments that examined a wide range of biomass including soft and hardwoods, grasses, and agricultural residues. Analysis of the combined data set confirmed a second order polynomial model is predictive of the mass yield based on the change in carbon concentration where both parameters are expressed on a dry, ash-free basis. This model, Delta M = 4.29 Delta M-2 3.66 Delta M + 0.98 (R2 = 0.935), is predictive for torrefaction experiments with initial masses of 500 g and greater and for mass yields as low as 60%. The uncertainty in this second correlation is +/-4.6% (w/w). Since this expression relies only on the concentration of carbon in each of the feed and output product streams, this model could be used to predict mass yield of continuous torrefaction in real-time if the carbon content of each stream were sampled periodically and if the raw biomass ash content were sampled intermittently. The relation between change in total mass and change in mass of carbon meanwhile was confirmed with the inclusion of all of the literature review data; the linear regression model was found to be Delta M-c = 0.37 Delta M, + 1.26 which had an R-2 of 0.93. In practical terms, this expression indicates that for the average of these experiments, the first 3.4% of mass loss in torrefaction occurs without loss of significant carbon, and below 97% mass yield, carbon will consistently represent close to 37% of total mass loss, a figure which appears to hold to as low as 40% mass yield (w/w).