Biomass is being researched as a possible alternative to fossil sources of energy, in order to avoid externalities from fossil fuel use that affect the environment and the economy. Some biomass-based energy production systems may produce unwanted externalities in their own right, such as increasing the production pressure on the agricultural land base, resulting in a rise in prices of food commodities. Using marginal land for biomass production has been suggested as a solution. However, the definition of what constitutes marginal land is poorly understood. This paper provides a theoretical foundation for identification of marginal lands, and analyzes recent literature to assess how current usage of the term marginal correspond to the theoretical framework. Then, the paper devises empirical models that test possible methods of identification of the extensive margin of agricultural land in 19 counties in the state of Michigan. The models find that dynamic variables such as price changes have a statistically significant effect on land use change into and out of cropland. Land quality and regional effects are also statistically significant.