Public interest in local food continues to grow, but few analyses have examined the capacity for the US population to be supplied through local and regional food systems. This paper extends earlier work that demonstrated a method for mapping potential foodsheds and estimating the potential for New York to meet the food needs of the state's population centers. It provides a methodology for addressing the question, 'If land is limited, which foods should be grown locally?' A spatial model was developed to allocate the available agricultural land of New York State (NYS) to meet in-state food needs for six distinct food groups (grains, vegetables, fruits, dairy, meat and eggs) across the eight largest population centers. An optimization routine was used to allocate land to maximize economic land use value (LUV). Eleven scenarios were examined, ranging from a baseline level of consumption of New York produced foods to a 100% local diet. Across the 11 scenarios, the amount of food supplied, the LUV attained, and the area of land allocated increased as the 'willingness' to consume local products increased. This approach dictated that land was preferentially devoted to higher-value food groups relative to lower-value groups, and no scenario used all available land. Under the 100% local scenario, 69% of total food needs (on a fresh weight basis) were supplied in-state with an average food distance of 238 km. This scenario provided food from only four of the six groups, namely, dairy, eggs, fruit and vegetables. These results suggest that a much larger proportion of total food needs (on a weight basis) might be provided from in-state production than was found in previous work. LUV serves as a compelling optimization function, and future work should investigate the degree to which maximizing returns to land complements or conflicts with social and environmental goals of local and regional food systems.