538 – Environmental Exposure and Land Use Modeling
A Parcel-Level Model for U.S. Agricultural Land Use
Jonathan Lisic
United States Department of Agriculture
Agricultural land use prediction at the parcel level is important in many applications such as natural resource management, survey development, and agricultural economics. Unfortunately, in the United States, there is no publicly available and complete set of agricultural land use parcels covering the entire contiguous 48 states. Within this research, a computationally feasible method for forming and classifying parcels from geospatial data sources using a modified mean shift algorithm will be presented. These parcels will be incorporated into a Bayesian hierarchical multinomial probit model to predict agricultural land use. A key component of this model is the incorporation of spatially correlated crop rotation preferences through a spatial autoregressive prior. This research is of immediate use to the United States Department of Agriculture's National Agricultural Statistical Service (NASS) for survey planning and collection in the June Agricultural Survey, and may provide a useful tool in further research in agricultural land use.