The objective of the study was to develop a science-based model, OFFSET (Odor from Feedlot – Setback Estimation Tool), to establish setback distances from animal production sites based on the use of an air dispersion model (INPUFF-2) and the actual odor emission data from these sites. Extensive research was conducted to obtain representative odor emissions from various animal facilities and to evaluate the air dispersion model. Odor emissions were measured from 280 animal buildings and manure storage units on 85 farms in Minnesota during 1998 to 2001. The geometric means of the odor emission rates for each type of odor source were obtained to represent odor emissions of that source. The efficiencies of some odor control technologies were summarized. The air dispersion model was evaluated for short-distance (< 0.5 km) odor dispersion prediction against the odor plumes measured by trained field assessors on 20 farms and also for long-distance (4.8 km) odor dispersion prediction against odor data recorded by trained resident observers living in the vicinity of livestock operations in a 4.8 X 4.8 km rural area. The relationship between odor detection threshold and intensity was obtained for swine and cattle odors in order to convert odor intensity to detection threshold. The results indicated that the INPUFF-2 model was capable of simulating odor dispersion downwind from animal production operations for low-intensity odors. Six stable or neutral weather conditions that favor odor transport were identified, and their historical occurrence frequencies in all 16 directions at six weather stations in Minnesota were obtained. The occurrence frequencies of these weather conditions were used to determine odor occurrence frequencies in the OFFSET model.