Nitrous oxide is a potent greenhouse gas. Approximately 80% of current U.S. emissions of nitrous oxide come from soil and agriculture, in particular from the soil microbial processes of denitrification and nitrification. A thorough understanding of denitrification drivers is necessary to accurately predict nitrous oxide emissions from a particular agroecological landscape. However, current predictive models of denitrification do not consider spatio-temporal variation of drivers, nor the size and structure of the denitrifier microbial population. This study identifies the environmental and biological drivers of denitrification at different topographical positions (summit, backslope and footslope) and seasons (autumn, winter, spring and summer) within an agricultural field. In addition, the spatio-temporal variation of denitrifier and nitrifier microorganism abundance in soil was evaluated. Soil environmental measurements of soil water content, nitrate (NO3-N) concentration, ammonium (NH4-N) concentration, soluble total nitrogen, soluble non-purgeable organic carbon, pH, electrical conductivity, total carbon and nitrogen, mineral fraction carbon and nitrogen, and particulate organic matter carbon and nitrogen were measured. The abundance of nitrate reductase (nirK) and bacterial and archaeal ammonia monooxygenase (amoA) gene copies in soil were determined by quantitative PCR and community structure of the same populations was determined by T-RFLP. A short-term assessment of potential denitrification and basal denitrification in soils were performed using the acetylene inhibition method. A multivariate stepwise regression analysis of potential and basal v denitrification was performed using the soil environmental and biological measurements as possible explanatory variables. The predictive power of both potential and basal denitrification models was at times improved when spatio-temporal variation was considered. For potential denitrification, the overall model R-square was 0.69, while the model R-square for separate topographical positions and seasons ranged from 0.26 to 0.84. The predictive power of the potential denitrification model was improved further at times when nirK abundance was considered. In addition, it was found that the abundance of denitrifier and nitrifier microorganisms varies across the landscape over time. This study demonstrates the need to account for spatio-temporal variation when modeling and predicting denitrification rates in order to improve estimates of denitrification rates and thus more accurately prediction soil nitrous oxide emissions.