An empirical model of nitrous oxide emission from agricultural soils has been developed. It is based on the relationship between N2O and three soil parameters – soil mineral N (ammonium plus nitrate) content in the topsoil, soil water-filled pore space and soil temperature – determined in a study on a fertilized grassland in 1992 and 1993. The model gave a satisfactory prediction of seasonal fluxes in other seasons when fluxes were much higher, and also from other grassland sites and from cereal and oilseed rape crops, over a wide flux range (< 1 to > 20 kg N2O-N ha−1 y−1). However, the model underestimated emissions from potato and broccoli crops; possible reasons for this are discussed. This modelling approach, based as it is on well-established and widely used soil measurements, has the potential to provide flux estimates from a much wider range of agricultural sites than would be possible by direct measurement of N2O emissions.