The nitrous oxide (N2O) budget has been the least well constrained of the global trace gas budgets. For biogenic sources the uncertainty is caused by their extreme spatial and temporal heterogeneity. For the anthropogenic sources political, economic and cultural factors are major uncertainties associated with scaling-up. Zero-, one- and two-dimensional models are useful tools to explain historic time series of observations. Although no unique solutions are found for emissions, analysis of historic emission sources can provide an understanding of the total source strength and its increase in time, and the relative importance of sources. The impossibility to find unique solutions also occurs in forward and inverse simulations with three-dimensional models. Common problems of three-dimensional atmospheric models is that time-averaged emissions and wind fields are used for large grids, or the meteorology is taken from a specific year. This may cause discrepancies between model results and observations, which are often not representative of the temporal and spatial scale of the atmospheric model. Although the uncertainty of atmospheric concentrations is much smaller than that of emission estimates, there are other problems in forward and inverse three-dimensional modelling related to the atmospheric measurements. The observational network is scant, and at some stations the sampling frequency is low, often with temporal gaps in the observations. Finally, individual observations are of limited accuracy and precision, and observations from different networks are often not easily compared because of differences in measurement techniques and standards used. Expansion of the monitoring network with continental stations and isotopic characterization of emissions may help to better constrain the N2O budget using atmospheric models.