This paper describes a statistical analysis of a large number of NO x and CO emission data from natural gas-fired industrial combustion plants in The Netherlands. Many of these emission data are based on stack measurements. The implied emission factors observed for these combustion processes show a high variability, both between different installations and within one installation over time. This variability is well described by lognormal probability distribution functions with a width of 1-2 orders of magnitude.Due to the large number of plants included in the national inventory of The Netherlands, time trends in the national NO x emissions of a few percent per year can be detected despite this large variability. A significant trend for CO could not be observed. This suggests that the variability is not disturbing monitoring of emission trends at the national level.At the individual plant level, this variability however introduces a high uncertainty in a priori emission estimates. This could severely hamper emission reporting in the framework of a plant level emission trading scheme.