Radon is increasingly being used as a tool for quantifying stability influences on urban pollutant concentrations. Bulk radon gradients are ideal for this purpose, since the vertical differencing substantially removes contributions from processes on timescales greater than diurnal and (assuming a constant radon source) gradients are directly related to the intensity of nocturnal mixing. More commonly, however, radon measurements are available only at a single height. In this study we argue that single-height radon observations should not be used quantitatively as an indicator of atmospheric stability without prior conditioning of the time series to remove contributions from larger-scale "non-local" processes. We outline a simple technique to obtain an approximation of the diurnal radon gradient signal from a single-height measurement time series, and use it to derive a four category classification scheme for atmospheric stability on a "whole night" basis. A selection of climatological and pollution observations in the Sydney region are then subdivided according to the radon-based scheme on an annual and seasonal basis. We compare the radon-based scheme against a commonly used Pasquill–Gifford (P–G) type stability classification and reveal that the most stable category in the P–G scheme is less selective of the strongly stable nights than the radon-based scheme; this lead to significant underestimation of pollutant concentrations on the most stable nights by the P–G scheme. Lastly, we applied the radon-based classification scheme to mixing height estimates calculated from the diurnal radon accumulation time series, which provided insight to the range of nocturnal mixing depths expected at the site for each of the stability classes.