The annoyance potential of odours can be assessed using dispersion modelling, thereby delineating separation distances between pollution sources and nearby communities. It is common practice in this context to assume constant emissions over time, although odour emissions are often characterised by temporal variability. Here we show that the assumption of constant emissions can bias the separation distances towards underestimation, as compared to more realistic scenarios incorporating time-varying emissions. We identify three primary factors driving the level of such underestimation: wind direction frequency, degree of emission variability and percentile compliance level of odour impact criteria. Accordingly, the underestimation was more significant in the prevailing wind directions. With greater variability of the odour emission rate, the separation distances tended to be larger. The higher the percentile, the greater the underestimation of separation distances. In particular, the 90th percentile showed superior skill in counteracting the source emission variability when compared to the 98th, 99th and 99.5th percentiles. The findings are achieved using a Lagrangian particle dispersion model. Meteorological input data are due to locally-derived wind and turbulence measurements at a site in Central Europe (Austria). Discrete representation of odour emissions over time (hourly resolution) was accounted for by employing a Monte Carlo-based method (inverse transform sampling). This work provides a better understanding of the extent to which accounting for temporal variability in odour emissions can be most useful.