This paper examines the accuracy with which trace gas fluxes, from a source that disturbs the local wind and microclimate, may be estimated from measured concentrations, above or downwind from the source. The familiar flux–gradient methods, even if carefully applied within the near-surface constant-flux-layer, nevertheless posit horizontally-uniform wind and stability. Errors result if the windflow is actually advective (i.e. disturbed), so that its state is evolving in the alongwind direction. We take as an illustration the case of a gas evaporating uniformly (Q, kg m −2 s −1) from a small lagoon. We modify the Rao–Wyngaard–Cote local advection model, verify it against existing observations of disturbed flows, then calculate the fields of windspeed, temperature and tracer concentration over land and lake. From these “data” we calculate several estimates of the (known) source strength Q. Results by integration of the horizontal flux ( Q IHF) prove the most satisfactory, followed by those using a source–receptor relationship based on a backward Lagrangian stochastic method ( Q bLS). Flux–gradient estimates Q FG can be very seriously in error, and should only be used with caution in disturbed flow. These findings have generality beyond the specific case of a lagoon flow.