An accidental release of HTO to the atmosphere from a reactor at the Chalk River Laboratories was assessed in a timely and efficient manner using a combination of predictive modelling and environmental sampling. A simple Gaussian plume model performed well in predicting the concentration of HTO in air. Doses to workers and to members of the public were well below acceptable levels at all times during the incident. The release was turned to advantage to study tritium behaviour in the winter environment. HTO concentrations were measured in air, falling snow, vegetation and the snowpack at many locations during and after the release. The rate of HTO deposition to snow is greatly enhanced when snow is falling. The rate of new snow accumulation exceeded the rate of HTO diffusion in snow, so that the snowpack retained essentially all of the tritium deposited to it until spring melt occurred. Snow core data were therefore used as a surrogate for air concentrations to study the dispersion of the airborne plume, which was strongly affected by the topography of the Ottawa River Valley.