The radioactivity released from Japan's damaged Fukushima Daiichi nuclear power plant has increased the urgency to fund tracer experiments that will improve models of atmospheric dispersion and reinforce confidence in emergency procedures.

The last major tracer experiments were conducted in the mid-1990s. So the predictive capabilities of current atmospheric-dispersion models have not been properly tested, hindering their evaluation and development.

To generate more observational data, multiple-scale atmospheric tracer experiments should use non-hazardous, climate-neutral substances and a realistic release term with varying source strengths. Modellers could estimate emissions in real time using a limited set of observations without knowing the actual release rates, and later improve their models and data-reconstruction methods on the basis of the real source terms and measurements.