To assess potential disturbance effects on fish from seismic air-gun surveys, we described several metrics to characterize the exposures from such surveys, including the number of emissions by area and time, and metrics based on accumulated sound exposure levels (SEL). For the SEL-based metrics we used both a simple spherical–geometrical model and a model that incorporated physical sound propagation properties such as bottom topography and the vertical difference in sound speed. We applied the metrics to two experiments in Norwegian waters (the Nordkappbanken and Vesterålen experiments) where fish distributions and fisheries were affected by the air-guns, but where the disturbance was stronger in the Nordkappbanken case. The metrics based on the number of emissions by area and time showed a stronger impact in the Nordkappbanken case. For the SEL-based metrics, the simple sound propagation model failed because of artificially elevated levels close to the emissions, but for the more complex propagation model, contrary to expectations, a stronger SEL was found in the Vesterålen case. We conclude that simple sound propagation models should be avoided and that the reliance on sound energy metrics like SEL for disturbance effects must be interpreted with caution.