An observed distribution of aggregations of the euphausiid, Meganyctiphanes norvegica is used to model the effects of krill swarm orientation and shape on the success of two proposed survey designs. The simulation involves a number of random paths of parallel or radially arranged transects passing through three distributions each of which is given four rotations. The results indicate that the coefficient of variation (CV) of mean krill density varies inversely with mean density, and is lowest for the survey design utilising parallel transects set at right angles to the long axis of the aggregations. Calculations based on the power of surveys to reliably detect changes in mean density indicate that with probability of Type I and Type II errors 0.1 and 0.2 respectively, about 100 transects would be required to detect changes of 40% in the mean if CVs are as high as those obtained in the simulations.