Locations of gamma dose rate sensors have often been chosen by administrative or geometrical criteria. Nowadays computational capacity allows for a more realistic basis. We use simulations of potential radioactive plumes based on weather data of one year to investigate the threats to regions without own nuclear power plants and to find good numbers and locations of sensors to detect such plumes. We optimise sensor locations by minimising a cost function that can take into account numbers of undetected plumes, their dose to the region in general, or on the population. Besides we assess the effect of administrative constraints, be it that sensors have to cover administrative units, or that optimisation is done for sub-regions separately. Finally we evaluate the robustness of the approach if less or other plumes are used. The main findings are that sensors at boundaries are often best, but also typical paths of plumes may be important, and that administrative constraints may necessitate much more sensors. The small numbers of sensors actually deployed in these regions seem sufficient. However, the latter may be an artefact of the low number of plumes we considered. Altogether, combined with other considerations, this approach can contribute to better decisions about gamma dose rate sensor locations.