Sensornets have been proposed consisting of thousands or tens of thousands of nodes. Economic and logistical considerations imply predeployment evaluation must take place through simulation rather than field trials. However, most current simulators are inadequate for networks with more than a few hundred nodes. In this paper we demonstrate some properties of sensornet application and protocols that only emerge when considered at scale, and cannot be effectively addressed by representative small-scale simulation. We propose a novel multi-phase approach to radio propagation modelling which substantially reduces computational overhead without significant loss in accuracy.