Overwater Emitter Localization with a Single Receiver using Neural Network Modeling.

We show that simple neural network models can predict the range of an overwater emitter in certain ducting environments using physics-based synthetic data and real-world collected data. Evaporative ducting provides a medium through which electromagnetic fields propagate further than in a free-space environment, and with significant multipath effects. The richness of the multipath signal and its range dependence allow for predictive models of emitter localization. We show the limitation of these neural networks in predicting range on real-world collected data in which ducting was apparent but its characteristics unknown.