A Hybrid Empirical–Neural Model for HFSWR False Alarm Reduction Caused by Meteo-Tsunami-Like Phenomena

The meteo-tsunami as an atmospheric phenomenon is still being researched, and its effects beyond physical ones (destruction if they hit the shore) are yet to be fully classified. One such effect is an increase in false alarm occurrence in high frequency surface wave radar (HFSWR) systems used for vessel traffic surveillance and control. Unfortunately, this effect is characterized by high-dimensional and highly nonlinear functional dependencies that cannot be described by closed-form mathematical equations. Since an artificial neural network is a highly parallelized distributed architecture with a fast flow of signals from input to output designed not to execute a predefined set of commands, but to “learn” dependencies during the training process and to apply that knowledge to solve unknown, but similar problems, it is a natural solution to the presented problem. Hybrid empirical–neural model-based probabilistic neural networks (PNN) used in this research proved to be quite capable of recognizing when an increase in false alarms can be expected based on the monitoring of atmospheric conditions in the HFSWR network coverage area and to eliminate them from the system, thus increasing the safety and security of all the actors in maritime traffic.