Modeling electromagnetic propagation over water from correlated environmental data and neural network models

Accurate computations of electromagnetic (EM) propagation in the lower atmosphere require sophisticated modeling techniques such as those employed in the JHU/APL-developed Tropospheric Electromagnetic Parabolic Equation Routine (TEMPER). Since environmental conditions affect the propagation behavior, they are an integral part of these models. However, running TEMPER or other propagation simulations may not be practically feasible when a database of long-term conditions is desired at one or more geographical locations using large amounts of environmental data. In this case, statistical models of propagation such as neural network models may prove as valuable time-savers.