FDTD analysis of 3D lightning problems with material uncertainties on GPU architecture

We present a computational framework for the prediction of electromagnetic fields in lightning problems which, unlike common methodologies, can efficiently incorporate the statistical variations that characterize the ground electric parameters. Our full-wave technique is based on the three-dimensional (3D) extension of the stochastic finite-difference time-domain (S-FDTD) method, and enables the extraction - in a single run - of fundamental statistical parameters, such as the mean value and standard deviation, that describe the generated fields. Aiming at efficient code implementations via paralellization, the increased computing power of modern graphics processing units (GPUs) is exploited. Numerical tests verify the validity of the proposed approach, as the corresponding results compare very well with those obtained via - more time-consuming - Monte-Carlo FDTD (MC-FDTD) simulations, for a wide range of parameter values.