Mathematical modeling of ground-penetrating radar: parallel computing applications

In heterogeneous sub-surface environments, the evaluation of GPR sections is complicated by the influence of near-field effects, antenna radiation patterns, velocity variations and surveying inconsistencies. Section interpretation can be exceedingly difficult, even with advanced processing methods, and therefore mathematical modelling has become an increasingly popular addition to traditional techniques. The Finite-Difference Time-Domain method (FDTD) is the most common, but to be of practical use the modelling scheme must incorporate realistic antenna configurations, complex sub-surface geometries and accurate material property descriptions. These additional components add computational complexity to the models and, at present, most single processor FDTD schemes are only capable of modelling relatively basic three-dimensional data sets in practical time scales. Modern parallel computing techniques have the potential to overcome these limitations by spreading the computational demand across a number of processors (or individual PC's). These PC 'cluster' machines provide the necessary computational power required to model more complex GPR problems in realistic time-scales. Consequently, the scope and run-time of current GPR FDTD modelling applications can be improved making them an accessible and affordable aid to GPR interpretation.