Fully Automated, Data Driven Suppression of Surface Scattering in Large Volumes of GPR Data

The occurrence of undesired surface scattering in GPR data is a well-known problem in GPR studies. Resulting false interpretation of such events is not uncommon. Two categories of approaches to suppress these types of air diffractions have been proposed by previous authors: a migration (semblance) based detection method combined with synthetic forward modelling plus subtraction of air diffractions, and a 2D (directional) filtering method. Both methods fell short for us in treating a large volume (32+ km of lines) of GPR field data from New Zealand. For the migration based approach, too much leftover surface scattering energy was observed in case of clustered and distorted events, and the 2D filtering approach was not an option given the tedious manual application and risk of removing real dips from the original data. Therefore, an improved, fully automated, data driven surface scattering suppression algorithm was developed. Because it uses the complex amplitude- and phase patterns of the original surface scattering for the forward modelled diffractions and subtraction thereof, this algorithm can better suppress the aforementioned clustered and distorted air diffractions. Application on the large volume of GPR field data yields exciting results, where desired subsurface features emerge from underneath the suppressed surface scattering.