In a well-controlled experimental setup, a very high precision for studying soil water dynamics can be achieved with Ground-Penetrating Radar (GPR). However, accessing detailed hydraulic information for a field scale 2D subsurface structure requires the usage of multiple antennas. As each antenna is a unique instrument, the evaluation of signals recorded with different antennas under otherwise identical conditions show characteristic differences. In this paper, we quantify these variations and discuss the arising uncertainties. While measurements with one single antenna are highly reproducible, signal shapes and travel times vary significantly as measured by different antennas. This currently limits the attainable accuracy.