This study explores practical methods of estimating the signal-to-noise ratio of challenging seismic data containing a signal too small to be identified visually, as often occurring in a desert environment with single-sensor surveys. As a result, the estimation can only be performed using an ensemble of traces. We compare several data-driven estimation approaches and reveal their practical limits using a controlled experiment. We identify a stacking-based method as the most robust with the broadest range of applicability. We relate the number of traces in the ensemble to the lowest reliably estimated absolute signal-to-noise ratio possible with this method using such an ensemble. We support the findings using synthetic and single-sensor field data with low signal-to-noise ratios down to -60 dB. The proposed methodology allows reliable data-driven estimation of very low absolute signal-to-noise ratios directly from pre-stack seismic data.
