Estimating picking errors in near-surface seismic data to enable their time-lapse interpretation of hydrosystems

Time-lapse applications of seismic methods have been recently suggested in the near-surface scale to track hydrological properties variations due to climate, water level changes, or permafrost thaw, for instance. But when it comes to travel-time tomography or surface-wave dispersion inversion, a careful estimation of the data variability associated with the picking process must be considered prior to any time-lapse interpretation. In this study, we propose to estimate picking errors that are due to the inherent subjectivity of human operators, using statistical analysis based on picking repeatability. Two seismic datasets were collected along the same profile under distinct hydrological conditions across a granite-micaschist contact at the Ploemeur hydrological observatory (France). Both datasets were recorded using identical equipment and acquisition parameters. A thorough statistical analysis is conducted to estimate picking uncertainties, at the 99% confidence level, for both P-wave first arrival time and surface-wave phase velocity. With the suggested workflow, we are able to identify 33% of the P-wave traveltimes and 16% of the surface-wave dispersion data, which can be considered significant enough for time-lapse interpretations. In this selected portion of the data, point-by-point differences highlight important variations linked to different hydrogeological properties of the subsurface. These variations show strong contrasts with a non-monotonous behaviour along the line, offering new insights to better constrain the dynamics of this hydrosystem.