Time-lapse monitoring is a powerful tool for tracking subsurface changes resulting from fluid migration. Conventional time-lapse monitoring can be done by observing differences between two seismic surveys over the surveillance period. Along with the changes in the subsurface, differences in the two seismic surveys are also caused by variations in the near-surface overburden and acquisition discrepancies. The virtual-source method monitors below the time-varying near-surface by redatuming the data down to the subsurface receiver locations. It crosscorrelates the signal that results from surface shooting recorded by subsurface receivers placed below the near-surface. For the Mars field data, redatuming the recorded response down to the permanently placed ocean-bottom cable (OBC) receivers using the virtual-source method allows one to reconstruct a survey as if virtualsources were buried at the OBC receiver locations and the medium above them were a homogeneous half-space. Separating the recorded wavefields into upgoing and downgoing (up-down) waves before crosscorrelation makes the resultant virtual-source data independent of the time-varying near-surface (seawater). For time-lapse monitoring, varying source signature for the two surveys and for each shot is also undesirable. Deconvolving the prestack crosscorrelated data (correlation gather) by the power spectrum of the source-time function results in virtual-source data independent of the source signature. Incorporating up-down wavefield separation and deconvolution of the correlation gather by the source power spectrum into the virtual-source method suppresses the causes of nonrepeatability in the seawater along with acquisition and source signature discrepancies. This processing combination strengthens the virtual-source method for time-lapse monitoring.