SUMMARYSince 2008, CO2 has been injected at the demonstration site for CO2 sequestration in Ketzin, Germany. Since 2009, a permanent array of seismic receivers installed by TNO at the injection site has recorded passive data continuously. It is the intention of TNO to use seismic interferometry (SI) by cross-correlation applied to the recorded ambient noise as a cost-effective technique for time-lapse monitoring. Under specific conditions, SI by cross-correlation can retrieve reflection data that may be interpreted in terms of subsurface layer properties. To assess the feasibility of this monitoring technique, we model base and monitor passive experiments for recording the response from noise sources and cross-correlate the recorded synthetic traces. The best results are obtained when the stationary-phase regions for the Green's functions are sampled densely enough by noise sources and, at the same time, the noise contributions from non-stationary regions cancel to a sufficient extent. Conversely, non-favourable noise conditions will cause wrong retrieval of the Green's functions and possibly deteriorate the monitoring potential of the technique. However, we show that being selective with the recorded noise in terms of illumination characteristics prior to correlation will improve the retrieved reflection data for further monitoring interpretation.