To construct a reliable and cost-effective monitoring system for injected in carbon capture and storage projects, we have considered a seismic monitoring approach using seismic noise from a fluid injection well. The passive seismic interferometry continuously monitors injected , enabling the detection of associated accidental incidents (e.g., leakage). We have applied three approaches: (1) crosscorrelation, (2) crosscoherence, and (3) deconvolution, to the passive seismometer data acquired during a fluid-injection experiment in Svalbard in the Norwegian Arctic. The crosscoherence approach enabled the construction of shot gathers similar to active-source data. Reflectors from the reservoir could be identified on common-midpoint (CMP) gathers constructed via seismic interferometry, and seismic velocity could be estimated from the time-lapse CMP gathers. High-frequency noise from fluid injection operations and low-amplitude background ambient noise were suitable for reconstructing virtual seismic data. However, we clearly found that the time variation characteristics of the noise influenced monitoring results, and thus the stable part of the noise should be used for monitoring. We further applied surface-wave analysis to the virtual shot gathers derived from seismic interferometry and investigated variations in S-wave velocity structure in a shallow formation. We observed clear time variations in seismic velocity in the shallow part of permafrost regions. The information derived from the surface-wave analysis is useful in evaluating the influence of shallow formations on monitoring results of deep reservoirs.