To mitigate global warming, it is necessary to reduce the large volume of CO₂ released into the atmosphere. Carbon capture and storage technology is a promising means of achieving this purpose. Those who inject and store CO₂ in sub-seabed geological formations may be legally required to monitor CO₂ leakage, even when the likelihood of such an event is very low. In the case that the leaked CO₂ is monitored by limited numbers of CO₂ censors set on the sea bottom, it is required to elucidate the leakage location and rate immediately, in order to evaluate its impacts on the marine ecosystems and to minimize the impacts. In this study, we developed a numerical method to predict the location and rate of a contaminant seeping from the seafloor, using the adjoint marginal sensitivity method, based on the data monitored by limited numbers of sensors. A time-reversed numerical test was successfully conducted, using information obtained by the time-forward simulation of tidal current and contaminant dispersion.