Monitoring of CO2 plumes is required to verify long-term sequestration of injected CO2, detect possible leaks from the storage zone, and infer deviations in plume path due to unforeseen large-scale geologic anomalies like flow barriers or high permeability streaks. Although several alternatives are available for monitoring plume path deviations during injection, the cost of monitoring still is a paramount concern for operators. The conjecture of this work is that dynamic data measured at injection wells are informative of the presence of heterogeneities large enough to affect plume paths. Hence geologic models updated using injection data can be used to predict better the trajectory of CO2 plumes and to design operational schemes that will ensure long term containment of the injected CO2. To test this conjecture, we have adapated a probabilistic history matching software (Pro-HMS) originally developed for oil field applications. The software assimilates injection data commonly monitored at active and inactive wells into models for the subsurface aquifer/reservoir. The algorithm yields an ensemble of realizations geologically consistent with the initial model such that the uncertainity in CO2 plume location can be easily assessed. We illustrate the approach with synthetic examples (a deep, heterogeneous aquifer with a long streak of large or small permeability) for which the reference bottomhole pressure and rate responses from injection and inactive wells were obtained from simulation. The updated models obtained after the history matching process detected the presence of the streak, and the subsequent estimation of CO2 plume location was much more accurate. Pressure data recorded at inactive wells were essential for detecting the streak. Monitoring such wells is thus expected to provide valuable information about geologic heterogeneities. The results confirm that injection rate and pressure data provide an inexpensive option for monitoring CO2 plumes.