Standard geophysical methods for monitoring CO2 injection have severe limitations in many cases. In particular, while the presence or absence of CO2 can be monitored during an injection, it is difficult to quantify changes in CO2 saturation in most reservoirs. We hypothesize that amplitude variation with offset (AVO) attributes may be able to provide more sensitive discriminators for CO2 presence. We propose a workflow useful for the prediction of the AVO response to a CO2 flood under arbitrary fluid saturation and pressure conditions. Using this workflow, we combine laboratory and numerical experiments to upscale and predict the effects of small-scale heterogeneities on AVO response at the seismic scale. This exercise demonstrates the ability of AVO to not only determine the presence or absence of CO2, but also to track changes in CO2 saturation over time.