Geological carbon storage (GCS) is a key strategy for reducing CO₂ emissions and addressing climate change. To ensure that injected CO₂ stays safely underground for hundreds to thousands of years, it's important to carefully evaluate potential storage sites before moving forward with a project. One major step in this process is defining the area of review (AoR), the region around the injection site that must be monitored and possibly managed to prevent leaks or environmental impacts. However, predicting how CO₂ and pressure will spread underground is very challenging due to limited data and the natural complexity of the subsurface. Traditional simulation tools used to estimate these outcomes are computationally expensive and cannot efficiently explore all the possible geological variations. In this study, we develop a fast and accurate deep learning-based surrogate model to replace traditional simulations. By training this model on a set of detailed simulations, we can rapidly predict how CO₂ and pressure behave under different conditions. This allows us to assess risks of storage failure and determine which geological factors most influence the outcome. The approach improves early site screening, supports smarter data collection, and reduces project cost and risk, helping make carbon storage safer and more effective.