This seminar will explain a methodology for combining site characterization and soil CO2 monitoring for detecting leaks at geologic CO2 sequestration sites. Near surface CO2 fluxes resulting from a leak are simulated using the TOUGH2 model for different values of soil permeability, leakage rate and vadose zone thickness. Natural background soil CO2 flux rates are characterized by a Bayesian hierarchical model that predicts the background flux as a function of soil temperature. A presumptive leak is assumed if the monitored flux rate exceeds a critical value corresponding to a very high (e.g., 99%) prediction interval for the natural flux conditioned on temperature. A probabilistic calculation then combines the probability distribution of random leak locations relative to monitoring coordinates, the predicted size of the flux relative to the leakage rate, and the probability that the total flux (natural + leakage) exceeds the critical value for detection. The methodology is demonstrated for an idealized site employing soil surface seepage monitoring. Extensions are discussed, including: the combination of inferences from multiple leak detection technologies (e.g., tracer gas, soil gas isotope, and groundwater chemistry monitoring); and the application to actual sites with nonhomogeneous leak probabilities and variable CO2 leak-flux response profiles.