NASA’s Orbiting Carbon Observatory 2 (OCO-2) was launched last July, and has been collecting data on carbon dioxide concentrations in Earth’s atmosphere ever since. OCO-2 observes spectral radiances at 3048 wavelengths on ground footprints that are approximately 1 square km. These radiances are used to infer CO2 concentration through a procedure called a retrieval. The OCO-2 retrieval is based on Bayes’ Theorem, and provides estimates of the posterior mean and variance of total column CO2 concentration, given the observed radiances, in each footprint. However, various conceptual and computational approximations are required in order to implement the procedure in an operational environment. Uncertainty quantification for OCO-2 seeks to quantify the statistical performance of the calculated posterior mean and (in particular) variance as estimators of the true posterior quantities. Our approach is based on Monte Carlo simulations of synthetic ensembles of state vectors under different geophysical conditions, and a stochastic treatment of uncertain retrieval algorithm inputs. This talk will give an overview the OCO-2  mission, our strategy for uncertainty quantification via Monte Carlo simulation, and some preliminary results.