Recently, inverse modelling has become an important tool in atmospheric science for obtaining “top down” estimates of emissions of trace gases such as, CO2, CO and CH4. In this approach, the mismatch between observations of the trace species and a forward model simulation of the observations is minimized to obtain an optimized estimate of the emissions that best fits the observations, given the uncertainty of the observations and the model. We are exploring ways of optimizing these to more reliably quantify surface fluxes of gases such as atmospheric CO and CO2.