The CSIRO Division of Atmospheric Research Limited Area Model (DARLAM) is used to study the emission and transport of trace gases in the atmosphere. Two simulations are performed, one for radon, which has relatively simple surface emissions, and another for CO2, which has more complex spatial and temporal distributions of surface fluxes. For this study, DARLAM was initialized from observed atmospheric conditions. Diurnally varying surface fluxes from photosynthesis and respiration were included as functions of temperature, incoming solar radiation, and vegetation type. Industrial sources of CO2 and an Australian inventory of radon fluxes were also included. This study provides an evaluation of the model's trace gas transport capability by comparing Cape Grim measurements with the modeled concentrations of radon and CO2. Correlations calculated using 8-hour averaging are generally high for radon (0.6--0.8), and lower for CO2 (0.2--0.6). An objective verification technique of categorical forecasting is used to show that the model can satisfactorily forecast the occurrence of radon events. The quality of the model CO2 simulation is less satisfactory. Extra simulations suppressing either island or mainland sources are used to identify whether the air is of island, oceanic, or mainland origin. Further simulations suppressing industrial or biospheric sources are then used to show that the model can handle isolated industrial sources, but biospheric sources and sinks are not handled well at present. We conclude that insufficient accuracy in the spatial and temporal representation of the biospheric CO2 sources and sinks is the main reason for the model's less satisfactory CO2 simulation. ¿ 2000 American Geophysical Union