We parameterize radon emissions in the TOMCAT global off-line three-dimensional chemical transport model (CTM) and compare modeled radon profiles with spatially and temporally matched observations obtained near Moffett Field, California, in June 1994. The CTM was forced using European Centre for Medium-Range Weather Forecasts analyses for April--August 1994. To identify the origin of modeled radon, we divided the radon sources into three regions. We performed CTM sensitivity experiments at horizontal resolutions of 2.8¿¿2.8¿ (latitude¿longitude) and 7.5¿¿7.5¿, and with and without moist convection and/or vertical diffusion. At the higher resolution the full CTM (i.e., including convection and vertical diffusion) generally agrees well with the observations in the free troposphere. The observations exhibit free-tropospheric radon peaks at altitudes where zonal wind speeds in the Pacific jet stream are greatest (generally at ≥7 km). In this region the modeled radon originates from Asia, and the observed variability in the radon concentration is reproduced. Thus, the model reproduces the position and strength of the jet. We identify observed radon peaks that may originate from convective lifting, as they do in the model results. If these observed peaks do originate from convection, then the full CTM captures the correct temporal variability in convection over Asia and the Pacific for the observation period. Quantitative comparison shows that the model-data agreement is degraded if we decrease the model resolution to 7.5¿¿7.5¿, or remove the parameterizations of convection and/or vertical diffusion. This suggests a resolution of near 2.8¿¿2.8¿ is needed in global models for realistic simulations of short-lived species. We have identified certain shortcomings with TOMCAT: the model underestimates the amount of convective cloud, the convective cloud top height, and the amount of vertical diffusion. We identify improvements to ameliorate these problems. ¿ 1998 American Geophysical Union