The simulation of deep convective mixing in the Goddard Chemical Transport Model (GCTM) is evaluated by comparing 1990--1992 distributions of upper tropospheric convective mass flux and cloud top pressure from the Goddard Earth Observing System data assimilation system (GEOS-1 DAS) with deep convective cloud fields from the International Satellite Cloud Climatology Project (ISCCP). Deep convective mixing in the GCTM is calculated using convective information from the GEOS-1 DAS. Therefore errors introduced when deep convection is parameterized in the GEOS-1 DAS affect the distribution of trace gases in the GCTM. The location of deep convective mixing in the tropics is fairly well simulated, although its north-south extent is overestimated by >5¿. The frequency of deep convective mixing also appears to be overestimated in the tropics, resulting in GCTM-calculated upper tropospheric concentrations of carbon monoxide in the tropics that are larger and less variable than those observed. The spatial extent of deep convective mixing in the subtropics is overestimated at several locations including the Caribbean throughout the year and the South Pacific Convergence Zone during June--August. The extent of deep convection is underestimated over midlatitude marine storm tracks. DAS-calculated cloud top pressures differ from ISCCP cloud top pressures by less than one-half a GCTM layer (35 hPa) at most longitudes in the tropics; however, cloud top pressures are overestimated by more than 35 hPa (i.e., the vertical extent of deep convection is underestimated) over wintertime midlatitude storm tracks and the Indian Ocean and underestimated by more than 35 hPa at locations that include the Gulf of Mexico during December--February and central South America during June--August. ¿ 1997 American Geophysical Union