Model simulations of tropospheric O3 require an accurate specification of the reactive odd nitrogen (NOx) source from lightning that is consistent in time and space with convective transport of O3 precursors. Lightning NOx production in global models is often parameterized in terms of convective cloud top heights (CLDHT). However, a closer relationship may exist between flash rate and other measures of convective intensity. In this study, flash rates are parameterized in terms of CLDHT, convective precipitation (PRECON), and upward convective mass flux (MFLUX) from the Goddard Earth Observing System Data Assimilation System (GEOS DAS). GEOS-based flash rates are compared to flash rates from the National Lightning Detection Network (NLDN) and Long Range Flash (LRF) network and the Optical Transient Detector (OTD). Overall, MFLUX-based flash rates are most realistic. PRECON- and CLDHT-based flash rates are too large in the tropics. The MFLUX- and PRECON-based flash rates are a factor of 3 too high (low) over the equatorial western Pacific (central and southern Africa), while CLDHT-based flash rates are much too low at nearly all marine locations. In many cases, biases in the flash rate distributions can be traced to biases in the GEOS DAS convective fields. Improvements in flash rate parameterizations will be tied closely to improvements in model physics as well as to increases in the amount of tropical data that are available for assimilation. Flash rates calculated from the 6-hour averaged CLDHTs are much less variable than observed, and O3 production rates calculated using the NOx produced from these flash rates are likely to be larger than observed.