Abstract. We use a global chemical transport model (GEOSChem) to evaluate the consistency of satellite measurements of lightning flashes and ozone precursors with in situ measurements of tropical tropospheric ozone. The measurements are tropospheric O 3 , NO 2 , and HCHO columns from the GOME satellite instrument, lightning flashes from the OTD and LIS satellite instruments, profiles of O 3 , CO, and relative humidity from the MOZAIC aircraft program, and profiles of O 3 from the SHADOZ ozonesonde network. We interpret these multiple data sources with our model to better understand what controls tropical tropospheric ozone. Tropical tropospheric ozone is mainly affected by lightning NO x and convection in the upper troposphere and by surface emissions in the lower troposphere. Scaling the spatial distribution of lightning in the model to the observed flashes improves the simulation of O 3 in the upper troposphere by 5-20 ppbv versus in situ observations and by 1-4 Dobson Units versus GOME retrievals of tropospheric O 3 columns. A lightning source strength of 6±2 Tg N/yr best represents in situ observations from aircraft and ozonesonde. Tropospheric NO 2 and HCHO columns from GOME are applied to provide topdown constraints on emission inventories of NO x (biomass burning and soils) and VOCs (biomass burning). The topdown biomass burning inventory is larger than the bottom-up inventory by a factor of 2 for HCHO and alkenes, and by a factor of 2.6 for NO x over northern equatorial Africa. These emissions increase lower tropospheric O 3 by 5-20 ppbv, improving the simulation versus aircraft observations, and by 4 Dobson Units versus GOME observations of tropospheric O 3 columns. Emission factors in the a posteriori inventory are more consistent with a recent compilation from in situ measurements. The ozone simulation using two different dy-