Biomass burning is one of the largest sources of atmospheric trace gases and aerosols globally. These emissions have a major impact on the radiative balance of the atmosphere and on air quality, and are thus of significant scientific and societal interest. Several datasets have been developed that quantify those emissions on a global grid and offered to the atmospheric modelling community. However, no study has yet attempted to systematically quantify the dependence of the inferred pyrogenic emissions on underlying assumptions and input data. Such a sensitivity study is needed for understanding how well we can currently model those emissions and what the factors are that contribute to uncertainties in those emissions estimates. <br><br> Here, we combine various satellite-derived burned area products, a terrestrial ecosystem model to simulate fuel loads and the effect of fire on ecosystem dynamics, a model of fuel combustion, and various emission models that relate combusted biomass to the emission of various trace gases and aerosols. We vary one key parameter of a simple fuel combustion model, the emission model, and the burned area product and assess its impact on the computed emissions fields and their uncertainties. We find that choice of burned area data set has by far the largest impact on interannual variability of simulated emissions. For total global emissions, burned area and combustion completeness have the largest impact on emissions for most species. <br><br> We conclude that reliable information on burned area is key for accurately modelling spatial and interannual variations of wildfire emissions, but uncertainties about the combustion process have a similar impact on the magnitude of global emission estimates. The results are important for chemical transport modelling studies, and for simulations of biomass burning impacts on the atmosphere under future climate change scenarios.