Due to its fast economic development, China’s emissions are in the spotlight of efforts to mitigate climate change and improve regional and city-scale air quality. Despite growing efforts to better quantify China’s emissions, the current estimates are often poor or inadequate. Bottom-up inventories are generally based on sectoral statistical information and therefore rely strongly on the accuracy of the input data. Complementary to bottom-up methodologies, inverse modeling of fluxes has the potential to improve those estimates through the use of atmospheric observations of trace gas compounds. Here we present comparisons of key pollutant emissions from different bottom-up inventories, and perform 20-year model simulations of the atmospheric composition over China using either the EDGARv4.2 or the MACCity bottom-up emission databases. The skill of the model to capture the observed variability and trends is assessed through comparisons with satellite NO2 observations retrieved from GOME, SCIAMACHY and OMI sensors through 1997–2008 and HCHO columns observed by OMI over 2005–2010. Next, we use a decade (2005–2014) of OMI HCHO columns to constrain the VOC emissions over China in a flux inversion framework built on the IMAGESv2 chemistry-transport model, and adjust the emissions of VOC precursors of HCHO in the model in order to reduce the discrepancy between the model predictions and the HCHO observations. The interannual and seasonal variability of the resulting top-down VOC fluxes (anthropogenic, pyrogenic and biogenic) is presented and confronted to past studies.