Levoglucosan (LG) emitted from non-biomass burning (non-BB) sources has given rise to biased or even unreasonable source identification results when adopting LG as a distinct marker of biomass burning (BB). The estimation of LG emission and its spatiotemporal variation for various sources are the keys to reducing uncertainty. This study first developed a LG emission inventory for China from 25 sub-type sources belonging to eight categories, with a 3 km × 3 km spatial resolution and monthly distribution. The total LG emission in 2014 was 145.7 Gg. Domestic BB and open BB contributed 39.2 and 34.3% of the total emission. Non-BB sources, including municipal solid waste burning (9.7%), firework burning (9.6%), meat cooking (5.4%), domestic coal burning (1.5%), ritual item burning (0.2%), and industrial coal burning (0.1%), contributed to 26.5% of the total emission. LG emission varied spatially and temporally. Non-BB sources have a significant spatiotemporal impact on BB source contributions, even in high BB emission regions or in sowing, harvesting, and winter heating seasons. The local BB contributions have been substantially overestimated by 4.28-369% in previous studies, wherein LG was solely referred to as the BB source. By 2018, LG emission from BB might decrease to 63.9% of its total emission. This high-resolution LG emission inventory can be greatly useful for source identification studies in China. It also supports future research on the modeling of smoke aging and pollution control.