Abstract: Data assimilation is to use the observed data to update the pollution sources from top to bottom, which could provide a fast way to update and verify the emission inventories, and also add further information about the temporal variation and some sources that could not easily be quantified by traditional bottom-up methods. The basic principles of several assimilation inversion methods, including the mass balance ,ensemble Calman filter and four-dimensional variation methods were reviewed, and their application methods in emission inventory inversion and optimization introduced. The acquisition method of observation data and the primary emission inventory, the calculation method of uncertainty, and the acquisition method of the relationship between pollution sources and observation data were also introduced. The application of pollution source assimilation inversion methods at different scales of pollution source inventory inversion, both at home and abroad, were summarized. The ensemble Kalman filter or the 4D variation method combined with adjoint model or decoupling direct method was used to determine the relationship between pollution sources and observation data, able to avoid the nonlinearities in the model to a certain extent. The more accurate observations and more precise air quality model were needed in the future to improve the effect of pollution sources inversion.