A major challenge for air quality forecasters is to reduce the uncertainty of air pollution emission inventory. Error in the emission data is a primary source of error in air quality forecasts, much like the effect of error in the initial condition on the accuracy of weather forecasting. Data assimilation has been widely used to improve weather forecasting by correcting the initial conditions with weather observations. Similarly, observed concentrations of air pollutants can be used to correct the errors in the emission data. In this study, a new method is developed for estimating air pollution emissions based on a Newtonian relaxation and nudging technique. Case studies for the period of August 1-25, 2006 in 47 cities in China indicate that the nudging technique resulted in improved estimations of SO2 and NO2 emissions in majority of these cities. Predictions of SO2 and NO2 concentrations in January, April, August, and October using the emission estimations derived from the nudging technique showed remarkable improvements over those based on the original emission data.