We show in this study that the non-linearity of NO_x chemistry implies that the local derivative (of a first-order Taylor expansion) is better suited in the inverse modelling of NO_x emissions over eastern China. We used single-grid-cell-based perturbation sensitivity calculation. The method leads to smaller NO_x emission over highly polluted regions and higher NO_x emission over rural regions. There are also more consistent results from using two different satellite instruments.