Prediction of ozone production, a photochemically non-linear process, is known to be biased in coarse-resolution chemical transport models (CTM). Other species subject to non-linear sources or sinks are also susceptible. Here we compute the resolution-dependent bias in predicted NO<sub>2</sub> column, a quantity relevant to the interpretation of space-based observations. We use 1-D and 2-D models to illustrate the mechanisms responsible for these biases over a range of NO<sub>2</sub> concentrations and model resolutions. We find that the behavior of calculated biases in NO<sub>2</sub> depends on the magnitude and spatial extent of the NO<sub>2</sub> source. We use WRF-CHEM to determine the resolution necessary to predict column NO<sub>2</sub> to 10 % and 25 % over the Four Corners power plants in NW New Mexico, Los Angeles, and the San Joaquin Valley in California.