In situ observations of NO2, NO, NOy, ClONO2, OH, O3, aerosol surface area, spectrally resolved solar radiation, pressure and temperature obtained from the ER-2 aircraft during the Photochemistry of Ozone Loss in the Arctic Region in Summer (POLARIS) experiments are used to examine the factors controlling the fast photochemistry connecting NO and NO2 and the slower chemistry connecting NOx and HNO3. Our analysis uses chemical coordinates to examine gradients of the difference between a model and precisely calibrated measurements to provide a quantitative assessment of the accuracy of current photochemical models. The NO/NO2 analysis suggests that reducing the activation energy for the NO+O3 reaction by 1.7 kJ/mol will improve model representation of the temperature dependence of the NO/NO2 ratio in the range 215--235 K. The NOx/HNO3 analysis shows that systematic errors in the relative rate coefficients used to describe NOx loss by the reaction OH+NO2→HNO3 and by the reaction set NO2+O3→NO3; NO2+NO3→N2O5; N2O5+H2O→2HNO3 are in error by +8.4% (+30/-45%) (OH+NO2 too fast) in models using the Jet Propulsion Laboratory 1997 recommendations . Models that use recommendations for OH+NO2 and OH+HNO3 based on reanalysis of recent and past laboratory measurements are in error by 1.2% (+30/-45%) (OH+NO2 too slow). The +30%/-45% error limit reflects systematic uncertainties, while the statistical uncertainty is 0.65%. This analysis also shows that the POLARIS observations only modestly constrain the relative rates of the major NOx production reactions HNO3+OH→H2O+NO3 and HNO3+h&ngr;→OH+NO2. Even under the assumption that all other aspects of the model are perfect, the POLARIS observations only constrain the rate coefficient for OH+HNO3 to a range of 65% around the currently recommended value. ¿ 2000 American Geophysical Union