Regional and global chemical transport models underpredict NO_x (NO + NO₂) in the upper troposphere where it is a precursor to the greenhouse gas ozone. The NO_x bias has been shown in model evaluations using aircraft data (Singh et al., 2007) and total column NO₂ (molecules cm⁻²) from satellite observations (Napelenok et al., 2008). The causes of NO_x underpredictions have yet to be fully understood due to the interconnected nature of simulated emission, transport, and chemistry processes. Recent observation-based studies, in the upper troposphere, identify chemical rate coefficients as a potential source of error (Olson et al., 2006; Ren et al., 2008). Since typical chemistry evaluation techniques are not available for upper tropospheric conditions, this study develops an evaluation platform from in situ observations, stochastic convection, and deterministic chemistry. We derive a stochastic convection model and optimize it using two simulated datasets of time since convection, one based on meteorology, and the other on chemistry. The chemistry surrogate for time since convection is calculated using seven different chemical mechanisms, all of which predict shorter time since convection than our meteorological analysis. We evaluate chemical simulations by inter-comparison and by pairing results with observations based on NO_x:HNO₃, a photochemical aging indicator. Inter-comparison reveals individual chemical mechanism biases and recommended updates. Evaluation against observations shows that all chemical mechanisms overpredict NO_x removal relative to long-lived methanol and carbon monoxide. All chemical mechanisms underpredict observed NO_x by at least 30%, and further evaluation is necessary to refine simulation sensitivities to initial conditions and chemical rate uncertainties.