Abstract. A box model using measurements from the Mexico City Metropolitan Area study in the spring of 2003 (MCMA-2003) is presented to study oxidative capacity (our ability to predict OH radicals) and RO_x (RO_x=OH+HO₂+RO₂+RO) radical cycling in a polluted (i.e., very high NO_x=NO+NO₂) atmosphere. Model simulations were performed using the Master Chemical Mechanism (MCMv3.1) constrained with 10 min averaged measurements of major radical sources (i.e., HCHO, HONO, O₃, CHOCHO, etc.), radical sink precursors (i.e., NO, NO₂, SO₂, CO, and 102 volatile organic compounds (VOC)), meteorological parameters (temperature, pressure, water vapor concentration, dilution), and photolysis frequencies.

Modeled HO_x (=OH+HO₂) concentrations compare favorably with measured concentrations for most of the day; however, the model under-predicts the concentrations of radicals in the early morning. This "missing reactivity" is highest during peak photochemical activity, and is least visible in a direct comparison of HO_x radical concentrations. We conclude that the most likely scenario to reconcile model predictions with observations is the existence of a currently unidentified additional source for RO₂ radicals, in combination with an additional sink for HO₂ radicals that does not form OH. The true uncertainty due to "missing reactivity" is apparent in parameters like chain length. We present a first attempt to calculate chain length rigorously i.e., we define two parameters that account for atmospheric complexity, and are based on (1) radical initiation, n(OH), and (2) radical termination, ω. We find very high values of n(OH) in the early morning are incompatible with our current understanding of RO_x termination routes. We also observe missing reactivity in the rate of ozone production (P(O₃)). For example, the integral amount of ozone produced could be under-predicted by a factor of two. We argue that this uncertainty is partly accounted for in lumped chemical codes that are optimized to predict ozone concentrations; however, these codes do not reflect the true uncertainty in oxidative capacity that is relevant to other aspects of air quality management, such as the formation of secondary organic aerosol (SOA). Our analysis highlights that apart from uncertainties in emissions, and meteorology, there is an additional major uncertainty in chemical mechanisms that affects our ability to predict ozone and SOA formation with confidence.