Chemistry of HOx radicals in the upper troposphere

Aircraft observations from three recent missions (STRAT, SUCCESS, SONEX) are synthesized into a theoretical analysis of the factors controlling the concentrations of HOx radicals (HOx=OH+peroxy) and the larger reservoir family HOy (HOy=HOx+2H2O2+2CH3OOH+HNO2+HNO4) in the upper troposphere. Photochemical model calculations capture 66% of the variance of observed HOx concentrations. Two master variables are found to determine the variance of the 24 h average HOx concentrations: the primary HOx production rate, P(HOx), and the concentration of nitrogen oxide radicals (NOx=NO+NO2). We use these two variables as a coordinate system to diagnose the photochemistry of the upper troposphere and map the different chemical regimes. Primary HOx production is dominated by the O(1D)+H2O reaction when [H2O]>100 ppmv, and by photolysis of acetone (and possibly other convected HOx precursors) under drier conditions. For the principally northern midlatitude conditions sampled by the aircraft missions, the HOx yield from acetone photolysis ranges from 2 to 3. Methane oxidation amplifies the primary HOx source by a factor of 1.1–1.9. Chemical cycling within the HOx family has a chain length of 2.5–7, while cycling between the HOx family and its HOy reservoirs has a chain length of 1.6–2.2. The number of ozone molecules produced per HOy molecule consumed ranges from 4 to 12, such that ozone production rates vary between 0.3 and 5 ppbv d−1 in the upper troposphere. Three chemical regimes (NOx-limited, transition, NOx-saturated) are identified to describe the dependence of HOx concentrations and ozone production rates on the two master variables P(HOx) and [NOx]. Simplified analytical expressions are derived to express these dependences as power laws for each regime. By applying an eigenlifetime analysis to the HOx–NOx–O3 chemical system, we find that the decay of a perturbation to HOy in the upper troposphere (as from deep convection) is represented by four dominant modes with the longest time scale being factors of 2–3 times longer than the steady-state lifetime of HOy.