Sensitivity and uncertainty analysis of atmospheric ozone photochemistry models. Final report, September 30, 1993--December 31, 1998

The author has examined the kinetic reliability of ozone model predictions by computing direct first-order sensitivities of model species concentrations to input parameters: S{sub ij} = [dC{sub i}/C{sub i}]/[dk{sub j}/k{sub j}], where C{sub i} is the abundance of species i (e.g., ozone) and k{sub j} is the rate constant of step j (reaction, photolysis, or transport), for localized boxes from the LLNL 2-D diurnally averaged atmospheric model. An ozone sensitivity survey of boxes at altitudes of 10--55 km, 2--62N latitude, for spring, equinox, and winter is presented. Ozone sensitivities are used to evaluate the response of model predictions of ozone to input rate coefficient changes, to propagate laboratory rate uncertainties through the model, and to select processes and regions suited to more precise measurements. By including the local chemical feedbacks, the sensitivities quantify the important roles of oxygen and ozone photolysis, transport from the tropics, and the relation of key catalytic steps and cycles in regulating stratospheric ozone as a function of altitude, latitude, and season. A sensitivity-uncertainty analysis uses the sensitivity coefficients to propagate laboratory error bars in input photochemical parameters and estimate the net model uncertainties of predicted ozone in isolated boxes; it was applied to potential problems in the upper stratospheric ozone budget, and also highlights superior regions for model validation