Modeling the long-term frequency distribution of regional ozone concentrations using synthetic meteorology

A new method is developed to generate the meteorological input fields requi red for use with photochemical airshed models that seek to predict the effe ct of pollutant emissions on the long-term frequency distribution of peak O -3 concentrations. Instead of using meteorological fields derived from inte rpolation of direct weather observations, this method uses synthetically ge nerated meteorological data. These synthetic meteorological fields are crea ted by first constructing a semi-Markov process that generates a time serie s of large-scale synoptic weather conditions that statistically resemble th e occurrence and persistence of synoptic weather patterns during specific m onths of the year. Then for each day within each synoptic weather category, local weather variables indicative of the meteorological potential for ozo ne formation are drawn from the approximated joint distribution of the summ ation of three pressure gradients across the airshed and the 850 mb tempera ture measured in the early morning. The synthetic initial conditions are co mbined with boundary values that are extracted from historical days that ma tch the chosen synoptic class, temperature, and pressure gradient values as closely as possible for use in a prognostic mesoscale meteorological model . The prognostic mesoscale meteorological model generates the meteorologica l input fields necessary for the photochemical airshed model. The airshed m odel driven by synthetically generated meteorological data is executed for a 31 day period that statistically resembles weather during the month of Au gust in Southern California using pollutant emissions data from the year 19 87. The procedure produced a frequency of occurrence of peak 8 h average oz one concentrations that compared well both to that produced by the determin istic model as well as to the O-3 concentrations observed over the August m onths of the years 1984-1990.