Evaluating Meteorological input parameters for seasonal photochemical modeling on different time scales

Recently, regional-scale photochemical models have been applied for seasonal simulations. Among the results from one of these studies (a seasonal simulation with the RAMS3b/UAM-V system for the summer of 1995) were the findings that high ozone concentrations are attributable to positive forcings from fluctuations having time scales from 1 day to several weeks, and that model performance with respect to ozone observations is poor for the intra-day time scale and best for longer-term fluctuations (Rao et al., 1998, Hogrefe et al., 1999).

In this study, we apply the concept of scale analysis to several of the input fields for the seasonal photochemical simulation. In particular, the RAMS3b-predicted temperature, wind speed and moisture fields are decomposed into four spectral components, (namely intra-day, diurnal, synoptic and baseline fluctuations), and compared to observations. The results reveal that the predicted fields exhibit less temporal and spatial variability than the observations. This is demonstrated through calculations of the relative contribution of the individual components to the overall variance, and through the calculation of spatial correlation structures. For example, while intra-day fluctuations account for about 23% of the variance in the observed wind speed time series, they account for only about 3% of the variance of the predicted time series. In addition, it is shown that correlations between observed and predicted meteorological fields are better on the longer time scales. Out of the investigated variables, temperature predictions have the highest correlation with observations, while wind speed predictions have the lowest correlations. These results suggest that the previously shown inability of the modeling system to predict ozone fluctuations on the intra-day time scale (Hogrefe et al., 1999) is at least partially attributable to the fact that also the meteorological input fields show poor correlations with observations on this time scale. On the other hand, the better model performance with respect to ozone on the longer time scales appears to be related to the ability of the meteorological model to capture phenomena having time scales equal or greater than one day. The implications of these findings for the regulatory use of photochemical modeling systems are discussed.