Model assessment and verification - Air Pollution Modeling and Its Application - page 366

and can be thought of as the day of the year that a process reaches a maximum. An
estimate of A t is the modulus of the wavelet filtered time series. Phase (ph t ) and
amplitude (A t ) of seasonal ozone are typically low frequency processes.
3. Results and Discussion
3.1. Observed and modeled ozone by time scale
Low-frequency ozone variation (trend) was defined with a low-pass filter (Fig. 1).
The trend is captured by CMAQ (R = 0.79 and 0.89 at ABT147 and SHN418,
respectively).
Observed and CMAQ seasonal wavelets and their amplitude for ABT and SHN
are shown i n Fig. 2. Correlations between observed and CMAQ seasonal variation
(0.97 and 0.99 at ABT and SHN, respectively), for the most part measure phase
differences.
Among the meaningful measures of agreement between observations and
model outputs are the correlation between seasonal amplitudes (0.71 and 0.76 at
ABT147 and SHN418, respectively) and the phase difference ( Fig. 3 ). CMAQ is
out of phase with observations, sometimes by more than 10 days. The CMAQ
phase has trended downward since 1999 at both sites (seasonal maximum coming
earlier).
Ozone seasonal amplitude is modulated in part by meteorology (Fig. 4) . The
dotted lines in Fig. 4 are linear combinations of meteorological variables that
include temperature, solar radiation, relative humidity and wind speed. The most
significant covariates are wind speed and solar radiation at ABT and SHN,
respectively.