Documentation of the Evaluation of CALPUFF and Other Long ...

The fitted Gaussian plume statistics for the GP80 600 receptor arc and the initial CALPUFF
sensitivity tests are shown in Table 3‐15, with the percent differences (or angular offset for the
plume centerline location) between the model predictions and observations also shown
graphically in Figures 3‐8 through 3‐12. Unlike the CALPUFF performance for the 100 km arc
that mostly overestimated the fitted plume centerline (Cmax) and observed maximum
concentrations at any receptor (Omax), the CALPUFF sensitivity tests under‐estimate the
Cmax/Omax values for the 600 km arc by 40% to 80% (Table 3‐15 and Figures 3‐8 and 3‐9). The
Cmax/Omax underestimation bias is lower (‐40% to ‐60%) with the “C” series (i.e., no
meteorological observations in CALMET) of CALPUFF sensitivity tests. The CALPUFF sensitivity
tests overstate the amount of plume spread (σy) along the 600 km receptor arc compared to
the plume that is fitted to the observations (Figure 3‐10). The “A” and “B” series of CALPUFF
experiments using the turbulence dispersion (CAL and AER) tend to overestimate the plume
spread along the 600 km arc by ~50% with the “C” series overestimating plume spread by
~100%. For many of the experiments, use of the PG dispersion option greatly exacerbates the
plume spread overestimation bias with overestimation amounts above 250% for EPA1998_PG
and its related BASEA_PG scenarios. Given the similarity of the “C” series (CALMET with no
meteorological observations) and MMIF CALPUFF sensitivity simulations, it is not surprising that
the MMIF runs also overestimate plume spread by ~100%.
The predicted plume centerline angular offset from the observed value has an easterly bias of 9
to 19 degrees (Figure 3‐12). The “A” series of CALPUFF/CALMET sensitivity runs tend to have
larger (> 15 degrees) plume centerline offsets than the “B” and “C” series of experiments,
indicating that using upper‐air meteorological observations in CALMET tends to worsen the
plume centerline predictions in the CALPUFF sensitivity runs. Surprisingly, the CALPUFF/MMIF
sensitivity runs, which also do not use the upper‐air meteorological measurements, have
angular offsets in excess of 15 degrees.
The observed cross wind integrated concentration (CWIC) across the plume at the 600 km arc is
matched better by the CALPUFF sensitivity tests than the maximum (Cmax/Omax)
concentrations (Table 3‐15 and Figure 3‐12). The EPA1998_PG and EPA1998_CAL overestimate
the CWIC by 30% and 15%, respectively. However, the BASEA_PG and BASEA_CAL experiments,
which are designed to emulate the EPA 1998 CALPUFF runs, underestimate the CWIC by ‐14%
and ‐38%, respectively. The use of meteorological observations in CALMET appears to have the
biggest effect on the CALPUFF CWIC performance with the “A” series (use both surface and
upper‐air observations) have the largest CWIC underestimation bias and the CALPUFF CWIC
performance statistics as upper‐air (“B” series) and then surface and upper‐air (“C” series) are
removed from the CALPUFF modeling. The CALPUFF/MMIF runs underestimated the CWIC by
approximately ‐30%.
44