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

observed CWIC. For example, using the CAL turbulence dispersion option, CALPUFF/MMIF
underestimates the observed CWIC at the 600 km receptor arc by ‐32% using the puff model
configuration and no puff splitting. Using the DPS and APS puff splitting approach reduces the
CWIC underestimation bias to ‐28% and ‐21%, respectively, And then adding the slug
formulation with the APS completely eliminates the CWIC underestimation bias (‐2%). In fact,
use of the APS and slug options with the CALPUFF/MMIF modeling platform results in the best
performing CALPUFF sensitivity test for estimating CWIC across the 600 km arc of all the
CALPUFF sensitivity tests analyzed (Tables 3‐15 and 3‐17).
Table 3‐17. Plume fitting statistics for the CALPUFF slug and puff splitting sensitivity tests.
CALPUFF slug and puff
splitting sensitivity test
Cmax Omax Sigma‐y Centerline CWIC
(ppt) MNB (ppt) MNB (m) MNB (deg) Diff (ppt‐m) MNB
Observed 0.3152 0.3068 16,533 369.06 13,060
CALPUFF/CALMET
BASEA_CAL 0.0875 ‐72% 0.0817 ‐73% 36,870 123% 27.55 18.49 8,084 ‐38%
BASEA_APS_CAL 0.1014 ‐68% 0.1029 ‐66% 35,510 115% 27.19 18.13 9,023 ‐31%
BASEA_SLUG_CAL 0.0728 ‐77% 0.0726 ‐76% 62,650 279% 22.49 13.43 11,430 ‐12%
BASEA_SLUG_APS_CAL 0.0673 ‐79% 0.0652 ‐79% 58,440 253% 23.56 14.50 9,855 ‐25%
CALPUF/MMIF
MMIF12KM_CAL 0.1029 ‐67% 0.1012 ‐67% 34,290 107% 26.43 17.37 8,842 ‐32%
MMIF12KM_DPS_CAL 0.1049 ‐67% 0.1016 ‐67% 35,960 118% 16.74 7.68 9,454 ‐28%
MMIF12KM_APS_CAL 0.1108 ‐65% 0.1076 ‐65% 37,120 125% 16.30 7.24 10,310 ‐21%
MMIF12KM_SLUG_CAL 0.1458 ‐54% 0.1462 ‐52% 35,190 113% 16.92 7.86 12,860 ‐2%
MMIF12KM_PG 0.0956 ‐70% 0.0887 ‐71% 39,120 137% 24.89 15.83 9,371 ‐28%
MMIF12KM_DPS_PG 0.1085 ‐66% 0.1143 ‐63% 41,610 152% 17.04 7.98 11,310 ‐13%
MMIF12KM_APS_PG 0.1085 ‐66% 0.1143 ‐63% 41,610 152% 17.04 7.98 11,310 ‐13%
MMIF12KM_SLUG_PG 0.1251 ‐60% 0.1115 ‐64% 41,770 153% 17.43 8.37 13,100 0%
3.5 CONCLUSIONS ON GP80 TRACER TEST EVALUATION
For the 100 km receptor arc CALPUFF/CALMET sensitivity simulations, the ability of CALPUFF to
simulate the observed tracer concentrations varied among the different CALMET configurations
and were not inconsistent with the results of the 1998 EPA CALPUFF evaluation study (EPA,
1998a). The best performing CALPUFF/CALMET configuration was when CALMET was run using
MM5 data and just surface meteorological observations and no upper‐air meteorological
observations. In general, the CAL and AER turbulence dispersion options in CALPUFF performed
similarly and performed better than the PG dispersion option. The performance of CALPUFF
using the MMIF tool tended to be in the middle of the range of model performance for the
CALPUFF/CALMET sensitivity tests; not as good as the performance of CALPUFF/CALMET using
MM5 and just surface observations data in CALMET, but better than the performance of
CALPUFF using MM5 data and no meteorological observations in CALMET.
The CALPUFF sensitivity modeling results for the GP80 600 km receptor arc were quite variable.
With two notable exception (the BASEA_PG and EXP2C configurations), the initial CALPUFF
sensitivity tests were unable to duplicate the observed tracer residence time on the 600 km
receptor arc as was seen in the 1998 EPA CALPUFF evaluation study (EPA, 1998a). However,
when the near‐source slug option was used, CALPUFF/CALMET was better able to reproduce
53