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

3.4.3 SLUG and Puff Splitting Sensitivity Tests for the 600 km Arc
One issue of concern with the initial CALPUFF sensitivity tests was the large differences
between the estimated residence time of the tracer on the 600 km receptor arc in the EPA 1998
and current CALPUFF simulations using the CALPUFF (CAL) turbulence dispersion options when
the same meteorological observations are used as input into CALPUFF. The 1998EPA_CAL
CALPUFF sensitivity simulation estimated that the tracer would remain on the 600 km receptor
arc for 13 hours, which compares favorably with what was observed (12 hours) but is almost
double what the BASEA_CAL simulation estimated (7 hours). In addition to updates to the
CALMET and CALPUFF models that have occurred over the last decade, a major difference in
the 1998 EPA and current CALPUFF 600 km arc simulations was that the 1998 EPA CALPUFF
modeling used the near‐source slug option, whereas the current analysis did not. Another
major difference between the version of CALPUFF used in the 1998 EPA and current study was
that CALPUFF now has the ability to perform puff splitting. In fact, it was the presence of puff
splitting in CALPUFF that caused EPA to comment that CALPUFF may be applicable to distances
further downwind than 300 km in the 2003 air quality modeling guideline revision that led to
CALPUFF being the recommended long‐range transport model for chemically inert pollutants
(EPA, 2003).
To investigate this issue, a series of slug and puff splitting sensitivity tests were carried out
using the BASEA_CAL CALPUFF/CALMET configuration by incrementally adding the near‐source
slug option (MSLUG = 1) and puff splitting option (MSPLIT = 1) to the BASEA_CAL model
configuration. CALPUFF slug and puff splitting sensitivity tests were also carried out using the
MMIF12_CAL and MMIF12_PG model configurations. Two types of puff splitting sensitivity
tests were carried out:
• Default Puff Splitting (DPS) whereby the vertical puff splitting flag was turned on for just
hour 17 (i.e., IRESPLIT is equal to 1 for just hour 17 and is 0 the other hours); and
• All hours Puff Splitting (APS) that turned on the vertical puff splitting flag for all hours of
the day (i.e., IRESPLIT has 24 values of 1).
Table 3‐16 displays the tracer residence time statistic on the 600 km receptor arc for the slug
and puff splitting sensitivity tests. Using the puff model formulation and no puff splitting
(BASEA_CAL), CALPUFF estimates that the tracer resides on the 600 km arc for 7 hours, which is
‐42% less than observed (12 hours). Using all hours puff splitting in CALPUFF, but still using the
puff model formation (BASEA_APS_CAL), does not affect the estimated plume residence time
statistic (7 hours). However, when the slug option is used (BASEA_SLUG_CAL) the residence
time of the estimate tracer on the 600 km receptor arc more than doubles increasing from 7 to
15 hours. And adding puff splitting (APS) to the slug model formulation increases the estimated
tracer duration on the arc by another hour (16 hours).
The sensitivity of the CALPUFF/MMIF model configuration 600 km receptor arc tracer residence
time statistic to the specification of the slug and puff splitting options is a little different than
the CALPUFF/CALMET BASEA model configuration. Whereas the CALPUFF/CALMET BASEA
model configuration saw little sensitivity of the estimated tracer concentration residence time
on the arc due to puff splitting, the implementation of default puff splitting increases the tracer
residence time from 6 to 8 hours (CAL dispersion) and from 7 to 11 hours (PG dispersion) with
all hours puff splitting increasing the residence time even more. The effect of the slug option
using the CALPUFF/MMIF modeling platform has a very different effect on the tracer duration
time on the arc using the CAL and PG dispersion algorithms. Using the CAL dispersion option
51