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

splitting flag near sunset (hour 17), artificially limiting the number of split puffs that are
generated by the model. However, for the ETEX evaluation puff‐splitting was enabled for each
simulation hour instead of the default option of once per day in order to allow for full
treatment of wind shear. The puff splitting feature of the CALPUFF modeling system does not
have a complementary puff “merging” feature which aggregates puffs according to specified
rules when they occupy the same space. Without the complementary puff merging capability,
the number of puffs generated by puff‐splitting can rapidly increase, resulting in extensive
computational requirements of the model and eventual simulation termination once the
maximum number of puffs allowed by the model is exceeded. Since the ETEX CALPUFF
application was of short duration, the number of puffs allowed was increased so no termination
occurred. However, the use of all hour puff splitting with CALPUFF in an annual simulation
could be problematic. The SCIPUFF Lagrangian puff model also performs puff splitting when a
sheared environment is encountered, however it can perform puff merging when two puffs
occupy the “same” space so does not suffer from the extensive computer time of CALPUFF
when aggressive puff splitting is desired.
The horizontal and vertical grid structures of CALPUFF were similar to the parent MM5 data.
Twenty‐seven (27) vertical levels were used in CALPUFF with each of the first 27 MM5 layers
matched explicitly to the CALPUFF vertical structure, through the lowest 4,900 m vertical depth
of the atmosphere. Additionally, 168 discrete receptors were included in the modeling analysis,
with the location of each corresponding to the location and elevation of the ETEX monitors.
AERMOD (EPA, 2004) turbulence coefficients, no complex terrain adjustment, and puff‐splitting
were selected for this analysis. A constant emission rate of 7.95 g/s was assigned for twelve
hours of release of the PMCH tracer. Plume rise and momentum were also simulated in
CALPUFF according to the release characteristics detailed on the ETEX website. CALPUFF
results were integrated for 90 hours, and model results were post‐processed in order to
generate 30 three (3) hour averages for each of the 168 discrete receptors.
For SCIPUFF simulations, the horizontal and vertical grid structures of the extracted MM5 data
were similar to the original MM5 data. Twenty‐eight (28) vertical levels were extracted,
encompassing a depth of approximately 5,000 m, similar to the CALPUFF simulations. Plume
rise and momentum were also simulated in SCIPUFF in the same manner as the CALPUFF
simulations. SCIPUFF results were also integrated for 90 hours, and model results were post‐
processed in order to generate 30 three (3) hour averages for each of the same 168 discrete
receptors.
FLEXPART simulations used a 375 x 175 horizontal grid at a resolution 0.16° (~18 km)
latitude/longitude. All MM5 vertical layers were extracted for the transport simulation. The
FLEXPART concentration grid consisted of 15 vertical levels from the surface to 1,500 m with 9
layers below the first 500 m. Emissions were released at 10 meters. Concentrations were bi‐
linearly interpolated to grid cells corresponding to the 168 ETEX monitoring locations that were
used.
HYSPLIT simulations used a 60 x 60 concentration grid with a horizontal resolution 0.25° (~28
km) latitude/longitude, consistent with NOAA’s model configuration for ETEX described on the
DATEM website. All MM5 vertical layers to 5000 meters were extracted for the transport
simulation. Emissions were released at 10 meters. The gridded concentration output was
linearly interpolated to the sampling locations utilizing software from NOAA’s Data Archive of
Tracer Experiments and Meteorology (DATEM) project. HYSPLIT was configured as a puff‐
106