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

4.4 MODEL PERFORMANCE EVALUATION FOR THE SRL75 TRACER EXPERIMENT
The Irwin (1997) plume fitting evaluation approach was used to evaluate CALPUFF for the SRL75
field experiment. There are two components to the Irwin plume fitting evaluation approach:
1. A temporal analysis that examines the time the tracer arrives, leaves and resides on the
receptor arc; and
2. A plume fitting procedures that compares the predicted observed peak and average plume
concentrations and the width of the plume by fitting a Gaussian plume through the
predicted or observed concentrations across the arc of receptors or monitors that lie on
the 100 km receptor arc.
Because only long‐term integrated average observed SF6 samples were available, the timing
component of the evaluation could not be compared against observed values in the SRL75
experiments.
Most of the CALPUFF sensitivity tests estimated that the tracer arrived at the 100 km arc on
hour 13 LST, 2½ hours after the beginning to the tracer release. The exceptions to this are the
CALPUFF/MMIF simulations using the 4 km MM5 data and CALPUFF/MMIF using the 36 km and
PG dispersion that estimated the plume arrives at hour 14 LST. With one exception, the
CALPUFF simulations estimated that the tracer resided either 5 or 6 hours on the arc. And with
two exceptions, it was the meteorological data rather than the dispersion option that defined
the residence time of the estimated tracer on the 100 km receptor arc. The exceptions were
for the PG dispersion sensitivity test that in two cases predicted the tracer would remain one
less hour on the arc; the CALPUFF/CALMET BASE sensitivity test using the PG dispersion
estimated that the tracer would reside only 4 hours on the 100 km receptor arc. Without any
observed tracer timing statistics, these results are difficult to interpret.
Table 4‐9 displays the model performance evaluation for the various CALPUFF sensitivity tests
using the Irwin plume fitting evaluation approach. The observed values were taken from the
1998 EPA CALPUFF tracer test evaluation report data (EPA, 1998a). Also shown in Table 4‐4 are
the statistics from the 1998 EPA report for the CALPUFF V4.0 modeling using Pasquill‐Gifford
(PG) and similarity (CAL) dispersion. Note that the EPA 1998 CALPUFF modeling used CALMET
with just observations so is analogous to the BASE sensitivity scenario that used CALPUFF V5.8.
There are five statistical parameters evaluated using the Irwin plume fitting evaluation
approach:
• Cmax, which is the plume fitted centerline concentration.
• Omax, which is the maximum observed value at the ~40 monitoring sites or maximum
predicted value across the ~200 receptors along the 100 km arc.
• Sigma‐y, which the second moment of the Gaussian distribution and a measure of the
plume spread.
• Plume Centerline, which is the angle of the plume centerline from the source to the 100
km arc.
• CWIC, the cross wind integrated concentration (CWIC) across the predicted and observed
fitted Gaussian plume.
The first thing we note in Table 4‐9 is that the maximum centerline concentration of the fitted
Gaussian plume to the observed SF6 tracer concentrations across the 12 monitors (2.739 ppt) is
almost half the observed maximum at any of the monitors (5.07 ppt). As the centerline
concentrations in a Gaussian plume represents the maximum concentration, this means that
63