United States
Environmental Protection
Agency
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S3-86/016 April 1986
&ER& Project Summary
Evaluation of the PEM-2 Using
the 1982 Philadelphia Aerosol
Field Study Data Base
Jia-Yeong Ku and K. Shankar Rao
This report describes an evaluation of
the Pollution Episodic Model Version 2
(PEM-2). The results of various statistical
measures and graphical comparisons of
model concentrations against measure-
ments obtained from 29 days of the
Philadelphia Aerosol Field Study are
presented. The PEM-2 is an urban-scale
plume model capable of predicting average
ground-level concentrations and deposi-
tion fluxes of one or two gaseous pollu-
tants or particulates at multiple receptors.
The two pollutants may be nonreactive or
chemically coupled through a first-order
chemical transformation. Hourly meteor-
ological data and emissions from 300
point and 289 area sources within the
Philadelphia metropolitan area were used
to calculate 12-h and 24-h average con-
centrations of fine and coarse total par-
ticulate mass, fine sulfate, and sulfur
dioxide.
Statistical tests for evaluation of model
performance include standard measures of
difference and correlation between mea-
surements and model estimates paired in
time and space. The particulate species
were simulated very closely; however, the
results must be interpreted with caution
due to the large contribution from the
regional background concentration which
far exceeded the local urban source con-
tributions. Differences between modeled
and observed concentrations are dis-
cussed, and the results are viewed in rela-
tion to a previous evaluation of the model.
This Project Summary was developed
by EPA's Atmospheric Sciences Research
Laboratory, Research Triangle Park, NC, to
announce key findings of the research pro-
ject that is fully documented in a separate
report of the same title (see Project Report
ordering information at back).
Introduction
The Pollution Episodic Model Version 2
(PEM-2) is an urban-scale model designed
to predict short-term ground-level concen-
trations and deposition fluxes of one or
two gaseous or particulate pollutants from
multiple point and area sources. The
effects of dry deposition, gravitational set-
tling, and a first-order chemical transfor-
mation are explicitly accounted for in the
model, if desired. PEM-2 computes the
horizontal and vertical dispersion param-
eters for both point and area sources from
Briggs' urban dispersion curves. Concen-
tration computations from area sources
are determined numerically for eight
downwind receptor grid squares. Optional
methods are also incorporated to consider
new plume rise and penetration equations
and to account for bouyancy-induced
dispersion.
This report describes an evaluation of
the PEM-2 model with data collected dur-
ing the Environmental Protection Agency's
Philadelphia Aerosol Field Study (PAFS).
This intensive field program was con-
ducted from 14 July to 14 August 1982
in Philadelphia, Pennsylvania. The primary
purpose was to obtain detailed data sets
on ambient surface concentrations of
gaseous and particulate pollutants, emis-
sions, and relevant meteorological
parameters with adequate time and spatial
resolution for the evaluation of short-term
source and receptor models for a major ur-
ban area.
The goal of the evaluation was to pro-
-------
vide an objective determination of the
accuracy of the model when tested in an
operational mode against a suitably exten-
sive urban data base. The model perform-
ance was determined by comparing
calculated average concentrations with
corresponding observed values for the
following pollutant species: fine par-
ticulate (FP) total mass, coarse particulate
(CP) total mass, fine sulfate, and sulfur
dioxide (S02). The FP size cut-off is 2.5
urn and the CP size range extends from 2.5
to 10 p(m.
An integral part of the PAFS program
was an air monitoring network consisting
of six sites strategically located within the
Philadelphia urban area. The concentra-
tions of the above pollutants were sampl-
ed on a 12h basis with average daytime
(6 AM — 6 PM) and average nighttime (6
PM — 6 PM) concentrations; except S02
was measured at hourly intervals.
Meteorological data input to the model
consisted of the hourly surface observa-
tions of wind speed and direction, and
temperature from the National Weather
Service Office at the Philadelphia Inter-
national Airport. Morning and evening
mixing heights were derived from high
resolution temperature soundings made
during the program. Hourly mixing heights
were obtained by interpolation, and stabili-
ty class was determined from the hourly
surface observations.
Hourly emission inventories of the four
modeled pollutants were specifically
developed with 300 major point sources
and 289 area sources. The area source
grid consisted of 17x17 square cells being
2.5 km on a side. Mobile source and minor
point source emissions were also prepared
on the same grid system and merged with
the area emissions in the evaluation.
Model Evaluation
The model was executed for the 29
days of the PAFS program. The model do-
main consisted of 32x32 cells with the
southwest corner of the grid located at
UTM coordinates 444.5 km north and
4380 km east in zone 18. The model cell
size of 2.5 km coincided with the area
source grid and the model domain encom-
passed the entire emissions inventory
area. The model code was specially
modified to allow for the input of hourly
emissions and its capacity was increased
to accommodate the 289 area sources.
Otherwise, no further changes were made
to the model's code.
The pollutant species were computed
from two model runs. In Run I, SO2 and
fine sulfate concentrations were com-
puted. The day and night transformation
rates of S02 were set at 7% and 3% per
hour, respectively. Daytime and nighttime
deposition velocities for SO2 were 3 and
2 cm/s, and 0.2 and 0.1 cm/s for fine
sulfate, respectively. Settling velocities for
both species were assumed to be zero. FP
and CP total mass concentrations were
modeled in Run II for each 12-h period. Day
and night deposition velocities were 0.2
and 0.1 cm/s for FP, and 0.5 and 0.5 cm/s
for CP, respectively. A settling velocity for
CP was set at 0.25 cm/s. These values for
the transformation rates, and deposition
and settling velocities were based on
typical values reported in the scientific
literature for large roughness features ap-
proximated by urban surfaces.
The calculated 12-h average concen-
trations resulting from the various local
emissions sources were added to their ap-
propriate background values. The latter
were determined as the lowest observed
12-h concentration at one of the outer four
PAFS sites located upwind of the
downtown area during the averaging
period. However, the background for S02
was assumed to be zero. It is acknowl-
edged that this determination of the
background is somewhat subjective and
limited due to the small number of upwind
sites. Additionally, the background values
were found to be a significant fraction of
the total predicted concentration.
The statistical measures of difference
and correlation between model and
observed concentrations identified by the
American Meteorological Society ap-
plicable to model evaluation and uncertain-
ty were determined. The evaluation report
contains complete tables of all statistical
results and various graphical displays. In
the following discussion, observed con-
centrations are denoted by 0 and pre-
dicted values are denoted by P. Averages
were determined from the total number of
pairs.
The results for FP total mass over the
entire data set show a mean ratio (P/O) of
1.24 and mean bias (0 — P) of -4.7 ng/m3
indicating the model slightly over predicted
observed values. A correlation coefficient
of 0.75 indicates a relatively high correla-
tion between predicted and observed FP
total mass. There was little difference be-
tween daytime and nighttime statistical
measures. The mean 0 and P concentra-
tions were 29.8 and 34.5 ng/m3,
respectively.
The model evaluation statistics for CP
total mass were a mean P/O of 1.29 and
a mean bias of -0.84; however, a correla-
tion coefficient of only 0.25 reveals a large
amount of randomness in the model-
observed pairs. The mean 0 and P values
were 14.6 and 15.4 j*g.m3 compared to a
mean absolute error of 5.7 yug/m3, which
indicates that PEM-2 determined CP total
mass to within a factor of two. The local
source contributions were more important
for CP total mass (i.e. background was
about twice as large on average) than for
FP total mass; however, the background
concentrations remained a decisive factor
in the overall CP mass levels in the urban
area.
The mean ratio of P/O for fine sulfate
was 1.45, and the mean bias of -1.31
m/m3 reveals the model slightly over-
predicts observations. The mean O and P
concentrations were 9.8 and 11.2 ng/m3,
respectively. There was a relatively high
correlation coefficient of 0.77 for both day
and night. The background value of fine
sulfate was also a significant part of the
concentration levels in the urban area.
The hourly S02 measurements provide
a demanding test for short-term models
like PEM-2. Graphs of the diurnal variation
of observed and predicted S02 concen-
trations revealed a greater overprediction
by the model at night than during the day
at the three sites closer to the central ur-
ban area where there are major area
source emissions and surrounding major
point sources. At the three sites on the
periphery of the urban area where area
emission are much smaller, the model
underestimated concentrations during the
day, but still overestimated SO2 at night.
For 12-h averaged values, the mean ratio
of P/O was 1.51; however, the correlation
coefficient was 0.1. The mean bias was
-5.71 f4g/m3 which also indicates that the
model overpredicted concentrations. The
model performed better during the day
than at night for the 12-h averaged results.
Due to the significant impact of the
regional background values to the overall
concentrations, a stepwise regression pro-
cedure was also performed to investigate
their impact on the model evaluation. This
regression method consisted of fitting a
linear relation between observed and
background values. The next step involved
the observed concentration as the inde-
pendent variable and the background and
modeled concentrations as dependent
variables. Residuals and correlation coef-
ficients were determined for both step-
wise relations for all pollutant species. The
results of the regression analyses revealed
an incremental improvement when in-
cluding the modeled concentrations in the
stepwise regression; however, the higher
correlations and smaller errors were not
statistically significant. The observed
ground-level concentrations of fine sulfate,
and FP and CP total mass were highly
-------
related to the background level. In
addition, estimates for background
concentrations based on the observed
concentrations upwind appears to be
adequate.
Conclusions
This report describes an evaluation of
the PEM-2 model with the PAFS data base.
The model performance was tested by
statistical comparisons of predicted con-
centrations against observed measure-
ments for four pollutant species. For
sulfate and fine paniculate total mass, the
values for the correlation coefficient and
index of agreement were high, and a large
part of the estimated error was unsys-
tematic. For coarse paniculate total mass,
the evaluation statistics were not as good
by comparison, and the larger systematic
error suggests room for improvement.
Overall, the model slightly overpredicted
for the particulate species.
The overall high degree of model per-
formance for the particulate species
should also be interpreted with caution
since the background concentrations, aris-
ing from the regional inflow, exceeded the
urban source contributions. Ideally, the
background values should be obtained
from monitoring stations at the grid boun-
daries, farther from the source region
being modeled. Nevertheless, the back-
ground particulate concentrations derived
from observed suburban sites and by wind
direction analysis for each 12-h averaging
period seem appropriate.
Jia- Yeong Ku andK. Shankar Rao are with the National Oceanic and Atmospheric
Administration, Oak Ridge, TN 37830.
James M. Godowitch is the EPA Project Officer (see below).
The complete report, entitled "Evaluation of the PEM-2 Using the 1982
Philadelphia A erosol Field Study Data Base," (Order No. PB 86-167 921 /A S;
Cost: $16.95, subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-86/016
OC00329 PS
U S EWVIR PROTECTION ASiNCY
REGION 5 LIBRARY
230 S DEARBORN STREET
CHICAGO IL 60604
------- |