You are on page 1of 13

Environ Monit Assess (2016) 188: 516

DOI 10.1007/s10661-016-5508-8

Comparison of AERMOD and CALPUFF models


for simulating SO2 concentrations in a gas refinery
Farideh Atabi & Farzaneh Jafarigol & Faramarz Moattar &
Jafar Nouri

Received: 27 January 2016 / Accepted: 26 July 2016 / Published online: 13 August 2016
# Springer International Publishing Switzerland 2016

Abstract In this study, concentration of SO2 from a gas complex conditions of topography, CALPUFF offers
refinery located in complex terrain was calculated by the better agreement with the observed concentrations.
steady-state, AERMOD model, and nonsteady-state
CALPUFF model. First, in four seasons, SO2 concen- Keywords SO2 concentration . AERMOD .
trations emitted from 16 refinery stacks, in nine recep- CALPUFF . Dispersion models . Gas refinery
tors, were obtained by field measurements, and then the
performance of both models was evaluated. Then, the
simulated results for SO2 ambient concentrations made
Introduction
by each model were compared with the results of the
observed concentrations, and model results were com-
Entering pollutants and also the rate of releasing pollutants
pared among themselves. The evaluation of the two
into the atmosphere have adverse effect on air quality.
models to simulate SO2 concentrations was based on
Pollutants may be released periodically or continuously
the statistical analysis and Q-Q plots. Review of statis-
from a point or non-point source or sources. The dispersion
tical parameters and Q-Q plots has shown that, accord-
of pollutants also depends on the way they are entering the
ing to the evaluation of estimations made, performance
atmosphere. Lack of information on the distribution of
of both models to simulate the concentration of SO2 in
pollutants makes the determination of their emissions dif-
the region can be considered acceptable. The results
ficult outside the site and the surrounding residential areas.
showed the AERMOD composite ratio between simu-
Because of the focus on oil and gas industry, Assaluyeh is
lated values made by models and the observed values in
the spotlight of many professionals of Environmental Sci-
various receptors for all four average times is 0.72,
ences and air pollution is considered as one of the chal-
whereas CALPUFF’s ratio is 0.89. However, in the
lenges facing this industrial area. Hence, better and more
effective control and reduction of air pollutants have been
the agenda of the relevant organizations. Generally, sulfur
oxides are the maximum amount of gases discharged into
F. Atabi : F. Jafarigol (*) : F. Moattar the atmosphere in addition to NOx around gas refineries
Department of Environmental Engineering, Graduate School of
Environment and Energy, Science and Research Branch, Islamic (Yannopoulos 2007). Sulfur dioxide is one of the important
Azad University, Tehran, Iran atmospheric pollutants, which is a major cause of urban air
e-mail: f.jafarigol@srbiau.ac.ir quality problem (Sahin et al. 2005; Karbassi et al. 2007). In
fact, many studies have emphasized that the critical con-
J. Nouri
Department of Environmental Management, Graduate School of centration of SO2 emitted from industrial facilities in the
Environment and Energy, Science and Research Branch, Islamic area countries can seriously affect air quality in the region
Azad University, Tehran, Iran (Abdul-Wahab 2006; Abdul-Wahab et al. 2011a). In this
516 Page 2 of 13 Environ Monit Assess (2016) 188: 516

study, SO2 emissions from a gas refinery in South Pars Gas during the year 2011 and then compared the field mea-
Complex located in Assaluyeh have been simulated by surement results with simulation results made by
two common atmospheric dispersion models AERMOD AERMOD and CALPUFF models. Based on the results
and CALPUFF. of each model verification, the performance of CALPUFF
AERMOD is composed of three preprocessors: mete- has been found superior than AERMOD in that case
orological preprocessor (AERMET), geological prepro- (Dresser and Huizer 2011). However, it should be empha-
cessor (AERMAP), and main processor (AERMOD) sized that the performance evaluation and comparison for
(Willis and Deardorff 1981). Preprocessor AERMET pro- both AERMOD and CALPUFF together have never been
cesses the hourly upper and surface meteorological data done for a gas refinery so far, especially for complex
and estimates the boundary layer parameters of the atmo- terrain conditions. Therefore, this study examines the
sphere (Gulia et al. 2015). The second preprocessor, performance of each model AERMOD and CALPUFF
AERMAP, analyzes topographic data in conjunction with in such circumstances for the first time. The focus of this
a layout of receptors and emission sources (Rood 2014). study is characterizing spatial and temporal patterns of
Finally, the model has done its calculation and provided its SO2 emissions, the transfer and fate of this gas in two
final results using the results of both the processors and the media, because it is located between Persian Gulf and the
complementary information about sources of emissions Zagros Mountains. Performance evaluation of the models
and receptor network (Cimorelli et al. 2004). has been conducted in two ways in the study: first, by
CALPUFF has been approved by the US Environmen- comparing the observed data and model output, modeling
tal Protection Agency as a nonsteady-state Lagrangian- accuracy at any point in the desired time period has been
Gaussian puff model for complex flow modeling including achieved in the designated recipient points for models
cases that are in the complex terrain and also in places (based on the coordinates UTM). In the second stage,
where the stationary and reverse flow are important the performance accuracy of the models compared with
(Holnicki and Nahorski 2015). Modeling system consists each other has been examined.
of three main parts CALMET, CALPUFF, and
CALPOST. CALMET is a meteorological model that
takes mid-scale meteorological data from prognostic me- Materials and methods
teorological models and calculates values for the finer
scales using data of terrain and land use in more precise Study area
networking (Abdul-Wahab et al. 2011b). CALPUFF is a
nonsteady-state model of fluid motion which has subrou- The South Pars gas field located on the border of Qatar
tines for networking of time changes, the meteorological some 100 km from the Iranian Gulf coast has been
conditions in the three dimensions, and effects of terrain, rapidly developing. This world’s largest gas field will
wetland and dry land subsidence, diffusion and dispersion be developed in 30 phases, of which this refinery en-
on the marine environment, and building downwash. compasses phases 6, 7, and 8. The South Pars forth
CALPOST is used for post-processing the output results refinery includes a condensate line, which transports
and processing simulation files (Snyder et al. 1985). the condensate to a single point mooring (SPM) for
Generally, few evaluations have been conducted to exportation. In this study, the fourth gas refinery located
compare the performance of CALPUFF and AERMOD in South Pars Gas Complex in Assaluyeh was selected
models. Dmitry Tartakovsky and colleagues have as a case study, due to the presence of various industries
reviewed the amounts of particulate emissions from a in the region, high density of air pollution in this area,
mine located in the hilly terrain of Israel by the atmospher- community concern about poor air quality in this area,
ic dispersion models, AERMOD and CALPUFF in 2013. special topographical, land use and climatological con-
For a wide range of weather conditions and topography in ditions of the study area and being close to Nayband
the study area, the values simulated by AERMOD com- gulf and National Park which are sensitive ecosystems
pared with field measurement results were in a better in the area (Fig. 1).
agreement than the results obtained by CALPUFF Generally, the data required for this study can be
(Tartakovsky et al. 2013). Alan Dresser and colleagues divided into three main categories: the data required
have collected SO2 emission values from the stacks of for modeling of atmospheric conditions, the data re-
both coal-fired power plants at eight receptor stations quired for modeling the dispersion and diffusion of
Environ Monit Assess (2016) 188: 516 Page 3 of 13 516

Fig. 1 Study area map

SO2 including inputs of AERMOD and CALPUFF, and (Vijay Bhaskar et al. 2008). AERMAP information
the ambient measured data of SO2 concentrations for the source has been digital elevation model (DEM) which
verification of the model’s simulations. A summary of has been created by Iran national cartographic center
the input data for the AERMOD and CALPUFF model- and has been used as input data of the preprocessor and
ing systems is provided in Table 1. After collecting data, converted to XYZ formats that can be read by
the ambient concentrations of SO2 in the region have AERMAP. The output of this preprocessor contains
been modeled and simulated by AERMOD and the coordinates of each receptor and its height above
CALPUFF and the results of the two models together sea level, roughness height at the receptor, and the
with the results of field measurements have been com- geographical coordinates of the source. The preproces-
pared using statistical parameters. sor is just implemented once for a desired region
(U.S.EPA 2004).
Meteorological, topographical and emission sources Since, DEM file cannot be used as input in
inputs CALPUFF the available global data has been used as
topographic file SRTM 3 (Shuttle Radar Topography
In this study, upper atmosphere data and synoptic sta- Mission) with an accuracy of 90 m and GTOPO with
tions have been obtained from Iran Meteorological Or- resolution of 900 m to provide data on the natural effects
ganization to verify the initial studies of climate and to CALPUFF. The GLCC file has been used for land use
provide input data of meteorological model CALMET data with a spatial resolution of 1 km. SRTM is an
and AERMET (IRIMO 2014). Assaluyeh Airport international project led by the National Geospatial-
(24 km from the reference point) and Jam-Tohid Intelligence Agency (NGA) and the National Aeronau-
(37 km from the reference point) stations are the closest tics and Space Administration (NASA). SRTM consists
stations to the study area and the emission sources. The of a special radar system which had been mounted on
meteorological parameters include surface (Assaluyeh space shuttle Endeavor processor during its 11-day mis-
Airport station) and upper air data (Jam-Tohid station). sion in February 2000. So, SRTM is one of the most
As seen in Fig. 2, seasonal wind roses in Assaluyeh complete digital topographic database of Earth with
airport meteorological station are similar. Specifically, high resolution. Terrain map in the study area has been
Fig. 2 demonstrates that the prevailing wind direction is displayed using the software CALPUFF in Fig. 3.
from the northwest to southeast in the station. Seasonal changes, especially changes in atmospheric
The implementation of air pollution model for each parameters such as temperature, the speed and direction
point requires topographic data and land use to deter- of wind, relative humidity, and boundary layer height in
mine the effects of dispersion and the surface roughness the area in different seasons will have a significant effect
516 Page 4 of 13 Environ Monit Assess (2016) 188: 516

Table 1 Summary of input data to the AERMOD and CALPUFF modeling systems

Type of input data Observation Source

AERMOD Summary of input data Wind direction, wind speed, cloud cover, pressure Hourly averages from synoptic stations of Assaluyeh
used by AERMET at sea level, standard pressure as superficial Airport (IRIMO 2014).
characteristics, and rainfall, temperature, dew
point temperature, and moisture as profile
characteristics
Surface characteristics: surface roughness, Bowen In the fourth section of user’s guide for the AERMOD
ratio, and albedo meteorological preprocessor (AERMET), values of
these parameters are presented based on user type and
vegetation and Theory of Pain (1987). The values of
these three parameters are introduced monthly, quarterly
or yearly.
Summary of input data Gridded terrain data The data from Mapping Organization of Iran has been used
used by AERMAP as the input data of the preprocessor in this study.
Other requirements of Location of the receptors, emission sources, Characteristics of stacks: stack dimensions, output stack
AERMOD model specifications of weather files temperature, emission flow, and velocity, etc.
Location of the monitoring stations has been identified as
discrete receptors in the model.
CALPUFF Summary of input data Surface meteorological data (cloud cover, ceiling Hourly averages from synoptic stations of Assaluyeh
used by CALMET height, relative humidity, station pressure, Airport (IRIMO 2014).
temperature, wind direction, and wind speed)
Upper air meteorological data (wind speed, wind Hourly averages observed vertical profiles from Jam-Tohid
direction, temperature, pressure, and elevation) station (IRIMO 2014).
Geophysical data (gridded fields) Terrain elevation, land use categories, albedo, Bowen ratio,
surface roughness length, soil heat flux, and vegetative
leaf area index.
Summary of input data Meteorological and geophysical data Produced binary file by CALMET
used by CALPUFF Location of the receptors, emission sources Characteristics of stacks: stack dimensions, output stack
characteristics temperature, emission flow and velocity, etc.
Location of the monitoring stations.

on the dispersion of pollutants (plume). Therefore, the Simulation has been done at a height of 1.5 m above
modeling has been done in four seasons to investigate ground level (breathing height) for all receptors.
the effects of climatological changes. After reviewing the production process and deter-
Receptors in two separate and network system have mining the sources of emissions during a field visit in
been presented to the model to cover all the sources in the study area, sampling of SO2 from active stacks have
the domain of study area. Then, one of the sources was been done by Testo 350-S/XL in four seasons (Table 3)
considered as reference point in the area with special in 2014 and three times per season (ASTM 2011b).
topographical and meteorological features. Network re- Since the diffusion coefficients indicate production and
ceptors have been specified in Cartesian coordinates emissions in normal operating conditions of a process,
within an area of 50 × 50 km2, and the locations of the the sampling has been done in normal operating condi-
monitoring stations have been introduced to the model tions of the system and refused in the abnormal condi-
as distinct receptors. Arrangements of all receptors in tions such as repairs or out of service equipment that
respect to the reference stack in the gas refinery have affect the performance of the process (Method 1997).
been selected so that it cover all sources and indicate the Measuring ambient SO2 concentrations have been done
atmospheric conditions in intermediate and micro-scales simultaneously with sampling from stacks for evalua-
as well as the effects of land use and topography in the tion of simulation results. For this purpose, nine moni-
study area. Location of the monitoring stations toring stations have been set in the area around the
(receptors) in respect to the reference stack has been refinery. The measurements have been carried out in
presented in Cartesian coordinate system in Table 2. four seasons of the year 2014 using LSI-Lastem Babuc
Environ Monit Assess (2016) 188: 516 Page 5 of 13 516

Spring 2014 Summer 2014

Fall 2014 Winter 2014


Fig. 2 Wind roses based on seasonal meteorological data from Assaluye airport in 2014

Fig. 3 Map of the three-


dimensional terrain within the
50 × 50 km (preprocessor output
CALPUFF with a precision of
1 km)
516 Page 6 of 13 Environ Monit Assess (2016) 188: 516

Table 2 Location of air quality monitoring stations (Jafarigol measurements for the desired receptors, using the statis-
et al. 2015)
tical parameters proposed by the U.S. Environmental
Station name Coordinates (UTM) Protection Agency (U.S.EPA 2003). The following pa-
rameters were applied in this study:
X (m) Y (m)
& The correlation coefficient (CCOF)
A 654,966.4 3,046,687
B 655,666.4 3,047,269
According to Equation (1), parameter CCOF shows
C 656,317.2 3,047,253
the relationship between the model results and field
D 656,192.9 3,046,577
measurements and as much as it is closer to 1, it shows
E 655,961.1 3,046,268
the desired accuracy of the model results.
F 655,548.3 3,045,966
G 655,545.9 3,046,695 N   
H 655,374 3,046,566 ∑ X i −X Y i −Y
i¼1
I 654,857.3 3,046,491 CCOF ¼   2  2 1 =2 ð1Þ
∑Ni¼1 X i −X ∑Ni¼1 Y i −Y

A. Measurements were done based on the average hour & Fractional bias (FB)
concentrations (ASTM 2011a).
Based on the framework provided by Olesen (2001),
Performance evaluation of the models statistical performance of the results has been expressed
by fractional bias parameters (fractional bias), which
In the present study, the simulated SO2 ambient concen- represents the tendency of the model to simulate more
trations (simulated by AERMOD and CALPUFF) have or less than the measured values in this study. The
been compared with the results of the field amount of fractional bias −0.67 has been equal to over

Table 3 Stacks characteristics and SO2 seasonal average emissions from gas refinery

Source name X-coord. (m) Y-coord. (m) Stack height (m) Stack diameter (m) Stack temp. (K) Average SO2 emissions (g/s)

Spring Summer Fall Winter

SP6-2-1104 235.13 43.92 16.8 1.30 527.20 0.08 0.1 0.06 0.09
SP6-2-2104 158.29 −19.53 16.8 1.30 523.25 0.1 0.3 0.23 0.14
SP7-2-3104 71.39 −90.81 16.8 1.30 509.10 0.11 0.18 0.2 0.19
SP7-2-4104 −5.45 −153.97 16.8 1.30 502.45 0.09 0.1 0.13 0.14
SP8-2-5104 −92.07 −225.55 16.8 1.30 494.95 0.05 0.08 0.12 0.09
SP8-2-6104 −168.10 −288.70 16.8 1.30 495.75 0.09 0.08 0.11 0.03
SP7-2-3107 0 0 12.745 1.75 658.65 0.28 0.3 0.19 0.25
SP7-2-4107 −77.38 −64.08 12.745 1.75 688.15 0.29 0.23 0.3 0.35
SP8-2-5107 −160.99 −135.32 12.745 1.75 678.88 0.4 0.29 0.34 0.37
SP8-2-6107 −245.01 −195.79 12.745 1.75 665.55 0.29 0.28 0.29 0.41
121-U-103 A −133.59 −521.57 42 2.80 426.65 0.16 0.1 0.2 0.15
121-U-103 D −214.24 −588.16 42 2.80 421.65 0.18 0.12 0.19 0.22
121-U-103 E −241.12 −610.05 42 2.80 405.48 0.15 0.1 0.16 0.13
120-16-GTG1 296.57 −756.38 20 3.63 760.75 1.33 1.52 1.86 1.29
120-17-GTG2 273.35 −775.14 20 3.63 749.15 1.68 1.23 1.54 1.72
120-17-GTG3 250.33 −793.94 20 3.63 768.15 2.21 1.98 1.85 2.6
Environ Monit Assess (2016) 188: 516 Page 7 of 13 516

Table 4 Comparison of the simulated values with seasonal average of observed values for SO2 ambient concentrations in spring 2014

Station name Observed (μg/m3) AERMOD CALPUFF

Simulated (μg/m3) Ratio to Observed Simulated (μg/m3) Ratio to observed

A 0.59 0.38 0.64 0.47 0.79


B 0.33 0.17 0.52 0.19 0.59
C 0.34 0.60 1.77 0.64 1.88
D 1.04 0.72 0.69 0.98 0.94
E 0.77 0.60 0.77 0.74 0.96
F 0.59 0.34 0.58 0.47 0.81
G 0.63 0.39 0.62 0.47 0.74
H 0.64 0.45 0.71 0.58 0.90
I 0.54 0.31 0.57 0.41 0.76

estimation by a factor of 2 and 0.67 equal to under AERMOD and CALPUFF results were also judged
estimation by a factor of 2. The deficit bias 0 indicates with quantile-quantile (Q-Q) plots (Fig. 5).One-hour
that the estimation of model is complete. A good model observed and simulated concentrations were ranked re-
performance should have a fractional bias of less than gardless of time and space. The middle line shows the
0.3 (Ghannam and El-Fadel 2013). slope, which at the slope of 1:1 observed and simulated
concentration are the same, at the slope of 1.5:1 simu-
lated concentration is 1.5 times of observed concentra-
Y −X tions, and at the slope of 1:1.5 observed concentration is
FB ¼ i i  ð2Þ
0:5 Y i þ X i 1.5 times of simulated concentrations, which in fact
shows the amounts of under estimation and over esti-
mation respectively (Myerson et al. 2003).
In which the parameters are as follows: Xi is the
simulated values, Yi is the measured values (observed),
X is the average of simulated values, Y is the average Results and discussion
of measured values, and N is the total measured number.
In this study, at first, the characteristics of each model
& Q-Q plots (CALPUFF and AERMOD) have been evaluated. Since

Table 5 Comparison of the simulated values with seasonal average of observed values for SO2 ambient concentrations in summer 2014

Station name Observed (μg/m3) AERMOD CALPUFF

Simulated (μg/m3) Ratio to observed Simulated (μg/m3) Ratio to observed

A 1.29 0.62 0.48 0.98 0.76


B 0.54 0.41 0.76 0.51 0.95
C 0.56 0.43 0.76 0.46 0.82
D 1.26 1.81 1.44 1.75 1.39
E 0.67 0.49 0.74 0.51 0.77
F 0.57 0.27 0.47 0.51 0.90
G 0.64 0.43 0.68 0.46 0.72
H 0.61 0.56 0.91 0.57 0.94
I 1.01 0.57 0.56 0.91 0.90
516 Page 8 of 13 Environ Monit Assess (2016) 188: 516

Table 6 Comparison of the simulated values with seasonal average of observed values for SO2 ambient concentrations in fall 2014

Station name Observed (μg/m3) AERMOD CALPUFF

Simulated (μg/m3) Ratio to observed Simulated (μg/m3) Ratio to observed

A 0.83 0.46 0.55 0.69 0.84


B 1.28 0.68 0.53 0.98 0.77
C 0.45 0.31 0.68 0.43 0.94
D 0.96 0.68 0.71 0.87 0.91
E 1.20 1.39 1.16 1.42 1.18
F 0.62 0.36 0.57 0.55 0.89
G 1.00 0.61 0.62 0.87 0.87
H 0.65 0.39 0.60 0.62 0.95
I 1.06 0.83 0.79 0.87 0.82

CALPUFF and AERMOD have been proposed for large calmer winds in the study area, which prevent helpful
(more than 50 km) and small ranges, respectively, they dispersion of air pollutants, occur more in fall and
have been selected for better comparison of an area with winter. AERMOD and CALPUFF simulated ratios ex-
dimensions of 50 × 50 km and then simulated results of cept one in each season represent an under estimation.
these models have been compared with each other. As The AERMOD average composite ratio for all four
indicated in Tables 4, 5, 6, and 7, the performance seasons for 1 h is 0.72, whereas CALPUFF’s ratio is
evaluation of the models is directly estimated as simply 0.89. The results show that CALPUFF has a less ten-
the ratio of simulated to observed concentration. Con- dency to under estimate actual 1-h concentrations than
ceptually, model precision is measured by the scatter of AERMOD.
the ratio of simulated to observed concentration in four Figure 4 present the simulated SO2 concentration
seasons. The averaging of observed concentrations of values for each receptor of study domain (50 × 50 km)
SO2 within the study area in spring period of 2014 for seasonal averaging period. In Assaluyeh, the prevail-
was 0.61 μg=m3 , in summer was 0.79 μg=m3 , in fall ing wind is almost all the time from the northwest to the
was 0.90 μg=m3 and in winter was 0.92 μg=m3 . The southeast as shown in Fig 2 confirming the highest
peak concentrations of SO2 occur in fall and winter ground level SO2 concentration to be the south east
because sulfur oxides, produced mainly in heating facil- from predominate source. This represents the depen-
ities, are mostly emitted in the fall and winter, and dency of pollutant dispersion patterns on the

Table 7 Comparison of the simulated values with seasonal average of observed values for SO2 ambient concentrations in winter 2014

Station name Observed (μg/m3) AERMOD CALPUFF

Simulated (μg/m3) Ratio to observed Simulated (μg/m3) Ratio to observed

A 0.68 0.25 0.37 0.55 0.81


B 1.16 0.43 0.37 0.92 0.79
C 1.36 0.90 0.66 1.08 0.80
D 1.45 1.25 0.86 1.31 0.90
E 0.40 0.61 1.53 0.60 1.51
F 0.84 0.47 0.56 0.64 0.77
G 0.77 0.56 0.73 0.58 0.76
H 1.26 0.62 0.49 0.57 0.45
I 0.37 0.17 0.46 0.26 0.70
Environ Monit Assess (2016) 188: 516 Page 9 of 13 516

Spring 2014 Summer 2014

Fall 2014 Winter 2014


Fig. 4 Seasonally simulation results of SO2 distribution in 2014

meteorological conditions. The meteorology and com- and 0.13, respectively. The AERMOD average FB
plex conditions of topography of the study area played a shows a bias toward under estimation, whereas
main role for SO2 emissions in the neighborhood of the CALPUFF average FB is relatively unbiased. For
refinery sources. all four seasons, the CALPUFF’s FB certainty limits
Performance evaluation of dispersion models al- are closer to zero than AERMOD’s, indicating more
ways can be performed by comparing the simulated precise estimations. The average values of CCOF for
results with the observed values. Therefore, AERMOD and CALPUFF has been 0.71 and 0.84,
AERMOD and CALPUFF results have been com- respectively. For all four seasons, CALPUFF’s
pared with observed values for the maximum 1-h CCOF were closer to 1 than AERMOD’s. Consider-
average concentration in nine receptors for this ing these factors, the models used has shown rea-
study. Statistical parameters CCOF and FB for the sonably satisfactory results to estimate the distribu-
two models have been shown in Table 8. Both tion of SO2 ambient concentrations.
models showed good agreement compared with the The Q-Q plots are probably the most illustrative in
measured values. The CALPUFF output has been terms of summarizing models performance in simulat-
also more consistent with the observed values. The ing the concentrations of SO2 compared to observed
average value of the fractional bias (FB) for SO2 in concentrations at nine receptors in the study area (Fig
AERMOD and CALPUFF models has been 0.33 5). Each figure includes the Q-Q plots of three sets of
516 Page 10 of 13 Environ Monit Assess (2016) 188: 516

Spring Summer
AERMOD
AERMOD
CALPUFF
CALPUFF
Scale=1:1
Scale=1:1
Scale=1.5:1
Scale=1.5:1
Scale=1:1.5
Scale=1:1.5
1.0 1.0
Observed

Observed
0.1 0.1
0.1 Simulated 1.0 0.1 Simulated 1.0

Fall Winter
AERMOD AERMOD
CALPUFF CALPUFF
Scale=1:1 Scale=1:1
Scale=1.5:1 Scale=1.5:1
Scale=1:1.5 Scale=1:1.5
1.0 1.0
Observed
Observed

0.1
0.1
0.1 Simulated 1.0
0.1 Simulated 1.0

Fig. 5 Comparison seasonal of CALPUFF and AERMOD simulation results at nine receptor stations in 2014

ambient sampling results in each season as input data. Models unrealistic estimation is because of the accu-
Each point in plots shows a pair of simulated and racy limitation of the devices and also due to back-
observed SO2 concentrations at the monitoring recep- ground SO2 concentrations in the study area and contri-
tors. In general, the simulated average concentrations of bution of the pollutants resulted from other nearby re-
SO2 made by AERMOD and CALPUFF models were fineries and industries or emissions from transportation
similar for all four seasons but CALPUFF model that have not been considered in this study. The distri-
showed better correlation to the observed concentrations butions of simulated SO2 concentrations show that
and a greater number of points within a factor of 1.5 of CALPUFF modeling results have better agreement with
the observed values compared to the AERMOD (Fig 5). the observed concentrations than AERMOD results, and
AERMOD underestimated almost all the concentrations the simulated values of both models are less than the
by more than a factor of 1.5. The highest observed measured values. In conclusion, within the complex
concentrations were within a factor of 1.5 of corre- conditions of topography, CALPUFF offers better
sponding CALPUFF simulated concentrations. results.
Environ Monit Assess (2016) 188: 516 Page 11 of 13 516

Table 8 Validation of AERMOD and CALPUFF results

Model Statistical parameters 2014 spring 2014 summer 2014 fall 2014 winter

AERMOD CCOF 0.64 0.69 0.78 0.73


FB 0.32 0.24 0.34 0.44
CALPUFF CCOF 0.79 0.86 0.88 0.82
FB 0.10 0.07 0.10 0.24

Conclusion of the model is done in the short term, CALPUFF will


have more favorable comparison with AERMOD. (3)
In general, accurate estimation of the air pollutants in AERMOD sensitive to specification of surface proper-
industrial areas has been always considered as a major ties and land use may involve substantial error associat-
step for decision-making concerning implementation of ed with its inability to vary surface properties appropri-
air quality management programs. In this study, ately for the facility source and especially background
AERMOD and CALPUFF dispersion models have been sources. Totally Lagrangian puff model showed better
used to study the dispersion of SO2 emitted from stacks performance than steady-state model especially in the
over a gas refinery for the four seasons in 2014. The complex terrain conditions. In most receptors,
simulated results made by the models have been com- CALPUFF exhibited higher correlation to observed
pared with the observed values in monitoring stations, values and a higher percentage of observations within
using a set of statistical relationships in order to verify a factor of 1.5 of the observations compared to
the accuracy and the performance of models. The results AERMOD.
showed that the overall performance of CALPUFF has In this study, we include also results of studies done
been superior to AERMOD. According to the results of with other sources. Generally, the average ratio between
modeling, different output concentrations have been the simulated and the observed gas pollutant concentra-
mainly obtained from the comparison of the Gaussian tions for AERMOD and CALPUFF varies between 0.7
model (AERMOD) and Lagrangian puff model and 1.41(Table 9). The average ratio of the AERMOD
(CALPUFF). Based on the comparison of the models and CALPUFF estimates to observed values in this
performance to simulate the complex conditions of to- complex topography study were met within the previ-
pography, the following are concluded: (1) CALPUFF ously published range (0.72 and 0.89, respectively). In
estimates greater amounts at the receptor points for an general, the concentration levels simulated by
average period, for example an hour, than AERMOD. AERMOD and CALPUFF have been less than the real
(2) When the simulation of average output concentration amounts. That means that the ambient SO 2

Table 9 Summary of AERMOD and CALPUFF evaluation for stacks and other sources

Study Emission source Simulation software Average ratio of the Notes


simulated to the observed

(Tartakovsky et al. 2013) Quarry AERMOD and CALPUFF Not reported Complex terrain
(Dresser and Huizer 2011) Power plants’ stacks AERMOD 0.85 Near field
(Dresser and Huizer 2011) Power plants’ stacks CALPUFF 1.01 Near field
(EPA 2010) Stacks AERMOD 0.85–1.09 Hilly terrain
(EPA 2010) Stacks CALPUFF 0.98–1.41 Hilly terrain
(Hrebenyk et al. 2007) Line sources with CALPUFF 0.7–0.9 Flat terrain
16 % area sources
(Orloff et al. 2006) Pads of crushed ore AERMOD 0.76 N/A
Jafarigol (present study 2014) Stacks AERMOD 0.72 Complex terrain
Jafarigol (present study 2014) Stacks CALPUFF 0.89 Complex terrain
516 Page 12 of 13 Environ Monit Assess (2016) 188: 516

concentrations are not only due to point sources emis- Hrebenyk, B., McEwen, B., & Limited, S. C. (2007). CALPUFF
air dispersion modeling assessment at EKATI diamond mine
sions but also mobile emission sources and other refin-
(peer reviewed). Vancouver: Independent Environmental
eries and petrochemical industries in the study area have Monitoring Agency.
contribution in this case. IRIMO (2014). http://www.weather.ir/. Accessed 26 June 2014.
Jafarigol, F., Atabi, F., Moattar, F., & Nouri, J. (2015). Predicting
Acknowledgments This paper uses computations which were ambient concentrations of NO2 in a gas refinery located in South
performed with the financial support from the South Pars Gas Pars Gas Complex. International Journal of Environmental
Complex (SPGC). The authors wish to thank the expert’s team Science and Technology, 1–10. doi:10.1007/s13762-015-0870-6.
for their assistance in the preparation of the emission data. Karbassi, A. R., Abbasspour, M., Sekhavatjou, M. S., Ziviyar, F., &
Saeedi, M. (2007). Potential for reducing air pollution from oil
refineries. Environmental Monitoring and Assessment, 145(1),
159–166. doi:10.1007/s10661-007-0025-4.
References Method, G. R. I. (1997). Determination of nitrogen oxides, carbon
monoxide, and oxygen emissions from natural gas-fired en-
gines, boilers and process heaters using portable analyzers.
Abdul-Wahab, S., Ali, S., Sardar, S., Irfan, N., & Al-Damkhi, A. (Vol. 7): Gas Research Institute Method.
(2011a). Evaluating the performance of an integrated Myerson, J., Adams, D., Hale, S., & Jenkins, L. (2003). Analysis
CALPUFF-MM5 modeling system for predicting SO2 emis- of group differences in processing speed: Brinley plots, Q-Q
sion from a refinery. Clean Technologies and Environmental plots, and other conspiracies. Psychonomic Bulletin &
Policy, 13(6), 841–854. doi:10.1007/s10098-011-0360-6. Review, 10(1), 224–237. doi:10.3758/BF03196489.
Abdul-Wahab, S., Sappurd, A., & Al-Damkhi, A. (2011b). Olesen, H. R. (2001) Harmonisation within atmospheric disper-
Application of California puff (CALPUFF) model: a case sion modelling for regulatory purposes, Belgirate. In Ten
study for Oman. Clean Technologies and Environmental years of harmonization activities: past, present, and future,
Policy, 13(1), 177–189. doi:10.1007/s10098-010-0283-7. 7th Int. Conf, Roskilde, Denmark, 2001: National
Abdul-Wahab, S. A. (2006). The role of meteorology on Environmental Research Institute (NERI).
predicting SO2 concentrations around a refinery: a case study Orloff, K. G., Kaplan, B., & Kowalski, P. (2006). Hydrogen
from Oman. Ecological Modelling, 197(1–2), 13–20. cyanide in ambient air near a gold heap leach field: measured
doi:10.1016/j.ecolmodel.2006.02.021. vs. modeled concentrations. Atmospheric Environment,
ASTM (2011a). Standard practice for general ambient air ana- 40(17), 3022–3029. doi:10.1016/j.atmosenv.2005.09.089.
lyzer procedures. ASTM D3249-95. West Conshohocken: Rood, A. S. (2014). Performance evaluation of AERMOD,
ASTM International. CALPUFF, and legacy air dispersion models using the winter
ASTM (2011b). Standard test method for determination of nitrogen validation tracer study dataset. Atmospheric Environment,
oxides, carbon monoxide, and oxygen concentrations in emis- 89, 707–720. doi:10.1016/j.atmosenv.2014.02.054.
sions from natural gas-fired reciprocating engines, combustion Sahin, U., Ucan, O., Bayat, C., & Oztorun, N. (2005). Modeling of
turbines, boilers, and process heaters using portable analyzers. SO2 distribution in Istanbul using artificial neural networks.
ASTM D6522–11. West Conshohocken: ASTM International. Environmental Modeling & Assessment, 10(2), 135–142.
Cimorelli, A. J., Perry, S. G., Venkatram, A., Weil, J. C., Paine, R. doi:10.1007/s10666-004-7262-5.
J., Wilson, R. B., et al. (2004). AERMOD: description of Snyder, W. H., Thompson, R. S., Eskridge, R. E., Lawson, R. E.,
model formulation. North Carolina: U.S. Environmental Castro, I. P., Lee, J. T., et al. (1985). The structure of
Protection Agency. strongly stratified flow over hills: dividing-streamline
Dresser, A. L., & Huizer, R. D. (2011). CALPUFF and AERMOD concept. Journal of Fluid Mechanics, 152, 249–288.
model validation study in the near field: Martins Creek doi:10.1017/S0022112085000684.
revisited. Journal of the Air & Waste Management Tartakovsky, D., Broday, D. M., & Stern, E. (2013). Evaluation of
Association, 61(6), 647–659. AERMOD and CALPUFF for predicting ambient concentra-
EPA U. S. (2010). Air quality document technical support docu- tions of total suspended particulate matter (TSP) emissions
ment: NJ 126. Office of Air Quality Planning and Standards. from a quarry in complex terrain. Environmental Pollution,
Research Triangle Park, NC 27711. 179(0), 138–145. doi:10.1016/j.envpol.2013.04.023.
Ghannam, K., & El-Fadel, M. (2013). Emissions characterization and U.S.EPA (2003). Guidelines for developing an air quality (ozone
regulatory compliance at an industrial complex: an integrated and PM2.5) forecasting program. EPA-456/R-03-002.
MM5/CALPUFF approach. Atmospheric Environment, 69(0), Research Triangle Park, NC: Office of Air Quality,
156–169. doi:10.1016/j.atmosenv.2012.12.022. Planning and Standards.
Gulia, S., Nagendra, S., & Khare, M. (2015). Comparative evalu- U.S.EPA (2004). User's guide for the AERMOD terrain prepro-
ation of air quality dispersion models for PM2.5 at Air cessor (AERMAP). North Carolina 27711: US EPA,
Quality Control Regions in Indian and UK Cities. MAPAN, Environmental Protection Agency.
1–12, doi:10.1007/s12647-015-0149-x. Vijay Bhaskar, B., Jeba Rajasekhar, R. V., Muthusubramanian, P.,
Holnicki, P., & Nahorski, Z. (2015). Emission data uncertainty in & Kesarkar, A. (2008). Measurement and modeling of respi-
urban air quality modeling—case study. Environmental rable particulate (PM10) and lead pollution over Madurai,
Modeling & Assessment, 1–15. doi:10.1007/s10666-015- India. Air Quality, Atmosphere & Health, 1(1), 45–55.
9445-7. doi:10.1007/s11869-008-0004-0.
Environ Monit Assess (2016) 188: 516 Page 13 of 13 516

Willis, G. E., & Deardorff, J. W. (1981). A laboratory study of Yannopoulos, P. (2007). Spatial concentration distributions of
dispersion from a source in the middle of the convectively sulfur dioxide and nitrogen oxides in Patras, Greece, in a
mixed layer. Atmospheric Environment (1967), 15(2), 109– winter period. Environmental Monitoring and Assessment,
117. doi:10.1016/0004-6981(81)90001-9. 135(1–3), 163–180. doi:10.1007/s10661-007-9641-2.

You might also like