Journal of Environmental Management 93 (2012) 121e129

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

Air quality impact assessment of multiple open pit coal mines

in northern Colombia
José I. Huertas*, María E. Huertas, Sebastián Izquierdo, Enrique D. González
Automotive Engineering Research Center, CIMA of Tecnológico de Monterrey, Eduardo Monroy Cárdenas, No 2000, Toluca, México

a r t i c l e i n f o a b s t r a c t

Article history: The coal mining region in northern Colombia is one of the largest open pit mining regions of the world. In
Received 28 April 2011 2009, there were 8 mining companies in operation with an approximate coal production of w70 Mtons/
Received in revised form year. Since 2007, the Colombian air quality monitoring network has reported readings that exceed the
30 July 2011
daily and annual air quality standards for total suspended particulate (TSP) matter and particles with an
Accepted 8 August 2011
equivalent aerodynamic diameter smaller than 10 mm (PM10) in nearby villages. This paper describes
Available online 22 September 2011
work carried out in order to establish an appropriate clean air program for this region, based on the
Colombian national environmental authority requirement for modeling of TSP and PM10 dispersion. A
Keywords:
Air quality modeling
TSP and PM10 emission inventory was initially developed, and topographic and meteorological infor-
Open pit mining mation for the region was collected and analyzed. Using this information, the dispersion of TSP was
Open cut mining modeled in ISC3 and AERMOD using meteorological data collected by 3 local stations during 2008 and
Clean air policies 2009. The results obtained were compared to actual values measured by the air quality monitoring
network. High correlation coefficients (>0.73) were obtained, indicating that the models accurately
described the main factors affecting particle dispersion in the region. The model was then used to
forecast concentrations of particulate matter for 2010. Based on results from the model, areas within the
modeling region were identified as highly, fairly, moderately and marginally polluted according to local
regulations. Additionally, the contribution particulate matter to the pollution at each village was esti-
mated. Using these predicted values, the Colombian environmental authority imposed new decontam-
ination measures on the mining companies operating in the region. These measures included the
relocation of three villages financed by the mine companies based on forecasted pollution levels.
Ó 2011 Published by Elsevier Ltd.

1. Introduction
In 2009, the study mining region in northern Colombia con-
sisted of 7 open pit mines with an approximate coal production of
33 Mtons/year. This area, in combination with a 37 Mtons/year
mine located nearby, but outside of the region of interest, is one of
the world’s largest open pit mining regions. The mining operations
have a positive economic impact on the region but may also have
negative environmental impacts. Therefore, the Colombian
national environmental authority promotes mining while ensuring
that environmental conditions meet quality standards.
Major air pollutants arising from open pit mining operations are
total suspended particulate (TSP) matter and particles with an
equivalent aerodynamic diameter smaller than 10 mm (PM10)
* Corresponding author. Tel.: þ52 722 2799990x2113.
E-mail address: Jose-Ignacio.Huertas-Cardozo@u-bourgogne.fr (J. I. Huertas).
(Sinha and Banerjee, 1997; Chakraborty et al., 2002). The coal
mining activities that generate these particles are drilling, blasting,
overburden loading and unloading, coal loading and unloading,
road transport over unpaved roads and losses from exposed over-
burden dumps, coal handling plants and exposed pit faces
(Chakraborty et al., 2002; Huertas et al., in press). Transportation of
materials has been identified as the main source of TSP and PM10
pollution (Cowherd, 1988; Ghose and Majee, 2000; Chaulya, 2004;
Trivedi et al., 2009; Huertas et al., in press). TSP and PM10 in open
pit mining regions reduce air quality and can cause silicosis, black
lung (CWP) and increased mortality. They also reduce visibility and
affect surrounding flora and fauna (Wheeler et al., 2000; NIOSH,
2005).
In 2006, the Colombian environmental authority initiated an air
quality network to monitor TSP and PM10 in the mining region.
Since 2007, this network has reported readings that exceed the
standards for daily and annual TSP and PM10 concentrations in
some regions near the mining region. In order to establish an

0301-4797/$ e see front matter Ó 2011 Published by Elsevier Ltd.

doi:10.1016/j.jenvman.2011.08.007
122 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129

appropriate clean air program for the region, the Colombian envi-
ronmental authority required a detailed study of the temporal and
spatial variation of surface level TSP and PM10 concentrations to
assess the impact of the mining operations on air quality in the
region and identify areas within the mining region that should be
classified as highly, fairly and moderately polluted based on
national legislation. This information would enable the environ-
mental authority to implement new decontaminating measures
based on the pollution classification. Also, the results of the study
could be used to estimate the contribution of each mine to the
pollution in each population center within the mining region,
thereby allowing the environmental authority to determine the
appropriate contribution of each mining company toward financing
decontamination measures.
It has been found that TSP and PM10 dispersion is strongly
dependent on meteorological conditions (Monn et al., 1995; Tayanc,
2000; Jones et al., 2002; Triantafyllou, 2001, 2003). The highest
concentrations of particulate matter are found within the mine
with concentrations gradually diminishing with increasing
distance from the mine (Jones et al., 2002; Trivedi et al., 2009).
However, an air quality model is required to describe the temporal
and spatial variations of particles around the sparse measurement
stations.
Holmes and Morawska (2006) conducted a review of the
different particle dispersion models available, including Box,
Gaussian, Lagrangian/Eulerian, CFD and aerosol dynamic models.
They concluded that the major weakness in particle dispersion
modeling was a lack of validation studies that compared the

Fig. 1. Above: Location of the two modeling regions used in this study (Google; 2011). Below: Location of emission sources (mines (open circles labeled with M’s) and public road),
air quality monitoring stations (ZM’s) and meteorological stations (2009 wind roses for automatic stations and the start for the observatory station). Altimetry is also shown.
 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129 123

predicted and actual values. Chaulya et al. (2002) compared FDM
(Fugitive dust model) and PAL2 (point, area and lines sources
model) during a winter season in a coal mining region in India. They
found a coefficient of correlation (R2) of 0.66 for PAL2 and 0.75 for
FDM when experimental data from 3 high volume samplers was
compared with model results. Trivedi et al. (2009) modeled TSP
using FDM and obtained an R2 of 0.71 using data from 5 monitoring
stations. Jaiprakash et al. (2010) used AERMOD with one meteo-
rological station and 20 monitoring stations, and obtained an R2 of
0.783 for TSP and 0.741 for PM10. Singh et al. (2006) modeled
particle dispersion in an iron ore mine using FDM and ISCST3
during a winter season. They used 7 available monitoring stations
and found an index of agreement of 82% for FDM and 44% for ISCT3.
NIOSH (2005) included the work of Reed (2003) and concluded that
the ISC3 model cannot accurately predict dust concentrations from
mining operations because it is a model designed for predicting
dust dispersion from stationary sources. AERMOD was also tested
but not at mine sites. However, none of these models are accepted
for modeling particle dispersion by environmental authorities such
as the US EPA (US EPA, 2009), except AERMOD. AERMOD is
currently the model recommended by the US EPA for modeling air
quality for the type of application studied here (US EPA, 2009). Until
2004, ISC3 was the model recommended by the US EPA (US EPA,
2009) and is still the model used by Colombian mining companies.
This paper describes the practical implications and difficulties
modeling air quality for a real application where input data is
available for long periods of time. It describes the work done for the
Colombian environmental authority to assess the impact of mining
operations in northern Colombia on air quality in the region. Data
from multiple meteorological stations collected over 2 years was
used to characterize the meteorology of the region. A standard
methodology was used to estimate emissions from 82 sources.
AERMOD and ISC3 were used in conjunction with guidelines
established by more than 26 environmental authorities. Actual TSP
and PM10 concentrations collected over more than two years from
19 monitoring stations were used to validate the results obtained
by the air quality models. This paper contributes describing the
goodness of using linear regression analysis to validate air quality
model results for cases with input data with high uncertainties.
Furthermore, to our knowledge, this is the first time an air quality
model of open pit mines has been used to establish new environ-
mental policies.
2. Methods
information was obtained from the Geographic Information
Resources System database (Webgis, 2008). The elevation infor-
mation has a resolution of 0.9 km, while the information regarding
land coverage has a resolution of 1 km (Scire et al., 2000) (Fig. 1).
2.1.1.2. Meteorology. Due to the region’s topographical features, the
meteorology of the region was characterized by data collected from
several meteorological stations. 30 sets of data from 7 meteoro-
logical stations were obtained from 2008 to 2009. Wind roses,
hourly averages, monthly averages and variances of the primary
meteorological parameters were obtained. A minimum of 95% data
availability and the absence of atypical data were used as criteria to
select the final meteorological information used for the modeling
process. Table 1 shows the characteristics of the meteorological
stations selected for this study and Fig. 1 their location. The Borrego
and Calenturitas meteorological stations characterized the flat area
while the Cerrolargo meteorological station characterized the
mountainous area. Cloud cover or cloudiness information was
obtained from the Agustin Codazzi observatory station located
approximately 57 km to the north of the mining region (Fig. 1). This
station records cloud cover at 07:00, 13:00 and 19:00, and linearly
interpolated for the rest of the time. Fig. 1 shows wind roses
obtained from these data sets for 2009 where a strong difference
was observed between the wind patterns obtained for the flat and
mountainous areas. Taking into account that the main pollution
problems are found in the flat area, only Borrego and Calenturitas
meteorological data were studied further.
Given the absence of radio soundings in the region, it was not
possible to use the meteorological pre-processing module of AER-
MOD to obtain secondary meteorological parameters, such as the
height mechanical and convective mixing layer. Therefore, they
were calculated using the methodology adopted by the National
Meteorological Institution (IDEAM, 2008; Sozzi and Ruiz, 1998).
Additionally, the methodology implemented in AERMET (US EPA,
2004) was also implemented here. A high correlation was found
between the secondary parameters found using these two meth-
odologies (Fig. 2). It was also found that the modeling results
showed no significant differences when using either of these two
methodologies.
2.1.1.3. Emissions. The main sources of emissions in the region are 7
mines (M1 to M7 in Fig. 1 and Tables 2 and 3) and one public road.
No other economical activities in the region could be considered as
relevant sources of emissions. There are no intense agricultural

2.1. Generation of air quality models Table 1

Meteorological information used to simulate TSP and PM10 dispersion in the study
The general methodology for modeling air quality is divided into area.

3 stages: data input (geography, meteorology and emissions), Name Borrego Calenturitas Cerrolargo Agustín
model set-up and post-processing. Codazzi

Type Automatic Automatic Automatic Observatory

2.1.1. Data input Location x 668.4 665.6 689.24 690.1
2.1.1.1. Geography. The open pit coal mining region is located in (UTM Km) y 1059.5 1067.3 1056.9 1089.3
northern Colombia, west of Los Andes Mountains and south of the Primary Velocity O O O e
Santa Marta Sierra Nevada. Fig. 1 shows the mining region and Parameters Direction O O O e
Max Velocity O O O e
locations of the emission sources (mines) and population centers. Max Direction O O O e
Two modeling regions were defined. The first region is defined by Temperature O O O e
the southwestern UTM 18P coordinate (610, 1010) with a coverage Pressure O O O e
area of 89  89 km2 and resolution of 1.32 km. The second region is Radiation O O e
Rainfall O O O O
defined by the southwestern UTM 18P coordinate (650, 1050) with
Humidity O O O e
a coverage area of 45  25 km2 and resolution of 0.5 km. Evaporation O O O e
The region’s topography is characterized by a large flat area Cloud Cover e e e O
bordered by a region of high mountains (Fig. 1). The area of main Resolution Minutes 30 15 30 3 readings
interest is the flat area where villages with high levels of TSP per day
% Availability 98.8 99.7 99.7 99.6
concentration are located. The elevation and land coverage digital
124 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129

Convective mixing height Mechanical mixing height

2500 2500
y = 1.1719x + 1.9762 y = 0.0007x 2 + 0.7392x - 28.647
R² = 0.9935

Zm Aermet (m)
2000 2000 R² = 0.9957

Zc Aermet (m)
1500 1500

1000 1000

500 500

0 0
0 500 1000 1500 2000 0 500 1000 1500
Zc Sozzi (m) Zm Sozzi (m)

Friction velocity Monin Obukov length

1.0 900
y = 1.007x + 3.561
y = 0.9638x - 8E-05 800 R² = 0.954
U* Aermet (m/s)

0.8 R² = 1 700

L Aermet (m)
600
0.6
500
0.4 400
300
0.2 200
100
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 0 200 400 600 800 1000
U* Sozzi (m/s) L Sozzi (m)

Fig. 2. Comparison of the secondary meteorological parameters obtained using the methodology adopted by the national meteorological authority (Sozzi) and the methodology
implemented in AERMET.

activities and the activities in the villages are limited to several
small commercial businesses.
Inside each mine there are 6 area sources of emissions (Pit,
dump, stock and pit-dump via, pit-stock via and beltway). All mines
(except M1 and M6) adopted the standard methodology reported
by Huertas et al. (in press) to estimate their annual TSP and PM10
emissions for 2008 and 2009, and to project their emissions for
2010. The results for 2009 are shown in Tables 2 and 3. The
methodology consist in inventorying for each sub-activity of the
coal production process, the TSP and PM10 emission based on the
intensity of the activity and the corresponding emissions factors
recommended by the US EPA and other environmental agencies
(Huertas et al., in press). The methodology standardizes the sub-
activities and the emission factors being considered. It also stan-
dardizes the way of allocating the emissions into the 6 area sources
listed before.
Mines 1 and 6 reported their TSP emissions using their own
methodology. They did not reported PM10 emissions. The accuracy
of all the mines emissions estimates and projections were assessed
by applying the techniques described by Huertas et al. (in press). It
consists of statistically crosschecking historical mines operational
parameters and emissions estimates to identify potential sources of
errors.
The public road is 36 km long and is used by most of the mining
companies to transport their coal in 6e8 tons trucks. In 2008,
segments of this road located in villages were paved by the mining
companies under instructions from the Colombian environmental
authority. Currently, 18 km of this road remains unpaved. A
collection of 40 straight sections of the unpaved road were
considered as area sources of emissions. The emission from each
section was estimated as the amount of re-suspended particles
caused by the pass by vehicles as function of their weight, vehicle
average speed and silk and humidity content of the road.
2.1.2. Model setup
Both AERMOD and ISC3 models were used in this study. Note
that both models are Gaussian models. Implementation of both
models followed the guidelines reported by MAVDT (2008), which
were obtained after consulting and analyzing the guidelines
established by more than 26 environmental authorities from
several countries for the process of conducting air quality assess-
ments for regulatory purposes.

Table 2
Open pit coal mine TSP emissions for 2009. Pit, dump, stock and pit-dump via, pit-stock via and beltway are the area sources considered within mine (M’s).

Sources TSP emission intensities g/(m2-s)

M1 M2 M3 M4 M5 M6 M7
Pit 5.0681E-05 3.4910E-06 1.1912E-05 9.1367E-04 9.8477E-06 2.6908E-05 1.2887E-05
Dump 8.6049E-06 5.4693E-06 5.8058E-06 4.3629E-04 1.0434E-05 6.6505E-06 7.0641E-06
Stock 3.5024E-05 2.0435E-04 1.3037E-03 2.0729E-03 2.9833E-03 1.4394E-04 5.3191E-05
Pit-Dump via 2.1887E-04 5.5485E-06 1.4318E-04 2.1503E-04 3.5172E-05 7.2359E-05 2.6565E-05
Pit-Stock via 4.3951E-05 1.0493E-07 1.4638E-05 2.3989E-07 1.1574E-06 8.8889E-04 9.9652E-06
Beltway 5.4694E-05 1.4430E-07 0.0000Eþ00 9.3328E-07 3.1034E-06 4.3277E-05 3.8399E-06
Total TSP emissions g/(m2-s) 3.5713E-04 2.1911E-04 1.4792E-03 3.6390E-03 3.0430E-03 1.1820E-03 1.1351E-04
Total TSP emissions kg/year 3.9276Eþ06 7.4812Eþ05 2.3139Eþ06 1.6727Eþ06 3.0230Eþ06 1.4819Eþ07 9.0708Eþ05
Total coal production ton/year 1.8148þ06 1.038Eþ06 5.984Eþ06 6.000Eþ05 4.730Eþ06 1.840Eþ07 1.600Eþ06
 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129 125

Table 3
Open pit coal mine PM10 emissions for 2009. Pit, dump, stock and pit-dump via, pit-stock via and beltway are the area sources considered within mine (M’s).

Sources PM10 emission intensities g/(m2-s)

M1 M2 M3 M4 M5 M6 M7
Pit N/A 9.1190E-07 3.6886E-06 2.1817E-04 2.2692E-06 N/A 2.7715E-06
Dump N/A 3.3140E-06 3.5163E-05 1.2939E-04 7.4352E-06 N/A 2.0763E-06
Stock N/A 1.3114E-05 1.1823E-04 3.7075E-04 3.1944E-04 N/A 3.3714E-06
Pit-Dump via N/A 1.0347E-06 7.6142E-05 5.5325E-05 9.6155E-06 N/A 4.9538E-06
Pit-Stock via N/A 1.9638E-08 3.0505E-06 6.1307E-08 3.2104E-07 N/A 1.8583E-06
Beltway N/A 2.6908E-08 0.0000Eþ00 2.4012E-07 8.4787E-07 N/A 7.1606E-07
Total PM10 emissions g/(m2-s) N/A 1.8421E-05 2.3627E-04 7.7394E-04 3.3993E-04 N/A 1.5748E-05
Total PM10 emissions kg/year N/A 2.5762Eþ05 2.5519Eþ06 4.1201Eþ05 8.0099Eþ05 N/A 1.5690Eþ05
Total coal production ton/year 1.8148Eþ06 1.0438Eþ06 5.9984Eþ06 6.0000Eþ05 4.7230Eþ06 1.8940Eþ07 1.6000Eþ06

N/A: Not available.

It was found that the AERMOD and ISC3 models limited the
number of receptors to 1200 and 1500 respectively. This restriction
limits the discrete receptor screen resolutions to levels that are not
acceptable when visualizing and determining maximum concen-
tration values in the vicinity of mining companies. In order to
increase the resolution, simulations considering all the sources
were run for a region of output receptors of 89  89 km2 and for
another of 45  25 km2. The output receptor region was also
divided into 4 contiguous areas and the model was run for each
area. These alterations made it possible to increase the receptor
screen resolution from 2.46 km to 1.32 km for the larger region, and
from 1 km to 0.5 km for the smaller region.
Even though PM10 has higher correlation with public health
than TSP, this work was limited to TSP because input data for PM10
was incomplete (mines 1 and 7 did not report PM10 emission), the
environmental authority was interested in TSP (local regulation for
open pit mining is based in TSP) and finally, because it is possible to
obtain PM10 results directly as a fraction of TSP results (emissions
Table 4). Additionally, there was a high statistical correlation
(r ¼ 0.857) between the values estimated by AERMOD and the
values measured by the monitoring system when using the Cal-
enturitas meteorology (Fig. 4, Table 4) while the levels of correla-
tion for ISC3 were lower (r < 0.705).
The coefficient of determination R2 evaluates the ability of the
model results of describing variability in the experimental
measurements independently of systematic variations. For the case
of AERMOD using the Calenturitas meteorology R2 ¼ 0.734 indi-
cates that there is a 1R2 ¼ 26.6% of the measured TSP concen-
trations that is not explained by the model results. It could be
calibration problems of the monitoring stations or un-considered
sources of emissions affecting no systematically all the
A
250
2009
TSP a nnua l g eo metri c mea n

factors for PM10 for open pit mining activities are usually expressed 200 2008
as a fraction of TSP, and AERMOD and ISC do not distinguish TSP 2007
co ncentra ti o n

from PM10 or any other inert pollutant). 150

(µ g / m³)

2.1.3. Post-processing 100 µg/m³

100
The two models were used to estimate daily TSP averages for
2008 and 2009 and project values for 2010. It was necessary to
50
develop a tool in Matlab to visualize the obtained results. To verify
that the models were accurately modeling the main factors
affecting particle dispersion in the region, model results were 0 ZM 1 2

ZM 1 3

ZM 1 4
ZM 1

ZM 2

ZM 3

ZM 5

ZM 6

ZM 7

ZM 9

compared to measured results and a sensitivity analysis was

developed.

3. Results and discussion B 120 2009

PM1 0 a nnua l a ri thmetic a v era g e

3.1. Comparison of predicted and measured data 2008

100
The AERMOD and ISC3 models were used to estimate TSP 2007
80 70 µg/m³
co ncentra tio n

concentrations at the locations where the environmental authority

(µ g / m³)

measured this pollutant during 2008 and 2009 (ZM’s in Fig. 1). It is 60
important to note that according to the US EPA (2005), AERMOD
produces better agreements when it is compared in long term 40
averages. Therefore, comparisons of predicted and measured data
should be made on an annual but not daily basis. Fig. 3 shows the 20
annual geometric mean TSP concentrations and annual average
PM10 concentrations obtained by the air quality monitoring 0
ZM 1
ZM 2
ZM 3
ZM 4
ZM 5
ZM 7
ZM 9
ZM 1 1
ZM 1 2
ZM 1 3
ZM 1 5
ZM 1 6
ZM 1 7
ZM 1 8
ZM 1 9

stations. Fig. 4 compares the predicted and measured TSP

concentrations in terms of annual geometric means, in accordance
with the procedures established by the Colombian national air
quality standards.
It was found that a correlation exists between the predicted and
measured results at a very high level of confidence (1-p ¼ 99.5%,
Fig. 3. TSP annual geometric mean (A) and PM10 annual arithmetic average (B)
concentrations measured by the air quality network in the monitoring zones (ZM’s) as
described in Fig. 1. The horizontal lines with the labels indicate the corresponding
national air quality standards.
126 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129

A 200
Borrego
monitoring stations such as nearby agricultural activities, unpaved
roads or uncovered land exposed to wind erosion. The stations ZM9
µg/m3)

180 y = 0.617x + 55.35 and ZM14 are the ones that exhibit the largest dispersion. The
160 R² = 0.418 Calenturitas result R2 > 0.7 can be considered as high for this application given
Measured TSP

140
concentration (µ

y = 0.597x + 40.17 the multiple sources of uncertainty involved in the modeling

120 R² = 0.497 process. Furthermore, the AERMOD correlation level is high
100
compared to modeling studies described previously (Chaulya et al.,
80
2002; Trivedi et al., 2009; Jaiprakash et al., 2010; Singh et al., 2006;
60
Reed, 2003).
40 The slope of the linear regression indicates the value by which
20 TSP concentration is underestimated. In this work, for the case of
0 the Calenturitas meteorological data, it was found that AERMOD
0 50 100 150 200 250 300 underestimates by 8.24 while ISC3 overestimates by 1/0.60 the TSP
ISC estimated TSP concentration ( µg/m 3) concentrations. Assuming that the air quality model and the
geographical and meteorological data are accurate, the underesti-
mation of the results with respect to the measured TSP concen-
B 200
Calenturitas
trations is due to inaccuracies of the emissions inventories.
180 y = 8.238x + 42.10 The constant of the linear regression indicates the background
concentration (µg/m3)

160 R² = 0.733 concentration estimated by the model. The air quality network uses
Borrego
Measured TSP

140 y = 5.870x + 58.56 station ZM11 to keep track of background concentration in the
120 R² = 0.614 region. In 2009 this station reported an annual mean concentration
100 of 28 mg/m3. The difference between the estimated and measured
80 background concentrations indicates the existence of additional
60 sources of pollutants that affect systematically all the monitoring
40 stations. In this work it was found that AERMOD and ISC3 estimate
similar background concentration which is w14 mg/m3 above the
20
measured background concentration.
0
Literature report several additional metrics to evaluate the
0 5 10 15 20 25
capability of air quality models to reproduce, in time and space at
AERMOD estimated TSP concentration (µg/m 3)
surface level, experimental observations (Walker et al., 1999; Mosca
et al., 1998; ASTM, 2000; Zawar-Reza et al., 2005; Li et al, 2011).
Fig. 4. Comparison of measured TSP annual geometric mean concentration values
from the air quality monitoring network and predicted values from ISC3 and AERMOD. Table 4 shows the values obtained for the most frequently used
Meteorology used was recorded at the Borrego and Calenturitas stations during 2009. metrics: Index of agreement, bias, fractional bias, geometric mean
bias, normalized mean square error, root of the mean square error

Table 4
Comparison of ISC and AERMOD results, using Calenturitas and Borrego meteorology for 2009, with measured annual mean geometric TSP concentration values from the air
quality network.

ISC AERMOD Perfect value Range

a
Calenturitas Borrego Calenturitas Calenturitas Borrego
m 0.60 0.62 8.24 1.00 5.87 1 N to N
b 40.2 55.4 42.1 0.0 58.6 0 N to N
R2 0.497 0.418 0.734 0.734 0.614 1 0e1
r 0.705 0.647 0.857 0.857 0.784 1 0e1
p 0.034 0.060 0.005 0.005 0.012 0 0e1
d 0.82 0.71 0.28 0.91 0.28 1 0e1
Bias 0.81 27.10 93.79 0.00 93.71 0 N to N
FB 0.01 0.31 1.74 0.00 1.73 0 2 to 2
MG 0.87 0.64 0.06 1.02 0.05 1 0

NMSE 0.12 0.27 14.15 0.04 13.92 0 0

RMSE 35.00 44.52 100.97 21.15 100.70 0 0

FAC2 89% 78% 0% 100% 0% 100% 0e100%

m: Slope of the linear regression.

b: Offset of the regression (model estimated background concentration).
R2: Coefficient of determination.
r: Coefficient of correlation.
(1-p): Level of confidence for which it can be stated that there exists a correlation between the predicted values from the air quality models and measured values from the air
quality network.
d: Index of agreement.
FB: Fractional bias.
MG Geometric mean bias.
NMSE: Normalized mean square error.
RMSE: Root of mean square error.
FAC2: Fraction of data for which 0.5 < Pi/Mi < 2, where P and M are predicted and measured data, respectively.
a
Data adjusted such as m ¼ 1 and b ¼ 0.
 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129 127

TSP concentration (µg/m3) 25 400 same constant such as m ¼ 1, and including the background

TSP concentration (µg/m3)

350 concentration, results become unbiased (column marked with * in
y = 1.192x - 0.958
Borrego -AERMOD

20 Table 4). Under these circumstances, results can only be evaluated

R² = 0.848

15
10
y = 16.40x - 6.989
5 R² = 0.826
0 0
0 5 10 15
Calenturitas- AERMOD TSP
concentration (µg/m3 )
Fig. 5. Comparison of the TSP annual geometric mean concentrations obtained from
AERMOD and ISC3 for the study villages using Calenturitas and Borrego meteorology
for 2009.
and fraction of data for which 0.5 < Pi/Mi < 2, where P and M are
predicted and measured data, respectively. All these metrics
measure at the same time dispersion and deviation from average
values. Table 4 shows that according to these metrics, the combi-
nation of AERMOD with Calenturitas represents the worst case
performance. This is due to the fact that this combination exhibit
the largest slope in the linear regression (m) and therefore it is the
most biased. As it was said before, this is due to un-accuracies in the
emission inventories. Affecting all the emission sources by the
Calenturitas-ISC
300 by their dispersion, and then, according to all these metrics, the
250 results obtained by AERMOD using the Calenturitas meteorological
200 data, are the ones with the lowest dispersion and therefore they are
the ones that best reproduces the experimental measurements of
150
TSP concentrations.
100
50 3.2. Comparison between models
Fig. 5 compares the results obtained by AERMOD and ISC3 using
20
the Calenturitas meteorology for 2009 at every village in terms of
annual geometric mean TSP concentrations. It shows a high
correlation between these two results (R2 ¼ 0.848 in Fig. 5). It
means that ISC3 results can be obtained from AERMOD results and
vice versa and therefore the use of either of them will lead to
similar conclusions. Similarly, Fig. 5 show a high correlation
between AERMOD results using the Calenturitas and Borrego
meteorological data. Attending US EPA recommendations, it was
decided to continue using AERMOD for the subsequent results.
Furthermore, it was decided to use AERMOD with the Calenturitas
meteorological data since that combination is the one that best
reproduces the experimental measurements of TSP concentrations.
3.3. Sensitivity analysis
A sensitivity analysis was carried out to evaluate the effect of
emission data on the model results. The results reported by

Fig. 6. TSP annual geometric mean concentration obtained from AERMOD using 2009 Calenturitas meteorology. The colored section represents readings exceeding the annual air
quality standards.
128 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129

Fig. 7. Identification of highly, fairly, moderately and marginally polluted areas within the study area based on AERMOD results using Calenturitas meteorology for 2009.

AERMOD when emissions increased by 10%, 20%, 50% and 100%

resulted in a sensitivity of 95.3%. This means that for every annual
increment of 1% (274.1 Mg) in TSP emissions from the mines an El Vallito
increment of 0.953 mg/m3 in the concentration of TSP at the surface
Casacará M1
level at every point of the modeling area will be observed. There-
El Cruce
fore, it can be concluded that the results presented here are highly M2
Poponte
sensitive to changes in emissions data. M3
Arenas Blancas
Becerril M4
3.4. Assessment of the impact of mining on air quality
La Victoria de San Isidro
M5
La Jagua de Ibirico
Fig. 6 shows the TSP results obtained by AERMOD using the M6
Prado
Calenturitas meteorology for 2009 and indicates that the air quality
model identifies areas near mines where the estimated annual Cuatro vientos M7
Receptor

geometric mean concentrations of TSP at the surface level exceed the ZM15 - Chiriguana
Vias
national annual standard. Regardless of the model and the meteo- ZM14 - Estados Unidos
rology used, the results showed that the areas with the highest ZM13 - El Hatillo
concentrations correspond to the location of the emission sources. ZM12 - Rincon Hondo
TSP concentrations decreased with increasing distance from the ZM11 - Minguillo
emission source. This observation is consistent with the experi- ZM9 - Jagua vía
mental results reported by Jones et al. (2002) and Trivedi et al. ZM7 - Plan Bonito
(2009). There were no areas of high TSP concentrations generated ZM6 - Boqueron
by the simultaneous contribution from different emission sources. ZM5- La aurora
The largest region with readings exceeding the annual TSP standard ZM4 - Palmitas
was located around mine 6. As expected, it was found that the results ZM3 - La loma 2
are independent of the resolution of the screen used.
ZM2 - Jagua Centro
ZM1 - La loma
3.5. Identification of polluted areas
0 100 200 300 400
The Colombian environmental authority can implement new
decontaminating measures only in areas that have been classified as Contribution from each mine ( g/m3)
highly, fairly, moderately or marginally polluted. Highly, fairly, Fig. 8. Contribution of each mine (M’s) to village air pollution predicted for 2010 by
moderately and marginally polluted areas have had a TSP 24 h AERMOD using Calenturitas meteorology from 2008 to 2009 and predicted emissions
average concentration above the national annual air quality standard for 2010.
 J.I. Huertas et al. / Journal of Environmental Management 93 (2012) 121e129
(100 mg/m3) for more than 75%, 50e75%, 25e50% or 0e25% of the
previous year respectively. AERMOD with the Calenturitas meteo-
rology for 2009 was used to identify these polluted areas (Fig. 7).
3.6. Contribution of each mine to the pollution at each village
Once the highly, fairly and moderately polluted areas were
identified, the environmental authority initiated plans to imple-
ment a set of new decontaminating measures for these areas. One
of the measures implemented was the relocation of three of the
most polluted villages. The cost of implementing such measures
was charged to the mining companies operating in the region
according to their level of contribution to the pollution at each
village as predicted for 2010 using the AERMOD model with Cal-
enturitas meteorology for 2008 and 2009 and predicted emissions
data for 2010. Fig. 8 shows the results obtained and also compares
the results against the annual TSP air quality standard, identifying
the villages that are expected to have air quality problems.
4. Conclusions
In order to establish a clean air program for one of the largest
open pit mining regions of the world, dispersion of TSP and PM10
was modeled by means of ISC3 and AERMOD for 2008 and 2009.
Dispersion predictions were also made for 2010. A high correlation
was observed between the results obtained using the models and
data measured by the monitoring stations, indicating that the
models accurately described particle dispersion phenomena in the
region. However, the results reported by ISC3 and AERMOD for the
same meteorology were quantitatively different. AERMOD esti-
mated smaller concentrations than ISC3.
AERMOD was used to identify highly, fairly, moderately and
marginally polluted areas. According to national regulations, only
in these areas can the environmental authority impose new
decontaminating measures. The model was used to quantify the
average contribution of each mine to the pollution of the pop-
ulation centers located nearby and this information was used to
determine the financial contribution required from each mining
company for the decontaminating measures.
Acknowledgments
This project was financed by the Ministry of the Environment of
Colombia. The authors also express their gratitude to engineers
Dumar Camacho, Natalia Navarrete and Jessica Garzon for their
contributions to this project.
References
ASTM, 2000. Standard Guide for Statistical Evaluation of Atmospheric Dispersion
Model Performance. American Society for Testing and Materials, Designation D
6589e00. ASTM.
Chakraborty, M.K., et al., 2002. Determination of the emission rate from various
opencast mining operations. Environmental Modeling & Software 17, 467e480.
Chaulya, S.K., 2004. Assessment and management of air quality for an opencast coal
mining area. Environmental Management 70, 1e14.
Chaulya, S.K., Ahmad, M., Singh, R.S., Bandopadhyay, L.K., Bondyopadhay, C.,
Mondal, G.C., 2002. Validation of two air quality models for Indian mining
conditions. Environmental Monitoring and Assessment 82, 23e43.
129
Cowherd, C.E., 1988. Control of Open Fugitive Dust Sources, EPA-450/3-88-008. US
Environmental Protection Agency.
Ghose, M.K., Majee, J., 2000. Assessment of dust generation due to opencast coal
mining-an Indian case study. Environmental Monitoring and Assessment 61,
255e263.
Jaiprakash, Singh, G., Pal, A.K., 2010. Air pollution dispersion modeling performance
for mining complex. Environment & We an International Journal of Sciences &
Technology 5, 205e222.
Holmes, N.S., Morawska, L., 2006. A review of dispersion modeling and its appli-
cation to the dispersion of particles: an overview of different dispersion models
available. Atmospheric Environment 40, 5902e5928.
Huertas J., Camacho D., Huertas M. Standardized emissions inventory methodology
for open pit mining areas. Environmental Science and Pollution Research, in
press.
IDEAM, 2008. Instituto de Hidrología, Meteorología y Estudios Ambientales de
Colombia. http://institucional.ideam.gov.co.
Jones, T., Blackmore, P., Leach, M., Matt, B.K., Sexton, K., Richards, R., 2002. Char-
acterization of airborne particles collected within and proximal to an opencast
coalmine: south Wales. UK Environmental Monitoring and Assessment 75,
293e312.
Li, L., Chen, C.H., Fu, J.S., Streets, D.G., Huang, H.Y., Zhang, G.F, et al., 2011. Air quality
and emissions in the Yangtze River Delta, China. Atmospheric Chemistry and
Physics 11, 1621e1639.
Mavdt-Ministerio de Medio Ambiente y Desarrollo Territoral de Colombia, 2008.
Guía Nacional de Modelación de la Calidad del Aire.
Monn, Ch., et al., 1995. Particulate matter < 10 mm (PM10) and total suspended
particles (TSP) in urban, rural and alpine air in Switzerland. Atmospheric
Environment 29, 2565e2573.
Mosca, S., Graziani, G., Klug, W., Bellasio, R., Bianconi, R., 1998. A statistical meth-
odology for the evaluation of long-range dispersion models: an application to
the ETEX exercise. Atmospheric Environment 32 (24), 4307e4324.
NIOSH-National Institute for Occupational Safety and Health, 2005. Significant Dust
Dispersion Models for Mining Operations IC 9478. (USA).
Reed, W.R., 2003. An improved model for prediction of PM10 from surface mining
operations (Dissertation). Virginia Polytechnic Institute and State University,
Department of Mining and Minerals Engineering, (USA).
Scire, J., Robe, F., Fernau, M., Amartino, R., 2000. User’s guide for the CALPUFF
dispersion model. US EPA.
Singh, G., Prabha, J., Giri, S., 2006. Comparison and performance evaluation of
dispersion models FDM and ISCST3 for a gold mine at GOA. Journal of Industrial
Pollution Control 22, 297e303.
Sinha, S., Banerjee, S.P., 1997. Characterization of haul road in Indian open cast iron
ore mine. Atmospheric Environment 31, 2809e2814.
Sozzi, R., Ruiz, F., 1998. Implemetación de metodologías para la meteorología y la
micrometeorlogía de dispersión de los contaminante en el aire Bogotá.
Tayanc, M., 2000. An assessment of spatial and temporal variation of sulphur
dioxide levels over Istanbul, Turkey. Environmental Pollution 107, 61e69.
Triantafyllou, A.G., 2001. PM10 pollution episodes as a function of synoptic
climatology in a mountainous industrial area. Environmental Pollution 112,
491e500.
Triantafyllou, A.G., 2003. Levels and trends of suspended particles around large
lignite power station. Environmental Monitoring and Assessment 89, 15e34.
Trivedi, R., Chakraborty, M.K., Tewary, B.K., 2009. Dust dispersion modeling using
fugitive dust model at an opencast coal project of western coalfields limited,
India. Journal of Scientific & Industrial Research 68, 71e78.
US EPA, 2004. User’s Guide for the AERMOD Meteotological Preprocessor
(AERMET).
US EPA, 2005. 40 CFR Part 51-Revision to the Guideline on Air Quality Models:
Adoption of a Preferred General Purpose (Flat and Complex Terrain) Dispersion
Model and Other Revisions. Final Rule.
US EPA, 2009. Technology Tranfer Network Support Center for Regulatory Atmo-
spheric Modeling. Obtenido de. http://www.epa.gov/scram001/.
Walker, S., Slordal, L., Guerreiro, C., Gram, F., Gronskei, K., 1999. Air pollution
exposure monitoring and estimation. Part II. Model evaluation and population
exposure. Journal of Environmental Monitoring 1, 321e326.
Webgis, 2008. Geographic Information Systems Resources. http://www.webgis.
com/.
Wheeler, A.J., Williams, I., Beaumont, R.A., Manilton, R.S., 2000. Characterization of
particulate matter Sampled during a study of Children’s Personal exposure to
airborne particulate matter in a UK urban Environment. Environmental Moni-
toring and Assessment 65, 69e77.
Zawar-Reza, P., Kingham, S., Pearce, J., 2005. Evaluation of a year-long dispersion
modeling of PM10 using the mesoscale model TAPM for Christchurch, New
Zealand. Science of the Total Environment 349, 249e259.