Journal of Geoscience and Environment Protection
2014. Vol.2, No.1, 6-11
Published Online January 2014 in SciRes (
Analysis of Contributions to NO2 Ambient Air Quality Levels in
Madrid City (Spain) through Modeling. Implications for the
Development of Policies and Air Quality Monitoring
Rafael Borge*, David de la Paz, Julio Lumbreras, Javier Pérez, Michel Vedrenne
Laboratory of Environmental Modelling, Technical University of Madrid (UPM), Spain
Email: *
Received November 2013
As environmental standards become more stringent (e.g. European Directive 2008/50/EC), more reliable
and sophisticated modeling tools are needed to simulate measures and plans that may effectively tackle air
quality exceedances, common in large cities across Europe, particularly for NO2. Modeling air quality in
urban areas is rather complex since observed concentration values are a consequence of the interaction of
multiple sources and processes that involve a wide range of spatial and temporal scales. Besides a consis-
tent and robust multi-scale modeling system, comprehensive and flexible emission inventories are needed.
This paper discusses the application of the WRF-SMOKE-CMAQ system to the Madrid city (Spain) to
assess the contribution of the main emitting sectors in the region. A detailed emission inventory was
compiled for this purpose. This inventory relies on bottom-up methods for the most important sources. It
is coupled with the regional traffic model and it makes use of an extensive database of industrial, com-
mercial and residential combustion plants. Less relevant sources are downscaled from national or regional
inventories. This paper reports the methodology and main results of the source apportionment study per-
formed to understand the origin of pollution (main sectors and geographical areas) and define clear targets
for the abatement strategy. Finally the structure of the air quality monitoring is analyzed and discussed to
identify options to improve the monitoring strategy not only in the Madrid city but the whole metropolitan
Keywords: Air Quality Modeling; Source Apportionment; NO2; CMAQ; Urban Air Quality; Ma drid
Despite recent efforts made in Europe, a large percentage of
the European population is currently exposed to air pollutant
concentrations above the European Union and World Health
Organization reference levels (EEA, 2013). This is mainly rele-
vant to urban areas, where the majority of the European popula-
tion lives, leading to adverse effects on public health. Even
from the regulatory point of view, there are still pending issues
in Europe. For instance, air concentrations of NO2 lags clearly
behind the decreasing trend of NOX emissions (Guerreiro, C. et
al., 2010). As a consequence, the main cities in Europe are
finding serious difficulties to comply with the air quality stan-
dards defined in the EU Directive 2008/50/EC for NO2. Non-
attendant areas and urban agglomerations are legally bound to
develop and apply air quality plans (AQPs) to meet legal stan-
dards. Such plans and strategies are based on abatement meas-
ures that usually entail significant social, economic and politi-
cal costs. Therefore, it is important to guarantee that any meas-
ure is actually aimed at the sectors causing air quality problems
so they can effectively improve air quality levels.
The inherent complexity of urban environments requires si-
mulation tools to assess air quality levels to be able to support
the analysis and evaluation of a variety of policies and emission
abatement measures previous to their implementation (Denby,
B. et al., 2011).
This evaluation must begin with a source apportionment
analysis that can provide essential information regarding the
basic emission abatement strategy, the maximum feasible air
quality improvement related to the main emitting sectors as
well as the external constrains. This paper reports on the source
apportionment carried out for the Madrid city (Spain) for the
development of an AQP to comply with NO2 standards in the
near future (Borge, R. et al., 2014). Despite satisfying a legal
requirement, this analysis is also useful to assess the layout of
the air quality monitoring stations and to identify options to
improve the monitoring strategy in the region.
Case Study
Madrid is the capital and largest city in Spain, located in the
centre of the Iberian Peninsula with a total population of 5 mil-
lion people in its metropolitan area. Madrid City Council and
surrounding municipalities are connected through a dense road
network and constitute a continuous urban area with a large
density of air quality monitoring stations (Figure 1).
Despite the experienced population and traffic increase, air
quality levels have improved in the city over the last decade.
However, some pollutants like nitrogen dioxide (NO2) still
exceed the limit values (LV) according to the European legisla-
tion. The NO2 annual average recorded in most of the city’s
traffic air quality monitoring stations is above the LV (40
µg/m3), as it can be seen in the records of the monitoring station
*Corresponding author.
Figure 1.
Madrid city metropolitan area and air quality monitoring stations. Air
quality monitoring stations of the Madrid City council and Madrid
Greater Area networks are represented by circles and squares respec-
tively. Traffic stations are represented in red, urban background in blue
and industrial stations in green.
(Figure 2). This phenomenon is basically attributed to heavy
traffic levels and to a strong dieselization of the fleet in recent
years (Kassomenos, P. et al., 2006).
Modeling System and Domains
The mesoscale modeling system is based on the Weather Re-
search and Forecasting (WRF) (Skamarock & Klemp, 2008),
the Sparse Matrix Operator Kernel Emissions (SMOKE) mod-
eling system (Institute for the Environment, 2009), and the
Community Multiscale Air Quality (CMAQ) (Byun and Ching,
1999; Byun and Schere, 2006). Details about specific configu-
ration and adaptation to the Spanish conditions can be found
respectively in Borge et al. (2008a, 2008b, 2010).
Four nested domains were used in order to capture interna-
tional, national, regional and local contributions to NO2 to am-
bient concentration in Madrid with a maximum resolution of 1
km2 (Table 1).
Special care was taken to keep the consistency among the
emission inventories used across the scales. This is a crucial
issue for the source apportionment exercise since it includes an
analysis of emission contributions from different geographic
areas. Further information about emission compilation and
harmonization can be found in Borge, R. et al. (2014). The
emission inventory for the innermost domain (Figure 1) is
based on a combination of top-down and bottom-up methods
specific for each of the main emission sources in the city. The
most relevant feature is the integration of a the City Concil’s
mesoscale road traffic model within the emission model so very
detailed, 1-h, 1-km2 resolved emissions can be computed for
Table 1.
Domains for the mesoscale modelling system.
Do ma i n Geographic scope X-Y dimensions
(km) Horizontal
resolution (km)
D1 Europe 6144 × 5376 48
D2 Iberian Peninsula 1200 × 960 16
D3 Greater Madrid Region 192 × 192 4
D4 Madrid Metropolitan Area 40 × 44 1
020406080100 120 140
Annual mean NO
concentration (µg/m
19th highest hourly NO
concentration (µg/m
Traffic stations
Urban background stations
Suburban background stations
Figure 2.
Observed NO2 values (corresponding to the annual and hourly NO2
limit values defined in the European AQ Directive) in the Madrid air
quality monitoring network for the years 2010-2012.
any given traffic situation (Borge, R. et al., 2012a). Total emis-
sions at SNAP group level in the innermost modeling domain
(Figure 1) are given in Table 2.
Source Apportionment Methodology
A zero-out methodology was followed in this application.
The contribution of a particular emission source or region can
be estimated through the brute force method (BFM), sometimes
referred to as single-perturbation method (Samaali, M. et al.,
2011 and references within). This method relies on the analysis
of the change in the pollutant concentration that would occur if
a given emitting source is removed from the simulation (usually
referred to as zero-out sensitivity runs). This approach has been
used in the past to isolate the response of complex, nonlinear
systems to one particular sector in source apportionment and
sensitivity analysis (Cohan, D.S. et al., 2005). This method has
inherent limitations in accurately describing sensitivities but it
may be useful to approximate the effect of potential emission
reductions in a particular source or origin area as pointed out
before in several studies (Koo, B. et al., 2009; Carmichael, G.
et al., 2010; Leung et al, 2007). This analysis was originally
performed separately for those locations inside and outside the
Madrid municipality, since this is a legal requirement. For this
study, both analyses have been merged so a more scientifically-
sound and coherent view can be given. Reductions of 100%
(zero-out) were simulated for the most relevant anthropogenic
Residential, commercial and institutional combustion (RCI)
Table 2.
Emissions in the Madrid metropolitan area (metric tons per year).
group* CO NH3 NOX PM10 PM2.5 SO2 VOC
01 225 0 243 50 29 1128 1
02 10004 0 3680 520 410 2731 1104
03 2238 0 10689 265 210 2494 1217
04 1083 130 108 51 32 70 3782
05 0 15 0 0 0 0 2056
06 0 212 0 0 0 0 48828
07 22070 250 27961 1506 1205 157 4365
08 2711 0 4171 360 360 287 7 69
09 441 2036 1769 26 26 6 5267
10 357 1543 56 90 13 0 17
11 32 605 125 0 0 0 4682
latoT 39161 4791 48802 2868 2285 6873 72088
*SNAP groups: 01Combustion in energy and transformation industries;
02Non-industrial combustion plants; 03Combustion in manufacturing
industry; 04Production processes; 05Extraction and distribution of fossil
fuels; 06Solvent and other product use; 07Road transport; 08Other
mobile sources and machinery; 09Waste treatment and disposal; 10Agri-
culture; 11Other sources and sinks (Nature).
(SNAP 02);
Industry (SNAP 03 & 04);
Road traffic (SNAP 07);
Other mobile sources (planes, trains, machinery) (SNAP
Sources outside the modeling domain (rest of the country
and international contributions).
The total impact and therefore, the maximum theoretical
benefits that can be harvested by implementing abatement op-
tions in these sectors were derived from the comparison of the
assessment of the individual runs with the base case (consider-
ing all emissions). All the emission processing and scenarios
for the sensitivity runs were done through the SMOKE system.
Further details on this procedure can be found in R. Borge et al.,
Results and Discussion
Individual results for the model grid cells where air quality
monitoring traffic stations are located are shown in Figure 3.
As expected, according to their classification, NO2 concentra-
tion levels at all these locations are clearly dominated by con-
tributions form road traffic, with shares ranging form 59.8% to
75.3% and an average value of 69.4%. It is worth noting, how-
ever that this contribution seems to be smaller for the stations
located at the right-hand side of the graph. Those stations
(COSL, ALCO, LEGA, GETA) correspond to monitoring sta-
tions of the Greater Madrid Region, that they are located in the
outskirts of the urban area. In contrast, the influence of sources
outside the Madrid Region (rest of the country) is larger, which
is totally reasonable. While national contributions inside the
Madrid municipality are around 10% of total NO2 concentration
levels, they reach a 15% as an average in the monitoring sta-
tions closer to the modeling boundary. A similar trend is seen
for international contributions, although the differences are
smaller (3.4% Vs 5.6%). In all cases local emissions (those
originated inside the region) are responsible for the majority of
the observed concentrations (around 85% as an average). The
second source in importance is the RCI sector, responsible for
4.5% of NO2 ambient concentration. This ratio exhibits a larger
variability, with larger values in the city center (RCAJ, CUCA)
except for some locations where other sectors (mainly SNAP
09) have a bigger influence locally due to the proximity to
waste management plants. The contributions from other mobile
sources are similar in all the traffic stations (around 4.2%). This
is probably due to the difficulty to allocate emissions from gar-
dening and cleaning machinery as well as construction and
industrial vehicles, that are spatially distributed all over the city
depending on the population and road network density. Finally,
the results point out that the influence of industry in the air
quality of the city center is negligible; (less than 2%). This
makes sense, since the industrial activity in Madrid (very mod-
erate) is located in the surroundings of the city. This can be
clearly spotted in the results corresponding to the stations of the
Greater Madrid Region, mainly Coslada (COSL) where 9.4% of
NO2 is produced by industry. It should be noted that the SNAP
01 sector is totally unimportant in Madrid since there are no
power plants in the region.
Results form non-traffic stations are shown in Figure 4. De-
spite most of these stations are considered urban background
locations, the structure of NO2 contributions is rather similar to
that shown in Figure 3. Contributions of road traffic emissions
to air quality background levels roughly represent 2/3 of total
pollution. As expected, industry has a bigger influence in these
environments not directly exposed to traffic emissions. This is
particularly true for those stations of the Greater Madrid Region
(again those in the right-hand side in Figure 4) where industrial
activity clearly constitutes the second most important pollution
source (16.3% as an average). Although Alcobendas (ALCB) is
the only station label as industrial, it does not present a special-
ly high contribution from this sector (14.7%), lower that other
stations such as Móstoles (MOST) and Torrejón (TOAZ)
(15.1% and 21.2% respectively). The contribution of other
sectors not explicitly considered in the analysis represents
around 5.5% of total NO2 values as an average, although it
presents quite some variability. Barajas Pueblo presents an
unusually high share for this sector (16.0%). This may be due
to the influence of emissions generated in the Madrid/Barajas
international airport neither included in the SNAP 08 nor the
SNAP 07 sector.
A comprehensive source apportionment study was performed
in the Madrid metropolitan area through the application of a
multi-scale, multi-pollutant air quality modeling system (WRF-
SMOKE-CMAQ). The study was based in the single-perturba-
tion method, i.e. removing emissions from each of the main
sectors one at a time and comparing the results with those of the
baseline scenario (complete emission inventory).
The results from a series of locations throughout the inner-
most modeling domain indicate the different relative weight of
the contributions of the sectors analyzed. Despite these differ-
ences, it can be seen that road traffic is the main contributor to
Other sectors
Other mobile sources
Road traffic
Figure 3.
Results of the NO2 source apportionment analysis performed at the location of traffic air quality monitor-
ing stations.
Other sectors
Other mobile sources
Road traffic
Figure 4.
Results of the NO2 source apportionment analysis performed at the location of (mostly) urban background
monitoring stations.
NO2 levels all over the region, specially in the city center with
contributions up to 75% of total concentration (86% if non-
local sources, those outside the Greater Madrid Region are not
considered). Contributions form other countries are relatively
unimportant (around 1% in most cases). Similarly, pollution
from outside the region has a small incidence (approximately
3% - 4%). National influences (sources in Spain but outside the
Greater Madrid Region) are more important, mainly in the city
outskirts, but it represents less than 10% of total NO2 as an
average) This means that NO2 levels observed in Madrid are
basically due to local sources, so air quality issues can be
tackled by local AQP. In particular, these plans should be clear-
ly directed to reduce emissions in the road traffic sector, re-
sponsible of 69.4% and 66.4% of nitrogen dioxide ambient
concentration observed in traffic and urban background loca-
tions respectively (as an average). These figures indicate that
abatement emissions in the road traffic sector may be particu-
larly effective. As shown in Table 2, the relative contribution
of the SNAP 07 sector is 57.3% in the geographical domain
analyzed (D4).
While their contribution in terms of emissions is well below
60%, their share in resulting NO2 ambient concentration levels
is close to 70%. This may be related to two factors: 1) emis-
sions form road traffic are released at ground level (unlike those
of the RCI or industrial sectors) and 2) their NO2/NOX emission
factor is higher due to a strong fleet dieselization phenomenon
(R. Borge et al., 2012).
This conclusion is clearly reflected in the Madrid AQP for
the horizon 2011-2015 (Madrid City Council, 2012). Consis-
tently with the results of the source apportionment exercise
performed, most of the measures in the plan (up to 70) were
targeted to the road traffic sector. This abatement strategy in-
cludes a Low Emission Zone (LEZ), reduction of road capacity
and pedestrianized areas in the city center, renovation of city
bus fleet to incorporate clean technologies (electric, hybrid
natural gas-fuelled buses), etc. According to the results shown
in this AQP would achieve a 40% reduction of NOX emissions
from the road traffic sector in the modeling domain. As a result
of all these measures, a global decrease of 31% in NOX emis-
sions is expected in the year 2014 within D4 (respective to the
emissions of 2007 used as reference scenario for the source
apportionment analysis presented here). Additional model runs
(R. Borge et al., (2014) indicate that annual NO2 levels may be
reduced by 34% as an average; approximately 15 μg/m3 in the
city centre, also with an important impact in the metropolitan
area (−7 μg/m3 as an average in the modeling domain). 1-hour
concentration peaks may also decline by 40% approximately in
most of the city (Figure 5).
The second most important contributor, far from road traffic,
is the RCI sector, responsible for approximately 5% of ob-
Figure 5.
Expected effect of the Madrid AQP in NO2 concentration. Annual average for year 2007, considered for
the source apportionment study (a) and 2014, temporal horizon of the AQP (b).
served NO2 in Madrid. A similar influence can be attributed to
mobile sources other than road traffic. The contribution of in-
dustry is low in general although it may be important in the
surroundings of the city. According to the source apportion-
ment study performed, the industrial sources in the region are
responsible of 5% and 16% of NO2 levels in the traffic and
urban background stations of the Greater Madrid Region, and
therefore may constitute a sensible target for additional meas-
ures, especially in the east area of the Madrid metropolitan area.
As for the air quality monitoring strategy, the results form
this study indicate that the relative amount of traffic stations,
although high, may be adequate since traffic is mainly respon-
sible for air quality problems in the region. However, further
analysis should be done to understand to what extend this sta-
tions may be providing redundant results (the source appor-
tionment results are very similar in all the stations) and there-
fore the air quality monitoring network may be simplified and
reduced, considering the minimum requirements established in
the Directive 2008/50/EC. On the other hand, the potential need
for some industrial stations in the eastern part of the Madrid
metropolitan area may be considered.
Acknowledgemen ts
The Madrid city Council provided the traffic model and
supported this study. The CMAQ modeling system was made
available by the US EPA and it is supported by the Community
Modeling and Analysis System (CMAS) Center. The authors
also acknowledge the use of emission datasets and monitoring
data from the Spanish and Portuguese Ministries of Environ-
ment as well as air quality monitoring data from the Madrid
city Council and Greater Madrid Region.
Borge, R., Alexandrov, V., del Vas, J.J., Lumbreras, J., & Rodríguez, E.
(2008a). A comprehensive sensitivity analysis of the WRF model for
air quality applications over the Iberian Peninsula. Atmospheric
Environment, 42, 8560-8574.
Borge, R., de Miguel, I., de la Paz, D., Lumbreras, J., Perez, J., &
Rodríguez, M. E. (2012a). Comparison of road traffic emission mod-
els in Madrid (Spain). Atmospheric Environment, 62, 461-471.
Borge, R., López, J., Lumbreras, J., Narros, A., & Rodríguez, E. (2010).
Influence of boundary conditions on CMAQ simulations over the
Iberian Peninsula. Atmospheric Environment, 44, 2681-2695.
Borge, R., Lumbreras, J., & Rodríguez, M. E. (2008b). Development of
a high-resolution emission inventory for Spain using the SMOKE
modelling system: A case study for the years 2000 and 2010. Envi-
ronmental Modelling & Software, 23, 1026-1044.
Borge, R., Lumbreras, J., Pérez, J., de la Paz, D., Vedrenne, M., de
Andrés, J. M., & Rodríguez, M. E. (2014). Emission inventories and
modeling requirements for the development of air quality plans. Ap-
plication to Madrid (Spain). Science of the Total Environment, 466-
467, 809-819.
Borge, R., Lumbreras, J., Pérez, J., de la Paz, D., Vedrenne, M., &
Rodríguez, M. E. (2012b). Emission inventories and modeling activi-
ties for the development of air quality plans in Madrid (Spain). 2012
USEPA International Emission Inventory Conference “Emission In-
ventories—Meeting the Challenges Posed by Emerging Global, Na-
tional, Regional and Local Air Quality Issues”, Tampa, 13-16 Au-
gust 2012.
Byun, D. W., & Ching, J. K. S. (1999). Science algorithms of the EPA
models-3 community multi-scale air quality (CMAQ) modeling sys-
tem. EPA/600/R-99/030, US EPA National Exposure Research
Laboratory, North Carolina.
Byun, D. W., & Schere, K. L. (2006). Review of the governing equations,
computational algorithms, and other components of the models-3
community multiscale air quality (CMAQ) modeling system. Applied
Mechanics Reviews, 59, 51-77.
Carmichael, G., & Wild, O. (2010) Global and regional modelling. In:
Hemispheric Transport of Air Pollution. Task Force on Hemispheric
Transport of Air Pollution.
Cohan, D. S., Hakami, A., Hu, Y., & Russell, A. G. (2005). Nonlinear
response of ozone to emissions: Source apportionment and sensitivi-
ty analysis. Environmental Science & Technology, 39, 6739-6748.
Denby, B., Douros, I., & Fragkou, L. (2011). Modelling of nitrogen
dioxide (NO2) for air quality assessment and planning relevant to the
european air quality directive. Forum for Air Quality Modelling in
Europe (FAIRMODE) Guidance Document.
European Environment Agency (EEA) (2013). Air quality in Europe
2013 report. EEA Report No 9/2013.
Guerreiro, C., Horálek, J., de Leeuw, F., Hak, C., Nagl, C., Kurfürst, P.,
& Ostatnicka, J. (2010). Status and trends of NO2 ambient concentra-
tions in Europe. ETC/ACC Technical Paper 19/2010.
Institute for the Environment. SMOKE v2.7 User’s Manual (2009).
University of North Carolina, Chapel Hill.
Kassomenos, P., Karakitsios, S., & Papaloukas, C. (2006). Estimation
of daily traffic emissions in a South-European urban agglomeration
during a workday. Evaluation of several “what if” scenarios. Science
of the Total Environment, 2-3, 480-490.
Koo, B., Wilson, G. M., Morris, R. E., Dunker, A. M., & Yarwood, G.
(2009). Comparison of source apportionment and sensitivity analysis
in a particulate matter air quality model. Environmental Science &
Technology, 43, 6669-6675.
Leung, F. Y. T., Logan, J. A., Park, R., Hyer, E., Kasischke, E., Streets,
D., & Yurganov, L. (2007). Impacts of enhanced biomass burning in
the boreal forests in 1998 on tropospheric chemistry and the sensitiv-
ity of model results to the injection height of emissions. Journal of
Geophysical Research, 112, D10313.
Madrid City Council (2012). Madrid’s air quality plan 2011-2015.
Approved by the Governing Board of Madrid City Council on April
26, 2012. General Directorate of Sustainability, Government Divi-
sion of Environment, Safety and Mobility, Madrid City Council.
Samaali, M., Bouchet, V. S., Moran, M. D., & Sassi, M. (2011). Appli-
cation of a tagged-species method to source apportionment of pri-
mary PM2.5 components in a regional air quality model. Atmos-
pheric Environment, 45, 3835-3847.
Skamarock, W. C., & Klemp, J. B. (2008). A time-split nonhydrostatic
atmospheric model. Journal of Computational Physics, 227, 3465-