Air Quality, Atmosphere & Health (2018) 11:993–1008

https://doi.org/10.1007/s11869-018-0604-2

Ozone and volatile organic compounds in the metropolitan area

of Lima-Callao, Peru
José S. Silva 1 & Jhojan P. Rojas 1 & Magdalena Norabuena 1 & Rodrigo J. Seguel 2

Received: 9 May 2018 / Accepted: 12 July 2018 / Published online: 26 July 2018
# Springer Nature B.V. 2018

Abstract
This study analyzes ozone formation in the metropolitan area of Lima-Callao as a function of meteorological patterns and the
concentrations of nitrogen oxides and reactive organic gases. The study area is located on the west coast of South America (12°S)
in an upwelling region that is markedly affected by the Southeast Pacific anticyclone. The vertical stability and diurnal evolution
of the mixing layer were analyzed from radiosondes launched daily during 1992–2014 and from two intensive campaigns in
2009. Vertical profiles show that during June–November, the subsidence inversion base ranges from 0.6 to 0.9 km above sea level
(asl). In contrast, during December–May, subsidence inversion dissipates, leading to weak surface inversions from 0.1 to 0.6 km
asl. At the surface level, compliance with the ozone standard of 51 parts per billion by volume (ppbv) is explained by the marine
boundary layer effect and by strong inhibition of ozone formation due to titration with nitric oxide. Day-of-the-week variations in
ozone and nitrogen oxides suggest a VOC-limited ozone-formation regime in the atmosphere of Lima. Furthermore, the pattern
of C6–C12 species indicates that gasoline-powered vehicles are the main source of volatile organic compounds (VOCs), whereas
the species with the greatest ozone-forming potential corresponded to the sum of the isomers m- and p-xylene. Mean benzene
concentrations exceeded the standard of 0.63 ppbv, reaching 1.2 ppbv east of Lima. Nevertheless, the cancer risk associated with
the inhalation of benzene was deemed acceptable, according to USEPA and WHO criteria.

Keywords Tropospheric ozone . Volatile organic compounds . Benzene . Marine boundary layer . Radiosonde

Introduction NO2), which are known as precursors (Fiore et al. 2015;

Alvim et al. 2017; Toro et al. 2013, 2014). Therefore, environ-
Tropospheric ozone (O3) and volatile organic compounds mental agencies commonly assess VOCs in ozone non-
(VOCs) are key components in atmospheric chemistry. At the attainment areas. The US Environmental Protection Agency
surface level, ozone is considered one of the main anthropogenic (USEPA) updated the list of 54 target VOCs measured at pho-
atmospheric pollutants (Schultz et al. 2017). Furthermore, ozone tochemical assessment monitoring stations (known as 54
is a key greenhouse gas due to its strong absorption in bands of PAMS) and currently recommends monitoring 25 organic spe-
terrestrial infrared radiation (Monks et al. 2015). cies that are mainly prioritized according to their contribution to
O3 is a secondary gas formed through photochemical oxi- ozone formation (USEPA 2017). Other criteria include the role
dation of VOCs in the presence of nitrogen oxides (NO + played as secondary organic aerosol precursors, and whether
the compounds are considered hazardous air pollutants.
In urban areas, large proportions of organic precursors de-
rive from mobile sources (Fujita et al. 2012; Seguel et al.
* Rodrigo J. Seguel 2012; Rappengluck et al. 2005). Biogenic VOC emissions
[email protected]
derived from vegetation may also contribute to the inventory
José S. Silva of O3 precursors (Chameides et al. 1988). Apart from photo-
[email protected]
chemical ozone formation, stratosphere–troposphere ex-
1
National Meteorology and Hydrology Service, Jr. Cahuide 785, Jesús
change favors irreversible ozone transport from the strato-
María, Lima, Peru sphere and is thus another source of ozone, particularly for
2
Environmental Department, Trade and International Advisory
the upper and middle troposphere (Seguel et al. 2018).
SAGU, Los Canteros 8666, Parque industrial de La Reina, Epidemiological studies have shown significant associa-
7880340 Santiago, Chile tions between contamination by O3 and damage to human
994 Air Qual Atmos Health (2018) 11:993–1008

health (Bell et al. 2004; Liu and Peng 2018). The most impor- Methodology
tant effects are decreased lung function, respiratory tract in-
flammation, pharyngeal irritation, functional changes, and Study area
asthma symptoms in susceptible people (Lippmann 1991).
Forouzanfar et al. (2015) estimated the number of deaths at- The MALC is located on the west coast of South America
tributable to ozone exposure in 2013 at 217,000, which is 63% (12.0°S, 77.0°W, 2672 km2), has a population of approximate-
higher than in 1990. ly ten million, and spans elevations from sea level to the west-
Due to the adverse health effects of ozone, many agen- ern slopes of the Andes mountain range (approximately
cies have established air quality standards following 1.5 km asl) (Fig. 1).
World Health Organization (WHO) guidelines, which set The MALC experiences scarce rainfall, high atmospheric
a value of 51 ppbv for ozone. Accordingly, Peru has strin- humidity, and persistent cloud cover due to strong subsidence
gent standards of 51 ppbv for ozone and 0.63 ppbv for inversion caused by the Southeastern Pacific anticyclone that
benzene (MINAM 2017). The latter is not only a key prevails over the region for much of the year.
organic ozone precursor but is also a known carcinogen On the Peruvian coast, the predominant surface wind from
(ATSDR 2007). the south-southwest favors the upwelling of cold waters along
In Peru, in 2010, the first Automatic Air Quality the coast, which, together with subsidence inversion of the
Monitoring Network, the so-called RAMCA (for its acronym subtropical anticyclone, results in dense cloud cover, particu-
in Spanish), was implemented in the Metropolitan Area of larly from June to November (cold period). In turn, weakening
Lima-Callao (MALC), comprising 10 stations that record of the anticyclone during the warm period (December–May)
hourly concentrations of gases and atmospheric aerosols increases the number of clear days. These conditions make the
(Silva et al. 2017) over a vast geographical area with complex MALC one of the driest regions in the world. For example,
terrain. The network is currently operated by the National during 2010–2015, average temperature was 23 °C (range 13–
Service of Meteorology and Hydrology (SENAMHI) under 32 °C), average relative humidity was 85% (range 70–95%),
the Ministry of the Environment (MINAM). and average wind speed was 4 m/s (range 1–5 m/s), predom-
Air quality in Lima is largely modulated by a persistent inantly from the south-southwest, with average annual rainfall
weather and climate pattern (Enfield 1981). Underhill (2015) of 10 mm.
determined that higher NO2 concentrations are found during
the warm period than during the cold period. In turn, CO
Radiosonde observations at the international airport
records and the prevailing wind show that polluted air masses
(1992–2014)
are transported to the northeast of the city (Tashiro and
Taniyama 2002).
In total, 3046 radiosondes were launched from the Jorge
Despite recent efforts, characterization of Lima’s pho-
Chávez International Airport by the Peruvian Corporation of
tochemical pollution is scarce in the scientific literature,
Commercial Airports and Aviation (CORPAC, for its acro-
and the lack of permanent VOC measurements is a key
nym in Spanish) on a daily basis at 7:00 LT (in this paper,
information gap that must be bridged to properly address
UTC-5 will be used as Local Time (LT)) to assess tempera-
their management and to determine the population’s ex-
ture, relative humidity, atmospheric pressure, and wind direc-
posure to hazardous organic species. Therefore, the ob-
tion and speed as a function of height up to approximately
jective of this study is to relate meteorological patterns to
150 hPa (15 km) (Table 1).
air quality in the Lima-Callao metropolitan area, and to
identify the atmospheric chemical processes that govern
the ozone formation regime in the city. To meet these Daily radiosonde campaigns in Las Palmas
objectives, atmospheric stability is first analyzed from
23 years of radiosondes that were launched between To monitor the daily evolution of the marine boundary layer,
1992 and 2014, followed by diurnal characterization of two radiosonde campaigns were conducted in 2009 by
the marine boundary layer during summer (February) and SENAMHI. Radiosondes were launched from the Las
winter (August) 2009. Subsequently, ozone behavior is Palmas meteorological station (Table 1) from February 16 to
analyzed as a function of precursors of NO x (NO + 28 during the warm period (33 launches) and from August 21
NO2) and reactive volatile organic compounds in periods to 30 during the cold period (31 launches). Three launches
and regions of interest. Finally, the levels of reactive or- were conducted each day, at 07:00, 13:00, and 19:00 LT,
ganic gases are contextualized, prioritized in terms of respectively.
their potential contribution to ozone formation, and com- Vertical profiles for meteorological variables (pressure,
pared with national standards and international temperature, and relative humidity) were recorded by Vaisala
guidelines. RS92-SGP radiosondes. Calibration of the pressure,
Air Qual Atmos Health (2018) 11:993–1008 995

Fig. 1 Map of the metropolitan

area of Lima-Callao. Red circles
indicate air quality monitoring
stations, black circles indicate the
study stations, and red squares
indicate radiosonde launch sites

temperature, and humidity sensors was performed with the temperatures at the base and top of the critical inversion layer,
Vaisala Ground Check Set GC25. respectively.

Analysis from sounding

Surface observations (O3, NO, NO2, and CO)
The potential temperature method (Theta method) was used to
estimate the marine boundary layer height. This method finds The hourly concentrations of O3, NOX, and CO were mea-
the lowest critical inversion that meets the Theta gradient and sured using Teledyne analyzers (models 400E, T200, and
lapse-rate criteria according to Heffter (1980). The two criteria T300, respectively; Teledyne Technologies, Inc., Thousand
are expressed as follows: Oaks, CA, USA). The analyzer operation includes zero and
span verifications, calibrations, detection of leaks, preventive
Δθ=ΔZ > 0:5 K=100 m‐1 ð1Þ maintenance, and documentation. Verifications are conducted
θt θb > 2K ð2Þ on a monthly or weekly basis. The data are transmitted by
telemetry to SENAMHI headquarters, where the data are val-
where Δθ / ΔΖ is the potential temperature lapse rate in the idated after correcting null entries, duplicates, and/or anoma-
inversion layer, and θ b and θ t refer to the potential lies. For further details, see Silva et al. (2017).
996 Air Qual Atmos Health (2018) 11:993–1008

2009 (07:00, 13:00,

Due to the predominant wind direction and the significant

February 16–28 and

contribution of mobile sources from downtown Lima, pollut-

2014 (07:00 LT)

August 21–30,

and 19:00 LT)

January 1992 to
ed air masses are transported to the northeast area of the city.
Therefore, the selected air quality monitoring stations were

October
ATE (Ate district), VMT (Villa María del Triunfo district),
Sounding

and PPD (Puente Piedra district), which are located 18 km

east, 16 km south, and 22 km north of the center of the
–

–
MALC, respectively (Fig. 1).
Descriptive statistical analyses of hourly measurements of
July 5, 20,21 and 23, 2017 (07:00–07:30;

July 5, 20,21 and 23, 2017 (07:00–07:30;

July 5, 20,21 and 23, 2017 (07:00–07:30;

O3, NOX, CO, and radiosonde data were performed using the
R open-source statistical programming language (R
Development Core Team, Vienna, Austria) and its Openair
12:00–12:30; 18:00–18:30 LT)

12:00–12:30; 18:00–18:30 LT)

12:00–12:30; 18:00–18:30 LT)

package (Carslaw 2013) under the open-source software

RStudio (RStudio Boston, MA, available from http://www.
rstudio.org/).

Analysis of volatile organic compounds (C6–C12)

During 2015 and 2017, two VOC C6–C12 campaigns were

performed to complement the information collected from the
air quality monitoring network (Table 1). The sampling sites
were the same stations described previously (ATE, VMT, and
–

PPD). The first exploratory campaign was performed three

times per day. The second campaign was conducted on con-
secutive days and three times per day. The diurnal periods
were selected to measure the levels associated with rush hours
September 3, 2015 (07:00–10:00;

September 2, 2015 (07:00–10:00;

12:00–15:00; 18:00–21:00 LT)

12:00–15:00; 18:00–21:00 LT)

12:00–15:00; 18:00–21:00 LT)

August 20, 2015 (07:00–10:00;

(7:00 and 18:00 LT) and favorable convective mixing condi-

tions (13:00 LT).
Air was pulled by constant airflow (0.1 L/min) and passed
through cartridges packed with Tenax TA., Carboxen 1000,
and Carbosieve SII. The C6–C12 species adsorbed on the car-
tridges were immediately analyzed in a gas chromatograph
VOC campaign

with a mass spectrometry detector (GC/MS) using the TO-1

method of the US Environmental Protection Agency (USEPA
Location and measurement periods for VOCs and radiosondes

1999).
The gas chromatograph mass spectrometer (Shimadzu,
–

model GCMS-QP2010 ULTRA) includes a thermal desorp-

Altitude

tion unit that volatilizes and pre-concentrates VOCs before

(m asl)

358

292

205

13

73

injection into the GC. The system was calibrated using a mix-
ture that contained C6–C12 compounds (carbon tetrachloride;
Station location

− 76.9186

− 76.9200

− 77.0700

− 76.9997
− 77.1109

benzene; toluene; ethylbenzene; o-xylene; m-xylene; p-

− 12.0261;

− 12.1664;

International airport Jorge − 12.0242;

− 12.1509;
− 11.8500;

xylene; styrene; n-propylbenzene; naphthalene; p-

(Lat; Lon)

isopropyltoluene, purchased from SUPELCO). The estimated

overall uncertainty of VOC measurements was approximately
10%.
Villa María del Triunfo

Determination of priority volatile organic compounds

Puente Piedra (PPD)

(C6–C12)
Las Palmas (LPA)
Chavez (JCH)
Monitoring station

The measured mixing ratio for each C6–C12 species was mul-
Ate (ATE)

(VMT)

tiplied by the maximum incremental reactivity (MIR) of the

Table 1

compounds (Carter 2010) to obtain their average reactivity-

adjusted mixing ratio. Thus, each C6 –C 12 species was
Air Qual Atmos Health (2018) 11:993–1008 997

categorized based on its proportional contribution to ozone displacement to higher latitudes produces clear skies over
formation according to the following formula: Lima that are occasionally interrupted by the passage of
troughs.
Proportional weighting of the compound The vertical temperature profiles obtained during
¼ ðC‐minÞ  ðmax‐minÞ⋅100 ð3Þ December–May of 1992–2014 showed that radiative heating
at surface level allows for the formation of weak surface in-
where C is the average MIR-adjusted mixing ratio for the versions. The bases of the temperature inversions occurred
compound, and min (max) are the lowest (highest) average between 0.1 and 0.6 km (max median 0.59 in December;
MIR-adjusted mixing ratios, respectively, for all compounds. min median 0.13 in March) (Fig. 2a), and the temperature
increased between 1 °C/100 m in May and 0.7 °C/100 m in
January (Fig. 2b).

Results and discussion

Daily evolution of the marine boundary layer
Weather patterns
The strength and persistence of the subsidence inversion sig-
nificantly determine the meteorological pattern over Lima.
A 23-year climatology of vertical stability based on radio-
Accordingly, radiosonde campaigns for the two distinct pe-
sondes launched from the Jorge Chavez International
riods (cold and warm) were designed to evaluate the daily
Airport at 7:00 LT is presented. The analysis identified two
marine boundary layer growth in more detail through vertical
markedly different periods: June–November and December–
temperature profiles, potential temperature, water vapor,
May.
mixing ratios, wind speed, wind vectors, and cloud layers
The vertical temperature profiles obtained between June
(Fig. 3).
and November during 1992–2014 showed that the bases of
the inversions frequently occurred between 0.6 and 0.9 km
(max median 0.82 in June; min median 0.67 in September) Cold period
(Fig. 2a), and that temperature increased between 2 °C/100 m
in August and 1.2 °C/100 m in November (Fig. 2b). The lapse rate shows that, from August 23 to 29, the atmo-
During this cold period (June–November), the strengthen- sphere experienced a significant temperature increase of ap-
ing and displacement of the subtropical anticyclone toward proximately 9 °C from a height of 0.6 to 2.0 km (Fig. 3a) as a
the central coast of South America intensifies the interaction result of adiabatic heating due to subsidence. Figure 3c shows
between the Humboldt Current and the descending air masses. the potential temperature gradient used to calculate the marine
As a result, a thick layer of cold and humid air is generated in boundary layer height according to the method of Heffter
the lower troposphere. The above processes, coupled with (1980). The potential temperature isotherm of 295 K marked
radiational cooling from the top of the stratus cloud deck, the height of the marine boundary layer.
leads to the development of a strong subsidence inversion During the period from August 21 to 30, 2009, the subsi-
(Albrecht 1981). dence inversion base reached a maximum height of 1.1 km asl
On the other hand, during the warm period (December– on August 21 (13:00 LT) and a minimum height of 0.46 km
May), the weakening of the subtropical anticyclone and asl on August 26 (13:00 LT). The average subsidence

Fig. 2 Variation of the first

temperature inversion base (a)
and vertical temperature gradient
(b) determined from radiosondes
launched at the international
airport by CORPAC during the
period 1992–2014 at 07:00 LT.
On each box, the central mark
indicates the median, and the
bottom and top edges of the box
indicate the 25th and 75th
percentiles, respectively. The
whiskers extend to the minimum
and maximum data points
998 Air Qual Atmos Health (2018) 11:993–1008

Fig. 3 Vertical profiles of temperature, potential temperature, water and February 16–28 (left panel) 2009. The cloud scale range considers
vapor, wind speed, and cloud layers measured using radiosondes clear sky (Clear), 1–2 oktas (FEW), 2–4 oktas (SCT: scattered), 5–7 oktas
launched from Las Palmas (Lima) during August 21–30 (right panel) (BKN: broken), and 8 oktas (OVC: overcast)
Air Qual Atmos Health (2018) 11:993–1008 999

inversion height at 7:00, 13:00, and 19:00 LT showed low daily maximum 8-h average (MDA8) ozone-mixing ratio does
variability and was located at 798 m (standard deviation not exceed 30.1 ppbv (PPD). Similarly, annual maxima do not
(SD) = 117), 796 m (SD = 207), and 677 m (SD = 154) asl, exceed 39 ppbv. Therefore, mixing ratios of surface O3 in the
respectively. Under these meteorological conditions, no sur- MALC comply with the national air quality standard of
face inversions were identified during the period August 21– 51 ppbv.
30, 2009. During the warm period, the mean diurnal cycle at station
Figure 3e shows the high water-vapor content of the marine ATE showed maximum nitric oxide concentration of 97 ppbv
boundary layer, whereas the observed wind speeds were low, at 7:00 LT, compared with 56 ppbv at 8:00 LT during the cold
generally less than approximately 4 m/s (Fig. 3g). season (Fig. 5). This pattern of higher concentrations during
The period from August 21 to 30, 2009 was mostly char- the warm period was also observed at VMT (south of the city)
acterized by cloudiness located below the subsidence inver- and at PPD (north of the city).
sion base. In general, a cloud layer with average thickness of NO and CO mixing ratios as a function of wind speed and
0.4 km was identified (Fig. 3i). Clear skies were observed only direction are presented in polar coordinates in Fig. 6. The
on August 26 and partly during the previous day. polar plot shows that in the warm period of 2017, the mixing
ratios of the primary gases NO and CO mostly overlap regard-
Warm period ing wind direction and speed, as is expected for pollutants
from the same source. Conversely, in the cold period, NO
Analysis of the marine boundary layer from February 16 to decreases and CO becomes concentrated (Fig. 5). Fine parti-
28, 2009 shows that the subsidence inversion was completely cles (PM 2.5) exhibited similar behavior than CO in ATE. The
dissipated. By contrast, surface inversions were developed. drastic decreases in nitric oxide compared to carbon monoxide
Over this period, sunny days were occasionally interrupted from June to November (cold period) can be explained by
by cloudiness, mainly during morning hours (7:00 LT). differences in the lifetimes of these species under conditions
The cross-section of temperature (Fig. 3b) shows that days of low atmospheric dispersion. Carbon monoxide has an av-
with clouds (February 16, 21, 22, 25, 27, and 28) inhibit the erage chemical lifetime of approximately 2 months in the tro-
formation of surface inversions. Conversely, clear days show posphere while for NOx the lifetime is about 1 day (Yin et al.
the development of weak surface inversions. The most notable 2015; Wenig et al. 2003). Hence, it is hypothesized that sub-
period was from February 17 to 19, which showed a temper- sidence inversion favor the accumulation of carbon monoxide
ature inversion of approximately 3 °C with base ranging from throughout the cold period and the oxidation of NOx to gas-
0.1 to 0.5 km asl. eous HNO3, followed by its uptake into condensed phase
Throughout the day, in general, the increase in surface would explain the significant drop of NOx. As described in
heating increases the marine boundary layer height, thus im- preceding sections, relative humidities are always high during
proving vertical mixing conditions. During this period, the the cold season, and precipitation in the form of a very light
marine boundary layer grew throughout the day a few hundred drizzle falls much of the time. For example, the relative hu-
meters. The average base of the thermal inversion at 7:00, midity in ATE, VMT, and PPD varies between 81 and 88%,
13:00, and 19:00 LT was located at 227 m (SD = 93), 350 m between 88 and 95%, and between 77 and 85% respectively
(SD = 287), and 336 m (SD = 167) asl respectively. during the cold period of 2016.
Figure 3f shows maximum water vapor mixing ratio at the
surface level, mainly on days with high surface temperatures.
Those days show a significant increase in water vapor (> 15 g/ Ozone inhibition
kg), with higher values than those recorded in winter.
Furthermore, wind speed at the surface level is higher during The mean diurnal cycle (Fig. 4) shows that the NO mixing
summer than winter (Fig. 3j). ratio at ATE is always higher than 27.7 ppbv (14:00 LT),
thereby preventing ozone accumulation during the day
Overview of O3–NOx–CO through titration with NO (reaction 1). The maximum ozone
concentration of 19.3 ppbv reached at 13:00 LT is lower than
Lima shows high atmospheric concentrations of nitric oxide the NO concentration at the same time (28.8 ppbv), which
and carbon monoxide throughout the year, especially at the illustrates the strong inhibitory effect of NO.
ATE station located downwind of the downtown area (Fig. 4). NO þ O3 → NO2 þ O2 ðReactionð1ÞÞ
By contrast, ozone measurements exhibited low mixing ratios.
For example, the annual average between 2015 and 2017 at The polar plot (Fig. 6) shows O3 response to NO levels.
the PPD station varied from 5.7 to 6.9 ppbv, and the mean During the warm period, quadrants with higher ozone levels at
diurnal cycle during the period did not exceed 18.9 ppbv ATE, PPD, and VMT are clearly accompanied by decreases in
(13:00 LT). The 3-year average of the annual fourth-highest NO. Inversely, quadrants with low ozone levels have higher
1000 Air Qual Atmos Health (2018) 11:993–1008

Fig. 4 Daily and weekly

variations in NO2, NO, CO, and
O3 mixing ratios for the period
2015–2017 (O3 period: 2012–
2017) at stations ATE, VMT, and
PPD

NO mixing ratios. This behavior is illustrated by the low O3/ in NOx reduces consumption of the •OH radical (reaction 2).
NO ratio found in all sites especially in warm periods (Fig. 6). Consequently, the greater availability of •OH would favor
The weekly variation of the primary pollutants CO and VOC oxidation, leading to O3 formation. This result suggests
NOx (Fig. 4) shows that the main decrease affects nitric oxide, a VOC-limited ozone-formation regime in the atmosphere of
given the longer lifetime of CO. Average mixing ratio indi- Lima.
cates higher NO concentrations on workdays (Monday–

Thursday) than at the weekend (Sunday), differing by 29, 9, OH þ NO2 þ M→HNO3 þ M ððReactionð2ÞÞ
and 19% at stations ATE, VMT, and PPD, respectively.
Conversely, for ozone, the average mixing ratio is higher on The observed Bweekend effect^ highlights that future re-
the weekend (Sunday) than at the workdays (Monday– strictions on nitrogen oxides without properly managing
Thursday) differing by 16, 13, and 7%, at the ATE, VMT, VOCs could lead to increased O3 concentrations, as described
and PPD stations, respectively. Over the weekends, a decrease in the literature (Fujita et al. 2003).
Air Qual Atmos Health (2018) 11:993–1008 1001

Fig. 5 Variations in O3, NO2,

NO, and CO mixing ratios and
PM 2.5 concentration during the
warm (left panel) and cold (right
panel) periods of 2016 at stations
ATE, VMT, and PPD

In summary, meteorological conditions of Lima do not Case study: VOCs in urban atmosphere of Lima
promote photochemical activity during most of the year. The
cloud cover affects actinic flux, and the moderate tempera- The air quality network of the MALC fails to measure relevant
tures do not favor ozone formation. However, the highest organic precursors. Therefore, the campaigns conducted to
ozone levels were typically found during the warm period measure C6–C12 species are the first efforts to address this
mainly due to a presence of photochemical precursors and issue in the MALC and results of C6–C12 speciation is pre-
the occurrence of thermal inversions. Statistically significant sented to illustrate the impact of several VOCs on air quality
differences between cold and warm periods (2014-2016) were and health. Specifically, knowledge of atmospheric VOC
found for all sites as determined by ANOVA (α = 0.05). levels in the MALC is critical for (1) understanding the pro-
From a chemical perspective, in Lima, NO is reacting rap- cesses that govern ozone formation, (2) assessing their contri-
idly with O3 to generate NO2 (reaction 1). Furthermore, at low bution to the formation of secondary organic aerosols, and (3)
VOC/NOx ratios, OH reacts predominantly with NO2 (reac- determining exposure to intrinsically hazardous substances.
tion 2), removing radicals, retarding O3 formation, and pro-
viding (in theory) a major source of HNO3 in fogs and parti- Pattern of C6–C12 species
cles. Hence, the high degree of ozone inhibition caused by
titration with nitric oxide, and the scarcity the radical •OH The results of C6–C12 species show that the highest mixing
explain the lower ozone concentrations observed at the ATE, ratios are oriented eastward to the MALC center (ATE) in
PPD, and VMT stations. accordance with the predominant wind direction. Figure 7
1002 Air Qual Atmos Health (2018) 11:993–1008

Fig. 6 Polar plot for: NO (ppbv), O3 (ppbv), CO (ppmv), and O3/NO ratio, during the warm and cold periods of 2017 at stations ATE, VMT, and PPD
Air Qual Atmos Health (2018) 11:993–1008 1003

Fig. 7 C6–C12 pattern assessed at

ATE, PPD, and VMT from
exploratory campaigns performed
on July 5, 20, 21, and 23, 2017
(top panel); comparison between
the 2015 pattern and that assessed
from the campaign conducted in
July 2017 (bottom panel)

shows that the pattern of C6–C12 species analyzed at ATE, of ozone associated with m + p-xylene is 23, 20, and 17 times
PPD, and VMT is relatively similar and comparable to cam- the productivity of benzene at ATE, PPD, and VMT, respec-
paigns conducted in August 2015 and July 2017 (Fig. 7). tively. By contrast, the species naphthalene, styrene, p-
Consequently, the pattern of C6–C12 species confirms the con- isopropyltoluene, n-propylbenzene, n-butylbenzene, and
tribution of a main source to the inventory of VOCs, whereas isopropilbenzene add little value in all sites.
the speciation of C6–C12 compounds identifies tracers that are Figure 8 shows the cumulative percentage of each C6–C12
typical of vehicular emissions, e.g., benzene, toluene, and xy- species, based on reactivity (Eq. 3). Categorization shows that
lenes. The city of Lima, with approximately 2.2 million motor 5 of the 14 species account for nearly 90% of the ozone for-
vehicles (Silva et al. 2017), experiences significant traffic con- mation potential. These species are m + p-xylene (36%), tolu-
gestion and gas emissions from vehicles throughout the day, ene (21%), 1,2,4-trimethylbenzene (14%), 1,2,5-
due to the lack of a comprehensive public transport system trimethylbenzene (13%), and o-xylene (8%). Furthermore,
(Tashiro and Taniyama 2002). these five VOCs prevail at the three sampling sites, and also
match the priority compounds proposed by the USEPA.

Priority C6–C12 in Lima

BTEX levels
The importance of each compound for ozone formation was
determined using the mixing ratios found in Lima (Table 2) In 2014, the national ambient air quality standard for benzene
and the maximum incremental reactivity (MIR) scale (Carter was reduced from 1.3 to 0.63 ppbv annual average in Peru.
2010). As shown in Table 2, adjusting the mixing ratio ac- Currently, no systematic benzene measurements are per-
cording to reactivity makes it possible to identify substances formed to evaluate compliance or exceedance of the standard.
with greater potential to increase ozone formation, compared In this context, the campaigns reported herein determine the
with those that are less productive despite having higher at- current levels of and exposure to this carcinogenic pollutant.
mospheric concentrations. For example, benzene occurs at Elevated levels of reactive organic gases (BTEX: benzene,
higher concentrations than o-xylene, but the latter has greater toluene, ethylbenzene, and xylene) have been found in the
ozone-forming potential. At ATE, PPD, and VMT, the species MALC. Table 3 shows that mean and median benzene con-
with the highest ozone formation potential is the sum of the centrations exceeded the standard during the study period,
species m-xylene and p-xylene (expressed as a sum because with the maximum concentration (1.2 ppbv) recorded at sta-
these isomers cannot be separated by current analytical tion ATE located east of Lima. For context, Table 3 also pre-
methods). As shown in Table 2, the theoretical productivity sents some statistical parameters gathered from BTEX
1004 Air Qual Atmos Health (2018) 11:993–1008

Table 2 Mixing ratio of C6–C12 species at Lima multiplied by maximum incremental reactivity (MIR, Carter 2010)

C6–C12 Ate PP VMT MIR (g O3/g VOC) Reactivity-adjusted

average mixing ratio
Median Mean Max Median Mean Max Median Mean Max
ppbv Ate PP VMT

m + p-Xylene 1.4 2.1 7.4 0.98 1.2 4.7 0.97 1.2 2.3 9.75 20 11 12
Toluene 2.3 2.7 9.1 1.2 1.6 5.3 1.2 2.0 4.9 4.0 11 6.5 8.0
1,2,4-Trimethylbenzene 0.30 0.59 3.1 0.35 0.43 1.3 0.27 0.40 0.84 11.97 7.0 5.2 4.7
1,3,5-Trimethylbenzene 0.30 0.55 2.8 0.28 0.36 1.1 0.30 0.41 0.95 11.76 6.5 4.2 4.8
o-Xylene 0.35 0.60 2.5 0.30 0.36 1.2 0.47 0.38 0.64 7.64 4.6 2.7 2.9
Ethylbenzene 0.46 0.62 2.1 0.29 0.37 1.2 0.41 0.41 0.75 3.04 1.9 1.1 1.3
Benzene 0.81 1.2 4.8 0.77 0.80 1.8 1.0 0.97 1.3 0.72 0.9 0.6 0.7
Naphthalene 0.07 0.13 0.53 0.06 0.19 1.1 0.06 0.08 0.14 3.34 0.4 0.6 0.3
Styrene 0.22 0.24 0.53 0.13 0.16 0.50 0.13 0.39 1.5 1.73 0.4 0.3 0.7
p-Isopropyltoluene 0.05 0.06 0.09 0.05 0.05 0.07 0.05 0.04 0.06 4.44 0.3 0.2 0.2
n-Propylbenzene 0.10 0.13 0.36 0.05 0.07 0.17 0.08 0.08 0.16 2.03 0.3 0.1 0.2
n-Butylbenzene 0.05 0.09 0.19 0.02 0.04 0.07 – – – 2.36 0.2 0.1 0.0
Isopropylbenzene 0.06 0.06 0.09 0.04 0.04 0.05 – – – 2.52 0.1 0.1 0.0

measurements in Lima and compares them to the latest official As mentioned in the previous section, m-xylene and p-
values reported for Santiago de Chile (SCL), another South xylene are the species with the highest ozone formation po-
American capital city, which has been designated an ozone tential among the C6–C12 species studied in the MALC, sim-
non-attainment area over the last 20 years (Seguel et al. 2018). ilar to the SCL, considering the 54 PAMS. The levels of m +
The mean benzene values found in the MALC exceed p-xylene at ATE are similar to those found downwind of SCL.
those reported for downtown SCL and approximately 16 km The same was observed for mean o-xylene levels.
downwind. In Table 3, upwind SCL is a suburban site located Nevertheless, the ozone standard is not exceeded at ATE, in
26 km from the city center. Mean toluene concentrations are contrast to downwind SCL. This result is attributed to the
similar for both cities, and the maximum toluene concentra- prevailing meteorological conditions in Lima and to the chem-
tions were found in ATE and in downtown SCL. Mean ethyl- ical regime characterized by the ozone reaction with NO and
benzene concentrations are relatively similar for both cities, low availability of the radical •OH consumed by high levels of
except for higher mean concentrations in central SCL. NOx relative to VOCs.

Fig. 8 Cumulative sum of the

proportional weight of each
C6–C12 compound. The red
line represents the cut point of
90%
 Air Qual Atmos Health (2018) 11:993–1008

Table 3 Statistical parameters for BTEX at Lima and Santiago (ppbv)

Median Mean Max

ATE PP VMT Upwind Downtown Downwind ATE PP VMT Upwind Downtown Downwind ATE PP VMT Upwind Downtown Downwind
(SCL) (SCL) (SCL) (SCL) (SCL) (SCL) (SCL) (SCL) (SCL)

Benzene 0.81 0.77 1.0 0.27 0.54 0.40 1.2 0.80 0.97 0.34 0.61 0.49 4.8 1.8 1.3 0.79 1.1 1.2
Toluene 2.3 1.2 1.2 0.98 2.1 1.7 2.7 1.6 2.0 1.0 2.8 1.9 9.1 5.3 4.9 1.9 6.2 4.2
Ethylbenzene 0.46 0.29 0.41 0.18 0.81 0.39 0.62 0.37 0.41 0.19 0.94 0.52 2.1 1.2 0.75 0.28 1.9 1.8
m + p-Xylene 1.4 0.98 0.97 0.48 3.0 1.0 2.1 1.2 1.2 0.53 3.3 1.7 7.4 4.7 2.3 0.80 6.9 6.9
o-Xylene 0.35 0.30 0.47 0.19 0.95 0.40 0.60 0.36 0.38 0.21 1.1 0.60 2.5 1.2 0.64 0.32 2.3 2.2
1005
1006 Air Qual Atmos Health (2018) 11:993–1008

Table 4 Comparison of mean

values for aromatic hydrocarbon Sites Benzene Toluene Ethylbenzene m + p-Xylene o-Xylene
mixing ratios for various cities of ppbv
Latin America (Sao Paulo: Alvim
et al. 2017; Santiago: Seguel et al. Sao Paulo 0.81 ± 0.02 4.40 ± 0.18 1.15 ± 0.03 0.65 ± 0.03 / 1.40 ± 0.13 0.63 ± 0.01
2013; Mexico City: Jaimes- 2008
Palomera et al. 2016) Santiago 0.48 ± 0.13 1.9 ± 0.9 0.55 ± 0.38 1.9 ± 1.4 0.64 ± 0.45
2009
Mexico City 1.63 ± 0.42 8.99 ± 5.18 0.82 ± 0.45 1.42 ± 0.58 1.09 ± 0.58
2012
Lima 1.0 ± 0.2 2.1 ± 0.5 0.46 ± 0.13 1.5 ± 0.5 0.45 ± 0.14
2017

For Sao Paulo, separate results for m-xylene and p-xylene were available

Lastly, Table 4 compares BTEX levels with other two Latin Conclusions
America megacities (Sao Paulo and Mexico City). In general,
a good agreement for m + p-xylenes and o-xylene is observed. Twenty-three years of radiosonde observations identified a
Toluene stands out as one of the aromatics that presents higher persistent subsidence inversion over Lima, which was stron-
concentrations and greater variation among the four cities gest during cold period (June–November) with an average
reported. base altitude of 0.74 km. During the warm period
(December–May), the subsidence inversion was dissipated
entirely, and weak surface inversions developed with an aver-
Risk assessment for benzene age base altitude of 0.29 km.
The daily evolution of the marine boundary layer showed
Benzene is a known carcinogen (WHO 2000). Therefore, the no surface inversions during the period August 21–30, 2009
USEPA and WHO have established ranges and levels of risk and a cloud layer with an average thickness of 0.4 km was
associated with its inhalation of 2.2·10−6–7.8·10−6 and 6·10−6, usually identified below the subsidence inversion base.
respectively (ATSDR 2007). Benzene measurements at ATE, During February 16–28, 2009, the marine boundary layer
PPD, and VMT were used to estimate the excess lifetime risk grew along the day reaching the higher average altitude of
of leukemia. Table 5 shows the exposure concentration 0.35 (SD = 0.29) at 13:00 LT.
modeled for a residential scenario. This concentration was In the warm period, efficient ozone formation was hindered
calculated by multiplying the 95% upper confidence limit by the intermittent occurrence of numerous cloudy days.
(95% UCL) of the arithmetic mean by the factor 0.41, which Similarly, on clear days with greater solar radiation and with
corresponds to a residential scenario. Specifically, an average the presence of precursors, a substantial increase in ozone was
lifetime of 70 years, exposure duration of 30 years, and an not observed, which was again attributed to strong inhibitory
exposure frequency of 350 days/year were considered to de- effect resulting from the titration of nitric oxide that is present
termine this factor (Table 5). at high atmospheric concentrations in the Lima-Callao metro-
The results show that the cancer risk ranges from 2 per hun- politan area. Maintaining compliance with ozone standards
dred thousand to 9 per million inhabitants. The estimated risk over time, in a continuously growing city, will require effec-
associated with benzene inhalation is considered acceptable, ac- tive measures for the balanced reduction of both NOx and
cording to the criteria established by the USEPA and WHO. VOC concentrations.

Table 5 Exposure levels and risk assessment for benzene according to EPA and WHO inhalation unit risk

Site N° Min. Max. SD Coefficient of Mean Median 95% Student’s t UCL Residential Inhalation unit risk
variation exposure
a
(Factor: 0.41) USEPA range WHO

μg/m3 2.2·10−6 7.8·10−6 6.0·10−6

ATE 12 0.25 15 4.4 3.6 3.9 2.6 6.3 2.6 6 in 1·106 2 in 1·105 2 in 1·105
PPD 12 0.81 5.6 1.5 1.9 2.5 2.5 3.5 1.4 3 in 1·106 1 in 1·105 9 in 1·106
VMT 12 1.7 4.3 1.1 1.1 3.1 3.3 3.8 1.5 3 in 1·106 1 in 1·105 9 in 1·106
a
Factor = (350 × 30) / (70 × 365): average lifetime = 70 years; exposure duration = 30 years; exposure frequency = 350 days/year
Air Qual Atmos Health (2018) 11:993–1008 1007

The pattern of C6–C12 species was relatively similar in the and metabolic risks or clusters of risks in 188 countries, 1990-2013:
a systematic analysis for the global burden of disease study 2013.
atmosphere of the Lima-Callao metropolitan area and was
Lancet 386:2287–2323
mainly attributed to mobile sources. In the group of C6–C12 Fujita EM, Campbell DE, Stockwell WR, Lawson DR (2012) Past and
species, five were classified as priority species for their ozone- future ozone trends in California’s south coast air basin: reconcilia-
f o r m i ng p ot en t i a l : m + p - x yl e n e, to l u en e, 1 ,2 , 4- tion of ambient measurements with past and projected emission
inventories. J Air Waste Manage Assoc 63:54–69. https://doi.org/
trimethylbenzene, 1,2,5-trimethylbenzene, and o-xylene.
10.1080/10962247.2012.735211
Further, the observed BTEX levels were high and similar to Fujita EM, Stockwell WR, Campbell DE, Keislar RE, Lawson DR (2003)
those reported in non-attainment areas (benzene (1.0 ppbv), Evolution of the magnitude and spatial extent of the weekend ozone
toluene (2.1 ppbv), ethylbenzene (0.5 ppbv), m + p-xylene effect in California’s south coast air basin, 1981–2000. J Air Waste
Manage Assoc 53:802–815. https://doi.org/10.1080/10473289.
(1.5 ppbv), and o-xylene (0.4 ppbv)).
2003.10466225
In terms of public health implications, the exploratory ben- Heffter JL (1980) Transport layer depth calculations. Second Joint
zene mixing ratios found in this study exceeded the standard Conference on Applications of Air Pollution Meteorology. New
of 0.61 ppbv. However, exposure to this carcinogenic pollut- Orleans, LA
Jaimes-Palomera M, Retama A, Elias-Castro G, Neria-Hernández A,
ant is considered acceptable, based on risk analysis conducted
Rivera-Hernández O, Velasco E (2016) Non-methane hydrocarbons
in accordance with USEPA and WHO guidelines. in the atmosphere of Mexico City: Results of the 2012 ozone-season
campaign. Atmos Environ 132:258–275. https://doi.org/10.1016/j.
atmosenv.2016.02.047
Acknowledgements Rodrigo Seguel acknowledges support from Lippmann M (1991) Health effects of tropospheric ozone. Environ Sci
CONICYT, FONDECYT Program, initiation into research 2013, Technol 25:1954–1962. https://doi.org/10.1021/es00024a001
Project No 11130177. Liu JC, Peng RD (2018) Health effect of mixtures of ozone, nitrogen
dioxide, and fine particulates in 85 US counties. Air Qual Atmos
Health 11:311–324. https://doi.org/10.1007/s11869-017-0544-2
MINAM (2017) Decreto supremo N° 003-2017-MINAM—Estandares
Funding This work has been funded by the SNIP project: expansion and Nacionales de calidad ambiental (ECA) para aire, Ministerio del
improvement of the monitoring network for air quality forecasting in Ambiente (MINAM). Republic of Peru. [WWW Document] doi:
Metropolitan Lima (N°199842). D.SNo 003–2017–MINAM
Monks PS et al (2015) Tropospheric ozone and its precursors from the
urban to the global scale from air quality to short-lived climate
References forcer. Atmos Chem Phys 15:8889–8973. https://doi.org/10.5194/
acp-15-8889-2015
Rappengluck B et al (2005) An urban photochemistry study in Santiago
Albrecht JA (1981) The twenty year average atmospheric structure at de Chile. Atmos Environ 39:2913–2931. https://doi.org/10.1016/j.
Lima, Peru. Dissertation. Dept. of Meteorology, Florida State atmosenv.2004.12.049
University, Tallahassee Schultz MG et al (2017) Tropospheric ozone assessment report: database
Alvim DS et al (2017) Main ozone-forming VOCs in the city of Sao and metrics data of global surface ozone observations. Elem Sci
Paulo: observations, modelling and impacts. Air Qual Atmos Anth 5:58. https://doi.org/10.1525/elementa.244
Health 10:421–435. https://doi.org/10.1007/s11869-016-0429-9 Seguel RJ, Mancilla CA, MAL G (2018) Stratospheric ozone intrusions
ATSDR (Agency for Toxic Substances and Disease Registry) (2007) during the passage of cold fronts over central Chile. Air Qual Atmos
Toxicological profile for benzene. Update. Public Health Service, Health:1–14 11:535–548. https://doi.org/10.1007/s11869-018-
U.S. Department of Health and Human Services, Atlanta, GA 0558-4
Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F (2004) Ozone Seguel RJ, Mancilla CA, Rondanelli R, Leiva MA, Morales RGE (2013)
and short-term mortality in 95 us urban communities, 1987-2000. Ozone distribution in the lower troposphere over complex terrain in
JAMA 292:2372–2378. https://doi.org/10.1001/jama.292.19.2372 Central Chile. J Geophys Res Atmos 118:2966–2980. https://doi.
Carslaw DC (2013) The Openair manual: open-source tools for analyzing org/10.1002/jgrd.50293
air pollution data. Manual for version 0.9. King’s College London. Seguel RJ, Morales SR, Leiva GM (2012) Ozone weekend effect in
London, UK. [WWW Document] URL https://goo.gl/iigCI9 Santiago, Chile. Environ Pollut 162:72–79. https://doi.org/10.
(accessed 5 January 2018) 1016/j.envpol.2011.10.019
Carter WPL (2010) Development of the SAPRC-07 chemical mecha- Silva J, Rojas J, Norabuena M, Molina C, Toro RA, Leiva-Guzmán MA
nism. Atmos Environ 44:5324–5335. https://doi.org/10.1016/j. (2017) Particulate matter levels in a south American megacity: the
atmosenv.2010.01.026 metropolitan area of Lima-Callao, Peru. Environ Monit Assess 189:
Chameides W, Lindsay R, Richardson J, Kiang C (1988) The role of 635. https://doi.org/10.1007/s10661-017-6327-2
biogenic hydrocarbons in urban photochemical smog: Atlanta as a Tashiro Y, Taniyama T (2002) Atmospheric NO2 and CO concentration
case study. Science 241:1473–1475. https://doi.org/10.1126/ in Lima, Peru. Environ Int 28:227–233. https://doi.org/10.1016/
science.3420404 S0160-4120(02)00018-1
Enfield DB (1981) Thermally driven wind variability in the planetary Toro R, Donoso C, Seguel RJ, Morales RES, Leiva MG (2013)
boundary layer above Lima, Peru. J Geophysical Res: Oceans 86: Photochemical ozone pollution in the Valparaiso region, Chile. Air
2005–2016. https://doi.org/10.1029/JC086iC03p02005 Qual Atmos Health 7:1–11. https://doi.org/10.1007/s11869-013-
Fiore AM, Naik V, Leibensperger EM (2015) Air quality and climate 0218-7
connections. J Air Waste Manage Assoc 65:645–685. https://doi. Toro R, Seguel RJ, Morales SRE, Leiva GM (2014) Ozone, nitrogen
org/10.1080/10962247.2015.1040526 oxides, and volatile organic compounds in a central zone of Chile.
Forouzanfar MH et al (2015) Global, regional, and national comparative Air Qual Atmos Health 8:1–13. https://doi.org/10.1007/s11869-
risk assessment of 79 behavioural, environmental and occupational, 014-0306-3
1008 Air Qual Atmos Health (2018) 11:993–1008

Underhill LJ et al (2015) Association of roadway proximity with indoor USEPA (2017) Final revised PAMS VOC target list. https://www3.epa.
air pollution in a peri-urban community in Lima, Peru. Int J Environ gov/ttnamti1/files/ambient/pams/targetlist.pdf
Res Public Health 12:13466–13481. https://doi.org/10.3390/ Wenig M, Spichtinger N, Stohl A, Held G, Beirle S, Wagner T, Jähne B,
ijerph121013466 Platt U (2003) Intercontinental transport of nitrogen oxide pollution
USEPA (1999) Method TO-1: method for the determination of vol- plumes. Atmos Chem Phys 3:387–393
atile organic compounds (VOCs) in ambient air using Tenax WHO (2000) Air Quality Guidelines for Europe 2000, second ed.
adsorption and gas chromatography/mass spectrometry (GC/ Copenhagen, World Health Organization Regional Office for Europe.
MS), in: Compendium of Methods for the Determination of http://www.euro.who.int/__data/asset s/pdf_file/0005/ 74732/E71922.pdf
Toxic Organic Compounds in Ambient Air, second ed. EPA/ Yin Y, Chevallier F, Ciais P, Broquet G, Fortems-Cheiney A, Pison I,
625/R-96/010b. Office of Research and Development, U.S. Saunois M (2015) Decadal trends in global CO emissions as seen
Environmental Protection Agency by MOPITT. Atmos Chem Phys 15:13433–13451. https://doi.org/
10.5194/acp-15-13433-2015