A Photochemical Modelling Approach to Investigate O3 Sensitivity to NOx and VOCs in the Urban Atmosphere of Delhi

Ambient air pollutants data were used to evaluate whether O3 formation at a specific site (Sirifort) in Delhi was limited by volatile organic compounds (VOCs) or oxides of nitrogen (NOx). For this purpose, a photochemical model OZIPR (Ozone Isopleth Plotting Research) based on Lagrangian trajectory model was applied to 9 ozone episodes that occurred at Sirifort from August to October 2006. Emissions data were estimated using an area-source box model. The results show that the prediction for peak O3 concentration agreed reasonably well with the observed data. O3-isopleth plots clearly reveal that O3 formation is more sensitive to VOC emissions for lower VOCs/NOx ratio, while for higher VOCs/NOx ratio, O3 formation is more sensitive to NOx emissions. However, for the purpose of practical O3 control applications at the observation site, it is concluded that VOCs emissions should be reduced while keeping a lower VOCs/NOx ratio.


INTRODUCTION
primary pollutants and which come from identifiable specific sources, such as vehicular traffic.This has led various researchers to employ different approaches in order to quantify this relationship.
Tropospheric O 3 , a well-known secondary pollutant is formed and sustained in the atmosphere due to a chain of complex reactions taking place among NO x and VOCs in the presence of sunlight.To prepare a control strategy for O 3 , it is crucial to investigate O 3 behavior in relation to its precursors (NO x and VOCs), which are mostly For the Greater Athens Area, using an extended database of air pollution and meteorological parameters in an Urban Airshed Model (UAM), Ziomas et al. (1998) concluded that ozone abatement strategy should focus mostly on controlling VOC emissions rather than controlling NO x (Blanchard and Stoeckenius, 2001) and compared O 3 predictions from six photochemical air-quality simulation models.For all simulations, peak ozone values increased following NO x control in 95% (median over all simulations) of the high-ozone (> 80 ppbV hourly mixing ratio in the base-case) grid cells having mean afternoon O 3 /NO z ratios less than 5:1, O 3 /NO y less than 4:1.Peak ozone levels decreased in response to NO x reductions in 95% (median over all simulations) of the grid cells having peak hourly ozone mixing ratios greater than 80 ppbV and where mean afternoon O 3 /NO z exceeded 10:1, O 3 /NO y was greater than 8:1.Arbilla et al. (2002) used an empirical kinetic modeling approach (EKMA) in order to simulate ozone concentrations for an urban downtown area with high vehicular traffic.The agreement between experimental and simulated results was quite good.The simulated ozone peak was obtained at 3:15 p.m. (23.0 ppb).A sensitivity-uncertainty analysis was performed and hypothetical scenarios were designed to illustrate the predictive potential of air quality models.Stein et al. (2005) examined the airborne measurements of sulphur dioxide (SO 2 ), total reactive oxides of nitrogen (NO y ), and O 3 taken from an instrumented aircraft to assess the governing photochemical processes of ozone formation.The sensitivity of ozone to changes in its primary sources was examined by simulating scenarios with varying rates of NO x and VOC emissions.The study of Stein et al (2003) show that for this particular case the measured and modeled upwind NO x sources are more effective than VOCs emissions for lowering O3.Cohan et al. (2006) applied nested grids of 36-, 12-and 4-km resolution to model an air pollution episode in Georgia., USA.A direct sensitivity analysis method used to compute the O 3 /NO y ratios seems to identify a photochemical regime in which reductions in response of ozone to emissions of its precursors: nitrogen oxides (NO x ) and volatile organic compounds (VOCs).All three grids predict that ozone production is limited primarily by the availability of NO x , and yield similar predictions of average ozone sensitivity to both regional and local emissions of NO x .The present study employs the Lagrangian trajectory model (Seinfeld and Pandis, 1988;Jang, 1999) because it allows us to study the response of the photochemistry in isolation

INPUT REQUIREMENTS OF OZIPR
from transport, which complicates the analysis in Eulerian grid model (Arbilla et al., 2002).In this context, the Ozone Isopleth Plotting Research (OZIPR) (Gery and Crouse, 1990)

OZIPR
(i) Sunlight Intensity: The OZIPR program uses a city's latitude, longitude, time zone, and day of the year being modelled to generate the appropriate diurnal pattern of photolytic reaction rates.Usually no changes need be made for this set of model inputs.
The OZIPR simulates complex chemical and physical processes of lower atmosphere through the use of Lagrangian trajectory model.The physical representation is a well-mixed column of air extending from the ground to the top of the mixed layer.This ideal air column moves with the wind (along the wind trajectory), but cannot expand horizontally.
Emissions from the surface are included as the air column passes over different emission sources, and air from above the column is mixed in as the inversion rises during the day.Very complex chemical mechanisms may be input into OZIPR to describe the chemical processes that occur within the modeled air mass.OZIPR performs a specified set of simulations to calculate ozone levels at fixed intervals.This allows for the plotting of fixed ozone concentration lines (isolines) as a function of initial precursors.
(ii) Dilution: In OZIPR model, dilution occurs as a result of the rise in atmospheric mixing height that typically occurs between early morning and mid-afternoon.The mixing height can be viewed as the top of a surface-based layer of air which is well-mixed due to mechanical and thermal turbulence.Specific input to OZIPR includes the early morning mixing height, the maximum afternoon mixing height, the time that the mixing height rise begins, and the time at which mixing height is finally attained.Procedure for estimating the early morning mixing height and maximum afternoon mixing height from available radiosonde measurements are outlined in EPA (1989).This was implemented using MATLAB ® program codes in the present study.Radiosonde measurements pertaining to Safdarjang, New-Delhi, measured with the help of Rutherford Appleton Laboratory, Chilton, and compiled by the British Atmospheric Data Centre (BADC, 2006) were used in this study.L = length of airshed, m; C = pollutant concentration in the airshed, mg/m 3 .Fig. 1 shows the map of New Delhi.A 10 km 2 enclosed square area around the observation site (Sirifort) is shown by a grey square on the map.An air parcel box was assumed having this area as the base area.The height of this box is given by mixing height determined by prevailing meteorological conditions.Air was assumed to be uniformly mixed within this box.For this assumed box model, the hourly mass emission rate were estimated by using Eq. ( 1).
(iii) Post-0800 Emissions: Post-0800 emissions refer to emissions occurring along the trajectory subsequent to the start of the model simulation (usually 08:00 AM).The actual inputs are expressed as emission densities (kg/km 2 -hr) of NMOC, NO x , and carbon monoxide (CO) concentrations that should be added each hour to the effect of fresh precursor emissions.
(iv) Ozone Transport: The two possible mechanisms by which ozone is transported into an urban area are: In the present study, Area-Source Box Model (Masters, 1998) was used to estimate mass emission rate (kg/km 2 -hr) from the available ambient concentration data (in g/m 3 ).The present study makes use of steady state solution of area-source model to estimate the mass emission rate from ambient pollutant concentration.The steady state solution of box model is given by (1).Advection of ozone along the earth's surface, and (2).Advection of ozone aloft, typically at night and during early morning hours, with downward mixing when the mixing layer increases later in the day.
Ozone transported at the surface is subject to surface reactions and scavenging by other species (e.g., NO) emitted during the night.As a result of night time atmospheric stability, ozone transported aloft does not come into contact with scavengers emitted during the night.Thus, overnight advection of ozone aloft is the more significant mechanism of transport from one urban area to another (EPA, 1989).(Masters, 1998) Where: Q = mass emission rate per unit area (mg/m 2 -s); u = average wind speed against one edge of the box, m/s; H = mixing height, m; This study mainly simulated the model for daytime only when the ozone aloft mechanism was less significant.As such, ignoring ozone aloft values does not change the final result significantly.This exercise was carried out for some of the well-tested US cities data as provided by EPA (2004).The change in simulated result is not more than ±5%.Hence, only surface ozone values were used in the present study; this is partly due to lack of ozone aloft data.
(v) Precursor Transport: Just as for ozone, precursor pollutants (NO x , VOCs) could be transported in both the surface layer and aloft.However, outside of urban areas, the surface layer is expected to be very shallow.Thus, long range transport of precursors in the surface layer may not be significant.Again, in the present study, for the reasons mentioned above, only surface precursor values have been used.
This study selects CB-4 mechanism as this mechanism uses a highly condensed method to represent reactions of VOCs, with the goal being to predict ozone from ambient mixtures as accurately as possible.It was evaluated against a large number of environmental chamber experiments (Gery et al., 1988).
For the purpose of carbon-fractionation, annual mean concentration data of VOCs for residential areas in Delhi pertaining to Srivastava (2005) have been employed.In the Srivastava (2005) study, the samples were collected in 2001 and analyzed to estimate a number of VOCs.Although there is a significant time gap between Srivastava (2005) and the present study, it can be a good representative for the purpose of estimating carbon fractions as per EPA (1989) recommendation.We have also compared this carbon-fraction estimation with the available CPCB hourly VOCs data at Sirifort from August to October 2006, which consists of only 5 VOCs.Within the range of ±0.05, o-Xylene, m,p-Xylene, Toluene, Benzene and Ethyl-benzene show an agreement with the carbon-fraction estimation based on Srivastava (2005).The small discrepancy shown here may also be due to the lack of other VOC data at Sirifort.Hence, the present study preferably uses carbon-fraction estimation based on Srivastava (2005) which is also more representative of VOC variability in Delhi, owing to its extensive broad-based sampling, monitoring and analysis using GC-MS techniques.
(vii) Temperature: Hourly temperature data must be utilized in OZIPR.Use of hourly temperatures allows reaction rates to be increased or decreased according to the hourly temperature.In the present work, hourly temperature data pertaining to Safdarjang, New Delhi, thankfully provided by Nautica Editrice Srl (1995-2006), was used.
(viii) Water Vapor: Ozone predictions are also sensitive to the amount of atmospheric moisture content.OZIPR estimates atmospheric moisture content using relative humidity values and ambient pressure level.
Hourly values of relative humidity were used for Safdarjang, New Delhi (provided by Nautica Editrice Srl, 1995-2006).
(viii) Biogenic Emissions: OZIPR also provides an option for biogenic emissions, such as isoprene, a-pinene, monoterpenes, and unknowns.EPA has prepared a computer program to estimate biogenic emissions rates on a county basis for the USA.This program estimates biogenic emission on the basis of day-specific meteorological parameters.However, in the absence of such data and facilities, EPA (1989) recommends to set it simply to zero, which was followed in the present case.

OZONE EPISODES, NO x , VOCs AND METEOROLOGICAL CONDITIONS
High ozone episode is a condition when the level of O 3 concentration exceeds a certain threshold and thereby posing threats to the human health.The World Health Organization has prescribed a standard of 100 g/m 3 (8-hour average) (WHO, 2006) for ambient O 3 concentration.In the present study, those conditions were considered as high ozone episodes when the ambient O 3 concentration repeatedly exceeded 100 g/m 3 level (Fig. 2).In Fig. 2       and a region that is NO x -limited where NO x reductions are more effective for reducing O 3 .Fig. 8 shows that for a NO x -value (> 0.04 ppm), O 3 production is more sensitive to the amount of VOCs than to the amount of emitted NO x (VOC limited regime).For a NO x -value (< 0.04 ppm), O 3 production is more sensitive to the amount of NO x -emission than to the amount of emitted VOCs (NO x limited regime).An examination of Fig. 9 reveals that O 3 formation is more sensitive to NO x for a higher VOC/NO x ratio, while for lower VOC/NO x ratio, O 3 formation is more sensitive to VOCs.The models were evaluated by comparing the model predicted peak O 3 with the observed peak O 3 concentrations.The observed vs. simulated peak O 3 concentration for all the ozone episodes taken together is shown in Fig. 7.In general, peak O 3 concentration has either been over-or under predicted by the model.However, the mean absolute percentage error (MAPE) of observed vs. simulated peak O 3 concentration varied from 15.2% on one occasion to 26.7% on another occasion.The MAPE for all the ozone episodes taken together is 21.3%.These MAPE values are well within the prescribed limit of 30% by EPA (1989) for OZIPR.Hence, the models can be adjudged to be sufficient to proceed with control estimate calculation.closer indicating a greater sensitivity.In Fig. 11, O 3 -isopleths are much closer and for a NO x value (> 0.04 ppm), O 3 production is more sensitive to the amount of VOCs than to the amount of emitted NO x (VOC limited regime) while for a NO x value (< 0.04 ppm) O 3 production is more sensitive to the amount of NO x than to the amount of emitted VOCs (NO x limited regime).The O 3 -isopleths in Figs. 12, 13 and 14 show a marked similarity.The isopleths are shifted towards the relatively higher VOC range.In addition, O 3 -isopleths in these cases are within the limited range of NO x concentration (up to 0.10 or 0.15 ppm).Figs. 15 and 16 also show a greater sensitivity of O 3 formation to VOCs for lower VOCs/NO x ratio, while higher VOC/NO x ratio, O 3 formation is more sensitive to NO x .

MODEL AND SCENARIO CONDITIONS
In all the cases except Fig. 8, it is remarkable to note that O 3 formation is relatively insensitive to VOC for a very high NO x concentration (say, > 0.26-0.27ppm).
Normally, ambient NO x concentration in Delhi varies within the range of 0.01-0.15ppm and very rarely exceeds 0.20 ppm (www.cpcb.nic.in).Therefore, the study of O 3 sensitivity to NO x , as well as to VOCs is important while delineating the conditions for controlling tropospheric ozone in the ambient environment of Delhi.In the present emission scenario, the analysis of the observed ozone episodes reveals that O 3 formation is more sensitive to VOC for lower VOC/NO x ratio, while for higher VOC/NO x ratio it is more sensitive to NO x ; although the range and threshold of NO x or VOC concentration (for O 3 sensitivity to VOC/NO x ratio) may vary from one episode to the other.These ranges and thresholds seem to depend upon the meteorological conditions, as one may notice that O 3 -isopleths of nearby days are more similar to each other than days which are further apart.For instance, O 3 -isopleths for 25 Sept,29 Sept and 4 Oct (Figs. 12,13 and 14) are very similar in nature, as well as in terms of range and threshold of NO x and VOCs concentration (for O 3 sensitivity to VOC/ NO x ratio).In the same way, O 3 -isopleths for 17 Oct and 21 Oct (Fig. 15 and 16) are similar in these characteristic terms.The typical behavior of O 3 -isopleths in Fig. 8 (13 Aug) shows that for lower VOC/NO x , O3 formation is greater, while a similar amount of O3-formation required much higher VOC/NO x ratio on the other days.This may be due to the fact that the early morning of 13 th August witnessed several thunderstorms (Nautica Editrice Srl, 1995-2006) and consequently higher O 3 concentrations at the start of the day.This might be the reason that even for low VOC/NO x ratio, O 3 -production seems to be higher on that day.Figs. 8 to 16 can be useful in the regulatory control of ozone.If VOC/NO x ratio is relatively high (> 1) and NO x -emission is within a certain range, the higher the VOC concentration, the higher the O 3 production.
While for the similar VOC/NO x ratio and NO x emission below the lower threshold value of the range, the higher the NO x emission, the higher the O 3 production.On the contrary, if the VOC/NO x ratio is relatively low (< 1) and NO x emission is above the lower threshold value of the range, O 3 production is sensitive to VOC concentrations.While for the similar VOC/NO x ratio and NO x emission below the lower threshold value of the range, O 3 production is more sensitive to NO x emission.Therefore, the most effective way to reduce the O 3 levels (or at least not to increase them) would be to reduce VOC emissions while keeping a relatively low VOC/NO x ratio.
It is worth mentioning here that the studies on O 3 , NO x and VOCs variation in the ambient urban environment of Delhi have so far been rather limited (e.g., Varshney and Aggarwal , 1992;Singh et al., 1997;Padhy and Varshney, 2000;Srivastava, 2005;Chelani and Devotta, 2006).Varshney and Aggarwal (1992) and Singh et al. (1997) deal with seasonal variation of O 3 in the Delhi's atmosphere, while the studies of Padhy and Varshney (2000) and Srivastava (2005) pertain to VOC variability.Chelani and Devotta (2006) forecasts NO 2 concentration using a hybrid model.However, so far no available reported studies deal with the interaction of O 3 , NO x and VOCs or O 3 sensitivity to NO x and VOCs in the context of Delhi.We expect this study to provide an insight into the nature of ozone episodes, its sensitivity to NO x and VOCs, and possible control measures in the urban atmosphere of Delhi.

CONCLUSIONS
A photochemical modeling approach was employed to examine the relationship between O 3 levels and concentration of NO x and VOCs aimed at suggesting ozone abatement strategy in an urban area of Delhi.Ozone formation was found to be more sensitive to VOCs emissions for lower VOC/NO x ratio.For higher VOC/NO x ratio, O 3 formation is more sensitive to NO x -emissions.However, for the purpose of practical O 3 control applications, it is concluded that VOC emissions should be reduced while keeping a lower VOC/NO x ratio in order to reduce the ambient O 3 levels at the observation site.
, encircled conditions depict the ozone episodes that occurred at Sirifort, New Delhi on various occasions.A total of 9 episodes from August to October 2006 were identified for the purpose of the present study.The corresponding NO x and VOCs concentrations are shown in Figs.3a & 3b.The observed meteorological conditions during the study period, such as temperature, relative humidity, and wind speed are shown in Figs. 4, 5 and 6.

Fig. 2 .
Fig. 2. Hourly average concentration of O 3 during the period of 3 August to 21 October 2006 at Sirifort.

Fig. 3 .
Fig. 3. Hourly average concentration of (a) NO x , and (b) VOCs, during the period of 3 August to 21 October 2006 at Sirifort.

Fig. 10
Fig. 10 also exhibit the similar characteristics with a difference that isopleths are relatively