Study of Extreme CO , NO 2 and O 3 Concentrations at a Traffic Site in Delhi : Statistical Persistence Analysis and Source Identification

Persistence in the air pollutant concentration events exceeding the standard threshold was studied by using detrended fluctuation analysis. Time series of gaseous pollutant concentrations of CO, NO2 and O3 observed during 2000–2009 at a traffic site in Delhi were considered in this study. Statistical persistence was observed in exceeding concentration events of CO and NO2, whereas O3 exceeding time series showed anti-persistent behavior due to its secondary nature. The formation of O3 formation was explored by examining its relations with precursor pollutants. A high proportion of the exceeded O3 concentration was observed to be CO limited. Back trajectory analysis was then carried out to trace the path of the air parcels with high ozone concentration events, suggesting two clusters of trajectories; one from south-east and the other from northwest of the study site.


INTRODUCTION
The extreme concentrations of air pollutants has been a matter of importance as it pose serious threat to human health and can damage the surrounding environment (Ercelebi and Toros, 2009).The study of extreme values is useful in planning and management for source emission reduction (Sharma et al., 1999).The extreme value analysis has been usually studied on the time series of number or total time of exceedances in a time intervals and duration of the events continuously exceeding the threshold concentration.The probability distributions are often fitted to obtain further projections and probability of exceedances (Ercelebi and Toros, 2009;Lonati et al., 2011), which requires the assumption of identically and independently distributed time series, whereas air pollutant concentrations are often nonstationary and auto-correlated.The concentration events exceeding the threshold over time reveals the information about the time structure of the exceeded levels, which can be further used to obtain the understanding of the dynamics of high air pollution events, which is useful for revealing the underlying persistence and temporal correlations.The persistence analysis is helpful in detecting the long-range correlations which is a useful property in detecting selforganizing behavior of the time series of air pollutant concentrations (Shi and Liu, 2009) and for making further projections.Several approaches such as rescale-range analysis, spectral analysis and detrended fluctuation analysis (DFA) have been used in literature for persistence or longrange correlations analysis (Zhu and Liu, 2003).DFA has advantages over other methods of persistence analysis such as spectral analysis and rescale-range analysis as it can avoid spurious detection of correlations due to non-stationarity (Shi and Liu, 2009) and can be applied even in the presence of trends in the time series (Varotsos et al., 2005;Shi and Liu, 2009).The long memory property using DFA has been inferred in air pollutant concentrations in several studies (Varotsos et al., 2005;Lu and Wang, 2006;Varotsos and Kirk-Davidoff, 2006;Varotsos et al., 2007;Weng et al., 2008;Lau et al., 2009;Yuval and Broday, 2010;Chelani, 2012).
Delhi, one of the most polluting cities in the world, has a hot and humid climate, with extremes in May-June followed by monsoon.The average temperature in summer ranges from 25 to 46°C, whereas temperature falls substantially down to as low as 3 to 4°C at the peak of winter which occurs during November to February.The main climatic influence in the area is the monsoon, typically from June to October with mean annual rainfall of 71.5 mm.The study site, ITO, is a traffic intersection site with huge traffic of about 113,000-176,000 vehicles per day and commercial buildings located near by the site.The air pollution problem in the city is a major concern as in-spite of adopting several control measures such as change in fuel; phasing out of 15 years old vehicles; shifting of industries to outer areas and lowering of sulphur and benzene content in automobile fuel, the concentrations of particulate matter, NO 2 and ground ozone (O 3 ) are increasing (Gurjar et al., 2004).In a study on ozone levels in Delhi, Singh et al. (1997) showed that the concentrations have either approached or exceeded the limit of World Health Organization air quality standards.The atmospheric emissions from Delhi contribute to acidification, eutrophication and the occurrence of large-scale pollution haze leading to perturbations of the climate and water cycle (Lelieveld et al., 2001;Gurjar et al., 2004).
In this study, the persistence in exceeding levels of CO, NO 2 and ground O 3 concentrations is studied using detrended fluctuation analysis.As high concentration event are important from health point of view, along with the temporal analysis, it is attempted to find out the origin of exceeded O 3 levels by using the relationship with precursors (CO and NO 2 ) and back trajectory analysis.The data is obtained from Central Pollution Control Board (CPCB), New Delhi (www.cpcb.nic.in).The details of sampling and analysis can be obtained from CPCB.The frequency of data is 24 hourly.The missing values, which were around 10%, were replaced by the mean of the time series.The exceeded levels are estimated by comparing with the standards provided by CPCB as 2000 μg/m 3 for CO, 80 μg/m 3 for NO 2 and 100 μg/m 3 for O 3 concentration.

DETRENDED FLUCTUATION ANALYSIS
The presence of long-range correlations or persistence in the time series can be detected by using DFA.Persistence is characterized mainly by the time series which follow the direction of historical patterns whereas anti-persistent time series follow reverse direction of historical observations.DFA permits the detection of intrinsic self-similarity embedded in a non-stationary time series and avoids the spurious detection of apparent self-similarity (Peng et al., 1994;Zhu and Liu, 2003).It calculates the root-meansquare fluctuation of integrated and detrended time series.To apply the DFA algorithm, the total length of the time series where y(i) is the time series with mean 〈y〉of all the samples,  is the time lag, k = 1, 2, …, N and N is the length of the time series.The integrated time series is divided into segments of equal length n and the least-squares line is fitted to the data in each segment.The y-coordinate of the straight-line segments is denoted by Z n (k), which is used to detrend the time series z(k) as z(k) -Z n (k) in each segment.The root mean square fluctuation of integrated and detrended time series is calculated by; Repeating the computations for all the segment sizes provides an increasing function relationship between the average fluctuation F(n) and the segment size n.A linear relationship on a log-log graph indicates the presence of scaling, i.e., F(n)~n α , where α is the scaling exponent, can be obtained as a slope of the line for all the segment sizes.The scaling exponent gives an indication of the nature of the time series.For 0 < α < 0.5, it indicates the presence of power-law anti-correlations in the time series, whereas 0.5 < α < 1 suggests the long-range power law correlations.Time series corresponds to white noise if α = 0.5.Even sometimes α ranges between 1 and 1.5, which suggests the stronger long-range correlations.

RESULTS AND DISCUSSION
The 24 hourly CO, NO 2 and O 3 concentrations observed over 2000-2009 are plotted in Fig. 1(a) and statistical summary of observed concentrations is given in Fig. 1(b).The maximum CO concentration (> 20000 μg/m 3 ) was observed in 2001 and 2009, whereas maximum NO 2 and O 3 concentrations (> 400 μg/m 3 ) were observed during 2002-2009.In India, the standard limits for air pollutant concentrations have been provided by Central Pollution Control Board (CPCB), New Delhi in 1994 and revised in 2010.The standards for CO, NO 2 and O 3 are 2000 (for 8h frequency), 80 (for 24h frequency) and 100 (for 8h frequency) μg/m 3 , respectively.The standards are not available by CPCB for annual average data.To have the idea about the status of gaseous air pollution in the area, the annual concentrations are compared with the available standards.Looking at the annual average concentrations, these are close to CPCB standards for CO and NO 2 , whereas O 3 concentration is below the standard (Fig. 1(b)).75 th percentile ranges between ~2800-5000 μg/m 3 , ~70-170 μg/m 3 and ~30-60 μg/m 3 for CO, NO 2 and O 3 suggesting exceeded standards for CO and NO 2 .The number of exceedances to standard limits during the study period are also plotted in terms of probability of exceedance in Fig. 2. The threshold exceedances are decreasing for CO concentration and increasing for NO 2 concentration indicating the positive impact of introduction of cleaner fuel from diesel to Compressed Natural Gas in vehicles in Delhi during 2001 on CO levels.Comparing the results with the CPCB guidelines, which states that "24 hourly or 8 hourly or 1 hourly monitored values, as applicable, shall be complied with 98% of the time in a year, 2% of the time, they may exceed the limits but not on two consecutive days of monitoring", it can be observed from Fig. 2 that probability of exceedance is much above the limit of 2% of the time for CO and NO 2 for all the years, whereas for O 3 , it is crossed only during 2007-2009.The analysis suggests that O 3 concentrations were not significant before 2007.The elevated O 3 concentration levels after 2007, although not much intensive are becoming a matter of concern in Delhi (http://www.cseindia.org/AboutUs/press_releases/press_20070614.htm).
The number of exceeded events of three pollutants in three seasons; summer, winter and monsoon are also estimated to gain insight into the seasonal variations in high concentration events.It is observed that the fraction of high CO and NO 2 concentration in winter was highest followed by summer and monsoon with average concentration of 4024 and 128 μg/m 3 in winter, 3082 and 116 μg/m 3 in summer and 3292 and 117 μg/m 3 in monsoon respectively.The number of events of exceeding O 3 concentration are highest in summer followed by monsoon and winter with average of 116, 117 and 212 μg/m 3 , respectively.Although the average concentration is high in winter, the frequency of episodes is observed to be larger in summer due to the effect of photochemical reactions in high temperature in summer.Further analysis is carried out to study the temporal correlation in exceeded events time series of three pollutant concentrations using detrended fluctuation analysis.The exceedance time series has mean ± standard deviation of 3511 ± 1539, 121 ± 56 and 149 ± 69 μg/m 3 for CO, NO 2 and O 3 concentrations, respectively.For DFA, the segment sizes were chosen in the range 10:10:1000 as it can provide detailed time scaling of the data.A code in MATLAB (version 7.0) was written to compute fluctuation function F(n) over different segment sizes n.As mentioned, log F(n) is then plotted against log n and the slope of the fitted straight line gives the measure of scaling component α.Fig. 3 shows the variations in log F(n) over log n.It can be observed that α = 0.7655, 0.8358 and 0.4782 for CO, NO 2 and O 3 concentrations, respectively.As α > 0.5 for CO and NO 2 , the exceedance time series for both the pollutants shows power-law persistence whereas α < 0.5 for exceedance time series of O 3 depicts power-law anti-persistence approaching white noise.
As well known, air pollution is a complex phenomenon and extreme air pollution events are governed by huge number of inter-related factors.The persistence or longrange correlations in the extreme event time series of CO and NO 2 is suggestive of dependence on the previous observations along with the interaction with other variables.Temporal uniformity in the source of high CO and NO 2 exceedance events can also be assumed with this persistence.In Delhi, mobile source emissions are the dominant contributors of CO pollution (Gurjar et al., 2004).In order to ascertain this, CO/NO 2 ratios were computed following Aneja et al. (2001).An average ratio of 40 was observed during 2000-2009 quite near to the value of 50 observed by Aneja et al. (2001).They attributed the high CO/NO 2 ratio to mobile source emissions in Delhi.Although change in fuel from diesel to CNG has been successfully implemented in Delhi, mobile source is still the significant   contributor of CO pollution, which has even not changed over the years.The shift to CNG along with the natural gas power plant contributes most of the emissions of NO x (Gurjar et al., 2004).

Anti-Persistence in O 3 Exceedances
The applicability of DFA on the exceedance time series of O 3 concentration with small sample size may however be suspected.Hence, the findings are confirmed by using spectral analysis.Although less sensitive technique for determining the temporal correlations compared with rescale range analysis and DFA, spectral analysis has been widely applied in air pollution literature (Shi et al., 2008).The details about the spectral analysis method are given elsewhere (Shi et al., 2008).Power spectrum is calculated over time period and the periodogram values are plotted in Fig. 4 for O 3 concentration, which shows the flat power spectrum with no significant frequencies almost corresponding to white noise (http://www.vanderbilt.edu/AnS/psychology/cogsci/chaos/workshop/Tools.html) which confirms the findings of DFA.
The anti-persistence in O 3 concentration exceedance time series suggests that the event at a time is independent of historical exceeding events, whereas persistent power-law behavior of CO and NO 2 time series of exceeding events suggests the temporal correlations.The exceeding events of O 3 concentrations are thus not derived from the previous exceeding events.As well known, precursor gases emissions in the presence of sunlight and favorable meteorological conditions generate ozone concentrations (Reddy et al., 2011).In a similar study using persistence analysis, Varotsos et al. (2003) and Chelani (2009) attributed the high ozone concentrations to solar radiation along with change in weather conditions.Kumar et al. (2008) attributed the ozone formation to NO x for a higher hydrocarbon/NO x ratio.In order to identify the origin of the exceeding O 3 concentration events, a simple linear relationship is assumed between the exceeding O 3 and low CO or NO 2 concentrations.Although the examination of origin of O 3 exceedance requires the data on both, NO x and VOCs concentration (Swamy et al., 2012), but due to the non-availability of VOCs concentrations, the corresponding CO and NO 2 concentrations were examined.If the corresponding observation of either CO (marked as C in the figure) or NO 2 (marked as N in the figure) is below the standard threshold, the particular pollutant is considered as significant in generating the exceeding event of O 3 concentration.If both are less than the standard threshold, both are considered as significant (marked as N-C in the figure) and if neither are significant, transport from nearby areas (marked as TRAN in the figure) is considered as the origin of O 3 events.Fig. 5 depicts the O 3 exceedance events marked with their origin.It is observed that CO contribution is highest (~37%) followed by transport from nearby areas (~32%) and NO 2 contribution (~29%).In a study on the persistence analysis of ozone concentrations at same site, Chelani (2009) suggested the causal behavior of hourly ozone concentration in all the months during 2005-2006 except during July and August.During these two months, the intermittence in ozone concentrations was proposed to be due to high VOCs rather than NO x whereas in other months the ozone concentrations were observed to be persistent and governed by meteorology.Although corresponding VOC data was not available in this study, in July and August, the significant source of exceeding ozone levels is observed to be CO rather than NO 2 confirming the findings by Chelani (2009).Comparing the results with Kumar et al. (2008), CO/NO 2 ratios were calculated as hydrocarbon data was not available.The contribution of NO 2 is observed for the events with high CO/NO 2 ratio.The ozone variation against NO 2 /NO ratio is also considered as one of the indicator of local ozone formation (Shao et al., 2009;Han et al., 2011).As can be seen from Fig. 5, which is drawn for the events assumed to be the result of NO 2 , an increase in O 3 levels with NO 2 /NO ratio with R 2 of 0.49 is indicative of NO 2 limited ozone formation.
In addition to the relationship with CO and NO 2 concentrations, the origin of high O 3 concentration is also analyzed with the help of back-trajectories.The contribution by nearby areas is investigated by tracing the path of the air parcels at the site on particular days.For this, backtrajectories were computed using Hybrid Single Particle Lagrangian Integrated Transport (HYSPLIT) model (Draxler  and Rolph, 2011) developed by NOAA (http://www.arl.noaa.gov/ready/hysplit4.html).For the air masses arriving at the receptor, back-trajectories were computed to identify the source regions for the days on which the transport is suspected.The analysis was performed with the GDAS meteorological dataset and the starting time of 0000 UTC, altitude of 500 m AGL and total run time of 24 hrs.It can be observed that there is a considerable variation in the origin of air parcel trajectories as can be seen from Fig. 6.Unlike to contribution of dust from the Thar deserts of Rajasthan (Ram and Sarin, 2010) which is located SW of Delhi, in general, two clusters of trajectories can be observed; one from SE sector of Delhi originated in Utter Pradesh state and other from NW sector originated in Haryana and Punjab state.Monitoring studies are however required in these source regions to ascertain the above observations.

CONCLUSIONS
Persistence in exceeding events of gaseous air pollutant concentrations is studied using detrended fluctuation analysis at a site in Delhi during 2000-2009.Statistical persistence is observed in exceeding concentration events of CO and NO 2 , where as O 3 exceeding time series showed anti-persistent behavior.This is mainly attributed to the secondary nature of O 3 concentrations.Further investigation by spectral analysis confirmed the anti-persistence in O 3 exceeding levels.The persistence in CO and NO 2 concentrations can be considered as the temporal dependence on the historical events.With the persistence property, the possibility of uniformity in the sources of high concentration events cannot be ruled out.In the study area although CNG has been successfully implemented in vehicles, mobile source is still the significant contributor of CO and NO 2 pollution in ambient air.To understand the formation mechanism of O 3 events, relationship with CO and NO 2 is studied as the data on VOCs concentrations is not available.Most of the high concentrations of ozone at the site are observed to be local in nature limited by CO.Transport from nearby areas also contributes to the high ozone events, which however needs further confirmation by monitoring the corresponding precursor concentrations at the site and in the pathways of air parcels.The analysis of back-trajectories revealed two clusters of trajectories; one from SE sector of Delhi and other from NW sector.The study is useful to understand the temporal behavior of the high concentrations of three pollutants and source origin of ozone high ozone concentrations which can be further utilized for making policy decisions to curb ground ozone pollution.

Fig. 1
Fig. 1(a).Time series of exceeded concentrations during 2000-2009.The x-axis indicates the exceeded observations during 2000-2009.The number of exceeded observations were 2566, 1943 and 41 for CO, NO 2 and O 3 concentrations, respectively.

Fig. 1
Fig. 1(b).Statistical summary of CO, NO 2 and O 3 concentration at traffic site in Delhi over 2000-2009.

Fig. 3 .
Fig. 3. Detrended fluctuation analysis of exceeding events of three air pollutant concentrations at a site in Delhi for the data during 2000-2009.

Fig. 5 .
Fig. 5. O 3 concentration (only for the exceeding days) marked with its origin.N denotes NO 2 generated, C denotes CO generated, TRAN denotes transport from nearby areas, N-C denotes both NO 2 and CO generated.

Fig. 6 .
Fig. 6.Back-trajectories at Delhi traffic station for the days of exceeded O 3 concentrations during 2000-2009 at starting time of 0000 UTC and altitude of 500 m AGL calculated using NOAA HYSPLIT model with meteorological data set-GDAS, total run time-24 hrs and location 28.54° 77.188°.The origin of the trajectories is marked with a cross and the trajectories are marked with red color.