Journal Archive

Two refineries (designated as Refinery 1 and Refinery 2) were selected for stationary source and air quality monitoring assignment. The two refineries selected are situated at different corners of India and also in different landuse pattern; one is located inland whereas the other is located at a coastal area. These refineries were chosen to understand emission patterns from refineries of different capacities and also to evaluate effects on ambient air quality in two different geographical areas and meteorological regimes. Refinery 1 is located in North Eastern part of India at Numaligarh in Assam, having a crude oil refining capacity of 3.0 MMTPA at the time of study. Superior kerosene oil (SKO), high speed diesel (HSD) and Aviation Turbine Fuel (ATF) are produced by Hydrocracker Technology for producing low sulphur products. Internally produced naphtha is used as fuel in the hydrogen generation unit (H₂U) and a fuel in captive power plant. Other product included Liquefied Petroleum Gas (LPG), Petroleum Coke, Parafin Wax and Sulphur. Low NO_x burners are in place to minimize NO_x generation from furnaces. There are two gas turbine generators (GTGs) with Heat Recovery Steam Generators (HRSGs), each having rated capacity of 30 MW.

Refinery 2 is located in the coastal area of Kochi in Kerala, situated in Southern part of India, that refined about 9.5 MMTPA crude oil at the time of study. This refinery has state of the art crude distillation unit and secondary processing units. Products from this refinery include LPG, Naphtha, Aviation Turbine Fuel, Kerosene, High Speed Diesel, Fuel Oils, Motor Spirit and Asphalt. Other products included Benzene, Toluene, Propylene, Poly Iso Butene, Bitumen and Sulphur.

Reconnaissance was conducted in the selected refineries for collection of secondary data on processes (Table 1), raw material consumption and environment management for planning on source emission and ambient air quality monitoring. Emission standards for particulates, SO₂ and NO_x prescribed by the State Pollution Control Boards in their ‘consent to operate’ letter to the industry were collected and studied.

The method prescribed by Bureau of Indian Standards (IS: 11255, Part 1 and 3-1985) was used for stationary source emission monitoring and determination of stack gas flow rate and concentration of analytes (BIS, 1985). Stationary source monitoring for particulate matter (PM) estimation was carried out under isokinetic flow conditions for a period ranging from 1-2 h under normal plant operations. A thermocouple sensor attached to a pyrometer and a modified “S-type” Pitot tube fabricated from SS 304 in conjunction with a stack monitoring kit (Model VSS-1) was used to estimate temperature of flue gas and differential flue gas pressure, respectively, from which flue gas velocity and flow-rate were calculated. A dry gas meter was used to record total volume of gas sampled. Particulate matter present in stack gas was collected in glass fibre thimble filters (19×90 mm; Whatman) capable of collecting particulates down to 0.3 μm and withstanding temperature up to 600°C. The thimble filters were conditioned at 50°C and 10% relative humidity (RH) in an oven followed by its storage in a humidity controlled dessicator before initial weighing. The same conditioning was also applied to the thimble filters after sampling and before final weighing. PM concentration in stack gas (mg Nm^-3) was estimated as gain in thimble weight against normalized volume (Nm³) of sampled stack gas.

A USEPA certified flue gas analyzer (Model Testo 350, Testo GMBH, Germany) fitted with electrochemical sensors was used for monitoring of O₂, SO₂, nitric oxide (NO), NO₂ and carbon monoxide (CO) in the stack gas. This analyzer was calibrated with standard certified concentrations of CO, SO₂, NO and NO₂ and was zero-calibrated with fresh air just before sampling, as per standard usage protocol. The fuel cell sensors deployed in the analyzer for SO₂ and NO₂ analysis, slowly and steadily decline in their output with time and therefore, must be recalibrated for a new zero at a pollution free ambient condition before they are used (Powrtech Solutions, Inc., https://www.powrtechsolutions.com/page/testo.htm; accessed on 25.2.2019). Subsequently, concentrations of PM, CO, SO₂, NO and NO_x measured in stack gas (mg Nm^-3) were first corrected to 6% carbon dioxide (CO₂) concentration and then integrated with stack gas flow rate (Nm³ h^-1) to estimate their emission rates (kg h^-1) and emission load (MT y^-1), considering continuous operation thoughout the year.

Ambient air monitoring (24-hourly) for a three-week period was conducted during the month of February for Refinery 1 and September for Refinery 2 at various locations around the refineries selected as per ASTM guidelines (ASTM, 2005). The locations of ambient air quality stations with respect to the refineries are depicted in Fig. 1. Fine Particulate Samplers (Model APM 550, Envirotech, Delhi, India) were used for monitoring PM₁₀ in ambient air. Ambient air enters APM 550 through an omnidirectional inlet designed to give aerodynamic cut-point for particles larger than 10 microns. The samplers were run at 16.7 LPM flow rate without Wins Impactor for PM₁₀ sampling. Calibration of flow rate of the instrument was undertaken by a Low Flow Calibrator (Model: APM-523, Envirotech) calibrated within 10-20 LPM flow range with error range of -0.4-8.20% full scale and expanded uncertainty (k=2) of +1.05% with traceability to FCRI, Palakkad. The combined (and expanded) uncertainty associated to atmospheric particulate measurements depends on uncertainty components (standard deviations) of relevant measurements viz. flow rate, time, mass, temperature, pressure, etc. Calibration for size is also critical. Therefore, uncertainties in impactor designing can be further added as one of the components of uncertainties (Aggarwal et al., 2013). Thermoelectrically cooled gaseous samplers (Model VTG II) were used to sample SO₂ and NO₂ by IS 5182 (Part 2): 2001 Method (BIS, 2001) and IS 5182 (Part 6): 2006 Method (BIS, 2006), respectively. The temporary ambient air quality monitoring stations were established with assured power supply, round the clock vigilance and facility for periodic sample collection. SO₂ and NO₂ were sampled at 1 LPM flow rate in impingers filled with designated absorbing media for SO₂ or NO₂. The IS 5182 Method (Part 2)-2001 (BIS, 2001) was used for SO₂ sampling and analysis. The impingers were calibrated by pipetting 35 mL absorbing reagent in 5 mL calibrated pipette and checking correctness of markings on the impingers. Sampling for SO₂ was undertaken for 24 hours continuously. This method allowed estimation of SO₂ in the range of 25 to 1,050 μg m^-3 and concentrations <25 μg m^-3 were measured by withdrawing higher air volumes. Likely NO_x interference was reduced by adding 1 mL of 0.06% sulphamic acid while ozone (O₃) was allowed to get decomposed by making the solution to stand for some time. Interference of trace metals was minimized by addition of 0.01% ethylene diamine tetra acetic acid (EDTA) to the absorbing solution before sampling. Calibration curve was drawn with the help of serial dilution of stock sulphite solution.

Fig. 1. 
Maps of study areas with marked refinery boundaries and ambient air quality monitoring stations (N1-N4) around Refinery 1 and (K1-K6) around Refinery 2.

Measurement of NO₂ was undertaken by sampling for 24 hours continuously following IS Method 5182 (Part 6): 2006 (BIS, 2006). The range of the method is reported to be 6 to 750 μg NO₂ m^-3 (0.003 to 0.4 ppm) while the analysis range is 0.04 to 2.0 μg NO₂ mL^-1. Under 50 mL absorbing reagent, sampling rate of 200 cm³ min^-1 for 24 h and absorption efficiency of 82%, method range is reported to be 6 to 420 μg NO₂ m^-3 (0.003 to 0.22 ppm). NO₂ concentrations (420 to 750 μg NO₂ m^-3) (0.22 to 0.4 ppm) are measured accurately by 1 : 1 dilution of sample. The positive and negative interferences of nitric oxide (NO) and CO₂ are low and therefore no correction was applied. Potential interference from SO₂ is minimized by letting SO₂ convert to SO4 = by adding hydrogen peroxide. In this method, reported intra-laboratory standard deviation was reported to be 8 μg m^-3 (0.004 ppm) while inter-laboratory standard deviation was 11 μg m^-3 (0.006 ppm) over a range of 50-300 μg NO m^-3 (0.027 to 1.16 ppm) (BIS, 2006).

On the other hand, CO was sampled in Tedlar Bags (SKC Inc., USA) passively through portable air sampling pumps and were analyzed ex situ in a CO analyzer (Model CO11, Environmental SA, France). Stability of CO in Tedler Bags is reported to be good and Tedler bags have been used earlier by various researchers to sample CO (Johnson, 2009; Chudchawal et al., 2000). SKC Tedlar bags are reported to have CO recovery rate of 90% within 48 hours after collection (Coyne et al., 2011). USEPA recommends Tedlar bags for determination of CO emissions from stationary sources in Method 10A (http://www.caslab.com/EPA-Methods/PDF/m-10a.pdf; accessed on 26.2.2019) and 10B (https://19january2017snapshot.epa.gov/sites/production/files/2016-06/documents/m-10b.pdf; accessed on 26.2. 2019). Standard recommendation for calibration by the manufacturer was followed and instrument calibration was done by two-point calibration process by zero air and a NIST traceable certified 100 ppm CO (Chemtron Laboratory, Mumbai, India). A suitable calibration coefficient was applied for correction of the obtained sample CO values. The instrument noise was 0.05 ppm and it had a lower detectable limit of 0.1 ppm CO (i.e. 100 ppb) and so anything below this concentration is reported as below detectable limit (BDL). This family of instrument complies with ISO 4224 and EN 14626:2005 standards, EPA, automatic reference method RFCA-206-147 in United States, TÜV No. 936/21206773/B, according to EN 14626. As per TÜV-Report (TÜV, 2008), the combined standard uncertainty and actual expanded uncertainty of CO analyzer (CO12M) in measuring CO had been found to be 0.1490-0.4433 μmol mol-1 and 7.11-10.29% which were good enough to fulfil the requirements of European Standard EN 14626.

Benzene was analyzed in a BTEX analyzer (Model VOC72M). This agreed with EN 14662-3 standard for measurement of benzene based on chromatographic separation of compounds in conjunction with photoionization detector (PID) (10.6 eV) (Environnement SA, http://www.hnunordion.fi/environnement/netissa/VOC72M_HNU.pdf; accessed on 25.11.2018). It is TUV Compliant following EN 14662-3. Sampling is done in a sorbent trap at a flow about 12 mL min^-1 that corresponds to a 165 mL sample volume in a 15-minute cycle. After sampling cycle, the trap is quickly heated to 35 to 380°C within 2 seconds to thermally desorb benzene and elute the same into GC column. Optimal separation in column is achieved by following a multi-ramp thermal cycle from 25°C to 160°C for flushing all the heavy compounds. The GC column is a stainless steel made (15 m×0.25 mm×1 μm, apolar). Measuring range of this instrument is maximum 1,000 μg m^-3 with a lower detectable limit of ≤0.05 μg m^-3 benzene and measuring noise of ≤0.025 μg m^-3 at 0.5 μg m^-3 benzene.

To record the prevailing meteorological conditions in the areas under study, meteorological data was collected from a portable meteorological station erected at a height of at least 10 meters at each refinery. Collection of meteorological data was carried out simultaneously with ambient air monitoring and windrose diagrams were prepared to understand and demarcate the zone (direction) of possible maximum pollutant concentrations during the study period. The study area maps are presented in Fig. 1.

USEPA’s Industrial Source Complex Short Term (ISCST3) Model (used by ISC-AERMOD software) that is based on Gaussian plume dispersion and suitable for single or multiple emission sources, was applied for predicting average 24-hourly ground-level concentration (GLC) as influenced by stationary source emissions (Cimorelli et al., 1998). Earlier ISC3 model has been used in several studies to predict concentration of pollutants (Bhanarkar et al., 2010; Bhanarkar et al., 2005; Bhanarkar et al., 2003; Abdul-Wahab et al., 2002). Windspeed and directions, two critical model input parameters, were recorded and processed according to the model requirement. The atmospheric stability classes were computed by using Turner’s classification (Hanna et al., 1982). By incorporating physical characteristics of emission source, emission rates, wind speed, wind direction, ambient temperature, stability classes and mixing height as inputs, dispersion modelling was carried out for predicting GLCs of pollutants within 5-km radius around the plant in winter. GLCs of SO₂ and NO₂ were modeled from their respective emissions from stacks by using ISC3 model and concentration contours over the study area were generated in order to identify the areas of concern. We have undertaken dispersion modeling by considering PM emissions from stacks, but in principle, we could consider PM primarily as PM₁₀, as refinery units are run on oil/ gas that are known to produce fine particles (Sánchez de la Campa et al., 2011; Kulkarni et al., 2007).

In Refinery 1, the total emission load of PM, SO₂ and NO_x were found to range from 87.9-221.2 MT y^-1 (UB and HRSG), 1.2-111.4 MT y^-1 (H₂U and SRU) and 14.3-2033.9 MT y^-1 (SRU and HRSG), respectively. PM concentration in flue gas ranged from 48.4 to 144.9 mg Nm^-3, the highest being from the utility boiler, but the emission load was highest (221.2 MT y^-1) in HRSG, followed by CCU (156.1 MT y^-1) and the lowest load was obtained from UB (87.9 MT y^-1) (Table 2). SO₂ was detected in the stack gas from all units, sulphur being a constituent in major raw materials. Highest SO₂ concentration and emission load were detected in SRU (646 mg Nm^-3 and 111.4 MT y^-1, respectively) followed by UB (83.8 mg Nm^-3 and 50.9 MT y^-1, respectively). Concentration of NO₂ was substantial in stack gas from all units, ranging from 57.8 to 445 mg Nm^-3 corresponding to emissions of 24.3 and 2,034 MT y^-1, respectively. In CDU, concentration and emissions of CO were 6.07 mg Nm^-3 and 5 MT y^-1, respectively. Concentration of PM, SO₂, NO₂ were well within the emission standards of MoEFCC in India.

In Refinery 2, the total emission load of PM, SO₂ and NO_x were found to range from 2.61-119.8 MT y^-1 (DHX 11 and COB), 9.65-656.3 MT y^-1 (DHX 11 and UB7) and 3.39-146.5 MT y^-1 (DHX 11 and CPP), respectively. The concentration of particulate matter in the stack gas was in the range of 8-99 mg Nm^-3 and the emission load ranged from 2.6 to 119 MT y^-1 (Table 3). SO₂ emission was highest in UB 8/9 (935 mg Nm^-3) and lowest in DHX11 (37 mg Nm^-3) while NO_x concentration in stack gas ranged from 13 mg Nm^-3 in DHX 11 to 235 mg Nm^-3 in UB6, respectively. CO concentration was 1-46 mg Nm^-3. Concentration of PM, SO₂ and NO_x were also well within the emission standards of MoEFCC. An overview of emissions from refineries around the world indicated that emission loads of most of the air pollutants observed in this work were comparable or lower than that found in some other European, Canadian and Asian refineries (Table 4). High SO₂ emissions in Refinery 2 as compared to Refinery 1 might be due to use of high sulphur fuel oil as well as high amount of crude processing/ higher production capacity of Refinery 2 as compared to Refinery 1. High emissions levels of SO₂ have been reported by Rao et al. (2006) in 2004 at Gujarat Refinery in India with crude oil processing capacity of 13.5 MMTPA. Karbassi et al. (2008) also reported high SO₂ emissions at Tabriz oil refinery which used liquid fuels containing high sulphur.

The windrose diagram prepared for Refinery 1 indicated that prevailing wind direction was from North and Northeast direction, with the wind speed prevailing within a range of 2-5 m s^-1 (Fig. 1). Winds from other directions were also observed on a few occasions with a predominance of North-Western direction. Calm condition was significantly prevalent, in 44% cases. The 24-hourly average levels of SO₂, NO₂, CO, benzene and PM₁₀ around the Refinery 1 prevailed within the limits promulgated in National Ambient Air Quality Standards (NAAQS) (CPCB, 2018a). Due to wind effect, higher concentration of pollutants are observed at sites located in downwind directions. SO₂ levels were always found to be low, while very low concentrations of NO₂ were found at one station (Table 5). On the other hand, CO was detected at all the locations. Benzene was detected at a few locations but was persistently low in concentration, ranging from 0.17 to 0.31 μg m^-3. Maximum PM₁₀ concentration was 65 μg m^-3 at location N1 followed by 53 μg m^-3 at N4.

As per the windrose diagram for Refinery 2, the prevailing wind direction was from West-South West with wind speed mostly falling in the range of 0.5-2.1 m s^-1 (Fig. 1). Wind from the North-Western direction was also conspicuous. Higher wind speed of 2.1-3.6 m s^-1 were observed on a few occasions. Weather Conditions prevailing near the refineries during the month of monitoring showed substantial day-to-day variability in relative humidity, especially in the minima, while temperature and pressure variability were comparatively lower. As the monitoring exercises were undertaken during February in Refinery 1 and September in Refinery 2, average temperature difference between the refineries was about 4-5°C while temperature in Refinery 1 showed a slightly increasing trend due to approaching summer. In general, slightly higher concentration of pollutants are observed at sites located in downwind directions. Around Refinery 2 also, 24-hourly average levels of SO₂, NO₂, CO, benzene and PM₁₀ prevailed within the limits prescribed as NAAQS. In general, higher concentration of pollutants are observed at sites located in downwind directions. SO₂ levels were persistently low, while very low concentrations of NO₂ were found at a few stations (Table 5). Low ambient concentrations of SO₂ and NO₂ could be due to the location of refineries in open land with almost no blockade that promoted good dispersion and dilution of pollutants. Also, negligible presence of polluting industries in the vicinity and low vehicular traffic ensured low levels of ambient SO₂ and NO₂. Low ambient levels of SO₂ and NO₂ have been reported by Central Pollution Control Board (CPCB) in 2008 and 2010 at a few cities and towns of India (http://cpcb.nic.in/openpdffile.php?id=UmVwb3J0RmlsZXMvTmV3SXRlbV8xNDdfcmVwb3J0LTIwMDgucGRm; http://cpcb.nic.in/openpdffile.php?id=UHVibGljYXRpb25GaWxlLzYyOF8xNDU3NTA1MzkxX1B1YmxpY2F0aW9uXzUyMF9OQUFRU1RJLnBkZg; both accessed on 22.2.2019). CO and benzene were detected at all the locations but were low in concentration. However, benezene concentrations were found to be generally higher in this refinery than Refinery 1. Maximum PM₁₀ concentration was to the tune of 69 μg m^-3 at site K3 followed by 61 μg m^-3 in site K1. Observed ambient air quality in our study and ambient air quality around other petroleum refineries in several other countries was found to be comparable (Table 4).

Ambient concentration of PM₁₀, SO₂ and CO were converted to respective air quality indices (AQI) as per USEPA’s concentration-AQI conversion principles and formulae (AirNow, 2018). It was noted that while AQI of CO never entered the zones of concern at any site and were always good, AQI of PM₁₀ were moderate at two sites, the rest being good (Table 6). In case of SO₂, no AQI could be calculated as SO₂ concentration values were outside the calculable range. AQI for NO₂ could not be calculated as the required 1-h average NO₂ concentration data needed for AQI calculation were not available. Further, AQI was also developed as per the formula AQI used by Central Pollution Control Board (CPCB) of India (CPCB, 2018b) (Table 6). Since all the eight pollutants included in AQI calculation was not monitored, AQI was calculated based on concentration of minimum necessary three pollutants amongst which one should be either PM_2.5 or PM₁₀. Sub-indices were also generated for each pollutant to evaluate air quality status for that particular pollutant. The pollutant-wise calculated sub-index values for PM₁₀ ranged from 38-65 and 38-69 for refinery 1 and refinery 2 respectively. However, CO showed sub-index values varying between 9-26 and 9-25, respectively, for refinery 1 and refinery 2. Similarly, the sub-index values for NO₂ ranged from 6-11 and from 5-8, respectively, for refinery 1 and refinery 2, whereas for SO₂, it ranged from 4-6 for both the refineries. From the above calculations, it became apparent that AQI had never been poor under any circumstances at the selected sites, indicating no risk of significant health impacts to inhabitants residing near these refineries. Considering that clean burning fuels were used in these refineries and low PM concentration were obtained in stack gas from various units, AQI of all air quality monitoring stations in both the refineries were found to be satisfactory. Analysis of particulate-bound SO₄⁼ and NO₃^- in filters containing either ambient particulates or stack gas particulates was not deemed crucial for drawing important conclusions. Hence, this aspect was kept out of scope of this work.

Air quality modeling exercise undertaken by ISCST 3 Model (used by ISC-AERMOD software) with the stationary source emission data of Refinery 1 revealed that major dispersion of PM, SO₂ and NO₂ occurred in southwest and northeast directions due to predominant winds patterns. However, the impact of refinery emissions was not significant and ambient air quality levels of these pollutants did not exceed NAAQS. The maximum modeled GLC of PM, SO₂ and NO₂ were 2.4, 2.6 and 14.9 μg m^-3, respectively (Table 7). The isopleths of predicted concentrations for SO₂ in Refinery 1 are presented in Fig. 2a. In Refinery 2, predicted air quality generated by the model indicates that the maximum GLCs of PM, SO₂ and NO₂ were 2.1, 17.1 and 4.9 μg m^-3, respectively (Table 6), which are lower than NAAQS and occurred primarily in the eastern direction. The isopleths of predicted concentrations for SO₂ in Refinery 2 are presented in Fig. 2b. Maximum GLCs of these pollutants was observed within 3-4 km in eastern direction.

Fig. 2. 
Isopleths and windroses superimposed on maps showing predicted GLCs (μg m^-3) of SO₂ in (a) Refinery 1 (b) Refinery 2 and wind patterns, respectively [circles are of 5 km radius around the centres of refineries].

Low to moderate difference was observed between observed and modeled GLCs of the air pollutants which has been earlier reported by other researchers also (Abdul-Wahab et al., 2002). The observed concentration of PM₁₀ in ambient air was found to be similar (Table 7) in both refineries as the total emission of PM from all stacks were similar in both. The concentrations for PM₁₀ in ambient air are higher than predicted values due to presence of other sources of particulates like vehicular emissions in nearby roads, fugitive dust emissions from nearby agricultural fields and road construction activities, emissions from other stationary sources that included small workshops and emissions from household biomass burning which were not considered by the model. The modelled values of PM by ISCST3 was also similar (2 and 3 μg m^-3, respectively), that depended on the stack PM emissions and hence validated. Though the emission of SO₂ was much more in Refinery 2 than Refinery 1, the ambient concentration of SO₂ was similar in both, probably because of higher conversion of SO₂ into sulphate in the ambient air of Refinery 2, which is situated near sea shore. Conversion of SO₂ to sulphate on sea salt in atmosphere is reported (Alexander et al., 2005) and hence, a predominance of this reaction might have played an active role in high conversion of emitted SO₂ near Refinery 2.