1 Introduction

SO2 is a toxic pollutant in the atmosphere. An excess SO2 amount in the atmosphere can cause respiratory and cardiovascular diseases and damage to buildings (Finlayson-Pitts and Pitts 2000 and references therein). SO2 reacts with OH radical and forms H2SO4 droplets, high amount of which causes acid rain. The sulphate aerosols are scattering type of aerosols and affect the climate (Charlson et al. 1990). These aerosols have negative radiative forcing in the atmosphere and counter the global warming by greenhouse gases and black carbon (Charlson et al. 1991; Kiehl and Briegleb 1993). SO2 has a lifetime of about two days to two weeks in the troposphere (von Glasow et al. 2009). It is removed from the atmosphere either by dry or wet deposition or conversion into aerosol particles.

Major anthropogenic sources of SO2 are coal burning power plants. There is a considerable reduction in SO2 emissions from regions like USA, Europe and China during the period 2005–2014, whereas in India the emissions showed a positive trend during the same period (Klimont et al. 2013; Lu et al. 2013; Krotkov et al. 2016). Klimont et al. (2013) have reported that SO2 emissions from India have increased by about 40% between 2005 and 2010. Lu et al. (2013) have calculated 61% increase in SO2 burden over India from 2005 to 2011–2012 using OMI satellite data, whereas, Krotkov et al. (2016) have reported nearly a factor of 2 increase in column SO2 burden from 2005 to 2014 using the same satellite data and stated that India has become the second largest SO2 emitting country after China in the world. The major sources of SO2 over India are coal burning power plants, which contribute nearly 50% of total emissions (Klimont et al. 2013; Lu et al. 2013). Usage of biofuels also emits SO2 into the atmosphere. Gadi et al. (2003) have estimated 0.4 ± 0.3 Tg of SO2 emissions from biofuels in India for the year 1990. The latest SO2 emissions from biofuels are estimated at 0.75 Tg for year 2010 (Klimont et al. 2017). There are also a few natural sources of SO2 like volcanoes and geothermal hot springs. Volcanoes contribute about 1.5–50 Tg per year of SO2 globally (Textor et al. 2004). However, SO2 burden over India from volcanic emissions is expected to be negligible.

Due to its high solubility, life-time of SO2 is less in the lower atmosphere (typically two days). However, under favorable conditions such as during low humidity or high wind speed conditions, SO2 can be transported to hundreds of kilometers from the place of emissions. A large SO2-rich pollution plume of East Asian origin was detected by aircraft based CIMS (Chemical Ionization Mass Spectrometry) measurements at 3–7.5 km altitude over the North Atlantic (Fiedler et al. 2009). Similarly, Tu (2004), Mallik et al. (2012) and Naja et al. (2014) have shown the episodic long-range transport of SO2 to their observational sites. However, not much information is available about the SO2 transport to southern India. SO2 is a part of legal air pollution criteria and hence its emission from industries and transport to downwind regions is a matter of regulation. In this context, it is important to have information on long-range transport as well as transport from natural sources at a given place. In the present work, we examine the transport pathways of SO2 to a rural site in south India (Gadanki, 13.48°N, 79.18°E) using FLEXPART model and the observed ratios of SO2 and NO2 at Gadanki. We also evaluate reanalysis SO2 data-set available from ECMWF (European Centre for Medium-Range Weather Forecasts) known as CAMS global reanalysis of chemical species against ground-based observations.

2 Observation site and instruments

Gadanki (13.48°N, 79.18°E) is a rural site surrounded by small hills. It has a tropical wet climate and experiences significant rainfall during both north-east and south-west monsoons. Temperatures are high during March to June with monthly mean values between 28° and 31°C and low during November to December with monthly mean values ranging from 21° to 23°C. During the study period, the highest monthly mean temperature is observed in May 2010 with 31.2° ± 4.3°C and the minimum is observed in December 2010 with 21.7° ± 3.2°C. The relative humidity varied from 55% ± 24% to 87% ± 11% during the study period. Majority of the rain occurs during July to December at Gadanki. Total amount of rainfall during north-east monsoon (October–December) for the years 2010 and 2011 are 298 and 330 mm which is 26% and 41% of the total rainfall in the respective years. Whereas during south-west monsoon (July–September), total rainfall received are 749 (65%) and 383 mm (47%) in the respective years. More description of the site and its meteorology can be found elsewhere (Renuka et al. 2014; Gadhavi et al. 2015).

2.1 In-situ measurements of SO2, NO2 and BC

SO2 mixing ratios are measured using Thermo Scientific – USA (model 43i) SO2 analyser at Gadanki since 2010. The accuracy of the SO2 analyser is 0.05 ppbv with a response time of 5 minutes. The SO2 analyser works on the principle of fluorescence. When SO2 molecules are irradiated with UV light, they go into excited state and re-emit the light at different wavelength than they were irradiated, while coming down to the ground state. The photon flux resulting from fluorescence which is being measured by the analyser is proportional to SO2 amount. The analyser has in-built zero and span calibration facilities. The periodic span calibrations are carried out using permeation tube traceable to NIST standard for every two months and zero calibrations are done every week. Air for zero calibrations is obtained by passing ambient air through activated charcoal.

NOx (NO + NO2) mixing ratios are measured using Thermo Scientific – USA (model 42i) NOX analyser during the study period. Accuracy of the analyser is 0.08 ppbv with a response time of 5 minutes. The zero calibration checks are done on a weekly basis while span calibration checks are done on a monthly basis. The detailed description of the NOX analyser and its calibrations can be found in Renuka et al. (2014).

Black carbon (BC) measurements are done using Magee Scientific Aethalometer (model AE-31) at Gadanki. It has seven lamps of wavelengths 370, 470, 520, 590, 660, 880 and 950 nm. The ambient air sample is drawn into the instrument with a flow rate 2.9 LPM and allowed to flow through a quartz fiber filter which collects the aerosols on it. It is illuminated by the lamps and the attenuated light is measured by the detectors. Amount of absorption of the light is converted to the concentration of aerosols loaded on the filter using Beer–Lambert’s law. The concentration measured at 880 nm wavelength is considered as BC. The instrumental error in BC measurements is estimated to be less than 10% (Gadhavi et al. 2015).

2.2 Satellite measurements of SO2

In addition to surface SO2 measurements from Gadanki, OMI (Ozone Monitoring Instrument) satellite data of columnar SO2 concentrations are also used in this analysis. The satellite data are obtained from the web-site, http://giovanni.gsfc.nasa.gov/giovanni/. The Ozone Monitoring Instrument (OMI) is a sensor onboard NASA Aura Satellite, which is in polar sun-synchronous orbit with nadir measurements. The spatial resolution of the sensor is 13 × 24 km2 which are regrided to 0.25° × 0.25° global grid as a Level 3 product (Krotkov et al. 2015). Local overpass time of the satellite at the equator is 13:45 hrs. The SO2 estimates from OMI are based on radiation measurements in ultraviolet (UV) parts of the spectrum (Li et al. 2017). For each OMI scene, four values of columnar SO2 concentrations are estimated using four different assumed vertical profile of SO2 with centre of mass altitude ranging from 0.9 to 18 km. The values used in the present work are for the profile with centre of mass altitude at 0.9 km.

3 FLEXPART model

FLEXPART (FLEXible PARTicle dispersion model) is a Lagrangian particle dispersion model used for calculating the source–receptor relationship of SO2 with Gadanki as a receptor site. The source–receptor relationship is also called potential emission sensitivity (PES) since it provides sensitivity of mixing-ratios at receptor site for emissions in different regions. Essentially, it is the average of time spent by trajectories in a given grid-box. The model tracks the particles in three-dimensional wind fields with options for both forward and backward modes and includes diffusion, turbulence and convection. Backward model runs are specifically very useful to obtain source–receptor relationships where number of receptor sites are small (Seibert and Frank 2004). We have used the model runs in backward-in-time (also known as retro-plume) mode to find the sources of SO2 being observed at Gadanki. The model set-up is summarised in table 1. The model is run for 10 days in backward mode for each day of the year 2011. NCEP-GFS-FNL meteorological data of resolution 1° × 1° are used for model simulations. Dry deposition of SO2 is simulated by setting the parameters such as diffusivity ratio (D) as 2, Henry’s constant as 100,000 units and reactivity (f0) as 0. Due to the specific geography, the wet deposition plays minor role at Gadanki (Gadhavi et al. 2015). Berglen et al. (2004) using OsloCTM2 chemical transport model have found that wet deposition contributes 1.7% of SO2 loss globally. The major removal process of SO2 from the atmosphere is reaction with OH to form H2SO4 which is readily converted into sulfate aerosol particles. In the present study, simulations are carried out with and without OH reactivity in the model to understand the role of OH field on SO2 concentrations at Gadanki. OH fields are taken from Bey et al. (2001). SO2 losses during transport other than reaction with OH radical and physical losses are not considered in the model.

Table 1 FLEXPART model set-up for retro-plume runs of SO2 from Gadanki.
Full size table

4 Emission inventory

The ECLIPSE version 5 emission inventory is used to simulate the SO2 mixing-ratios at Gadanki. The ECLIPSE (Evaluating the CLimate and air quality ImPacts of Short-livEd pollutants) global emission inventory has been developed using the GAINS Model (Greenhouse gas – Air pollution Interactions and Synergies Model; Amann et al. 2011). The inventory includes sources such as thermal power stations, residential combustion, transport, shipping, large combustion installations, industrial processes, waste and open burning of agricultural residues (except open forest fires). Emissions from forest fires are included from GFEDv3 (Global Fire Emissions Database; van der Werf et al. 2010). The inventory contains current as well as future projections for the emissions. In the present study, emissions for year 2010 are used. The inventory is available at 0.5° × 0.5° resolution which is reduced to 1° × 1° resolution by averaging. Emission fluxes of SO2 from the ECLIPSE-v5 inventory are shown in figure 1. The hotspots of SO2 with emissions greater than 100 kt yr−1 (kilotons) in the figure are mainly locations of major power plants in India and neighbouring countries. Total emissions of SO2 in India in this inventory are estimated to be 8.473 Tg yr−1 of which 51% is from energy sector, 40% is from industries and 7% is from domestic emissions. Transport sector contributes very little to the total emissions (0.9%) whereas biomass burning including forest fires contributes 0.85%.

Figure 1
figure 1

Total emissions of SO2 from ECLIPSE-v5 + GFEDv3 emission inventory for the year 2010. Map boundaries are for cursory region identification and may not be accurate.

Full size image

5 CAMS reanalysis data

One of the service products of Copernicus Atmospheric Monitoring Service (CAMS) is global reanalysis concentrations of various reactive gases including SO2. The CAMS reanalysis is produced using 4DVar data assimilation in CY42R1 of ECMWF’s Integrated Forecast System (IFS) and available through ECMWF data archive (Hennermann and Guillory 2018). The chemical mechanism of the model is documented in Flemming et al. (2015, 2017) but with few updates as listed on the website (Hennermann and Guillory 2018). It uses the MACCity anthropogenic emissions (Stein et al. 2014) and GFAS v1.2 fire emissions. The SO2 mixing ratios are available as profiles over 25 pressure levels and at 3 hourly time intervals over 1.0° × 1.0° lat.–long. grid over the globe. The SO2 values at 1000 mbar pressure level are compared with the ground measurements.

6 Results and discussions

6.1 SO2 mixing ratios

Frequency distribution of SO2 mixing ratio at Gadanki is shown in figure 2. Ninety percent of the observed values are less than or equal to 0.6 ppbv. There are not many SO2 observations from rural locations in India to compare against our values. Reddy et al. (2008) have measured CO and SO2 concentrations over part of peninsular India including one day at Gadanki in February 2008 during a field campaign known as ISRO-GBP Land Campaign-I. They have reported a value of 2.4 ppbv for SO2 mixing ratio at Gadanki with a median mixing-ratios 2.3 ppbv over 22 stations. This is quite high compared to our observations, which is 0.3 ppbv median for February. Overall, the SO2 mixing-ratios over Gadanki are indicative of its rural characteristic and absence of any strong influence of upwind cities.

Figure 2
figure 2

Frequency distribution of SO2 mixing ratios over Gadanki during study period.

Full size image

The in-situ SO2 observations in India are very few and exist mostly for urban areas. Monthly averages of sulphur dioxide concentrations over Delhi during 1998 were 23 ± 7 μg m−3 (Aneja et al. 2001), however, lower values of around 10.9 μg m−3 were reported during 2000–2006 (Kandlikar 2007). Kandlikar (2007) have also reported 33% decrease in SO2 concentration during 2001–2006. Mallik and Lal (2014) have also reported a decreasing trend in SO2 concentrations over Delhi during 2005–2009. They further reported that among six major cities (Delhi, Jodhpur, Durgapur, Kolkata, Guwahati, and Nagpur) in IGP (Indo-Gangetic Plane) region, Kolkata has the highest recorded maximum mixing-ratios of 13.66 ppbv and Jodhpur has the lowest recorded maximum of 3.4 ppbv. The SO2 mixing-ratios reported over Bhubaneswar (a major city in Odisha – an eastern state of India and in south of Kolkata) are 1.7–3.2 ppbv (Mallik et al. 2019). Bhubaneswar has significant regional source like power plants and steel industries in north and north-west of it. However, the reported values of SO2 mixing-ratios over Ahmedabad – a major city in western India are relatively low 0.95 ± 0.88 ppbv in spite of its huge population (more than 5 million) and high industrialisation (Mallik et al. 2012). Unlike urban sites, Naja et al. (2014) have found a strong seasonal variation in background SO2 concentrations with high values during pre-monsoon months and low during winter months at Nainital, which is a high-altitude site. This is because of the location of the site is higher than the boundary-layer height during the winter. Also, Naja et al. (2014) have found that MOZART-4 model simulations are under-predicting the observed SO2 values during pre-monsoon season over Nainital.

Diurnal variation of SO2 is shown in figure 3 where the boxes represent 25–75 percentile range, diamonds represent the mean values and horizontal lines inside boxes represent median values. Hourly mean varies between 0.1 and 0.3 ppbv with high values during day time and low values during late night hours. The skewness of the hourly median values toward low mixing ratios is due to the fact that the range is bounded by 0 at lower side and the high values are too infrequent. Nevertheless, a diurnal variation of SO2 mixing ratios is visible. The diurnal variation with high values in afternoon and low values in night along with skewness of the values towards lower mixing ratios suggest near absence of continuous localised source. If there was a significant localised source, then the boundary layer variation would have caused higher values in night-time and low values in afternoon due to better ventilation and higher OH radical concentration during afternoon. Daily mean of SO2 mixing ratios (upper panel) from January 2010 to April 2012 along with OMI SO2 values (lower panel) obtained from OMI satellite sensor are shown in figure 4. OMI values are in DU (Dobson Unit) and surface SO2 values are in ppbv. Daily mean of surface SO2 varied from 0.2 to 2.6 ppbv during the study period. Satellite-based OMI SO2 values ranged from 0 to 2.3 DU. Though the SO2 mixing ratios are low at Gadanki, it has discernible seasonal cycle with high values during pre-monsoon months (January–May) and low values during monsoon months (June–December). Both satellite and ground-based measurements show a similar seasonal variation.

Figure 3
figure 3

Diurnal variation of surface SO2 mixing-ratios observed at Gadanki for the period January 2010–April 2012. Boxes represent 25–75 percentile range, diamonds represent the arithmetic mean and horizontal lines inside boxes represent median.

Full size image
Figure 4
figure 4

Daily mean of surface SO2 (upper panel) and OMI columnar SO2 (lower panel) at Gadanki for the period January 2010–April 2012.

Full size image

Seasonal variation of surface SO2, NO2 and BC for the study period is shown in figure 5. All these pollutants showed high concentrations during January to May. There are no local sources of SO2 near Gadanki except the highway passing near-by which has a low traffic volume. The high SO2 mixing-ratios during January to May are mainly because of transport of SO2 from regional sources to the observational site. Gadanki has a prolonged rainy season from June to December and the low mixing-ratios of SO2 during June–December is due to removal of SO2 by reaction with available high concentration of OH radical during the season. Gadhavi et al. (2015) have discussed the seasonal variation of BC at Gadanki and shown that it is because of combined effect of change in source region from southwest to north India as well as increase in agricultural waste burning and forest fires in Spring season. The monthly mean variations in NO2 mixing-ratios over Gadanki are very low and hence its seasonal variation cannot be brought out confidently, nevertheless the increase in NO2 during October coincides with change in source region from Arabian Sea to south-central India (figure 7j, k). Similarly, SO2 mixing-ratios are also low and the change in source region should have affected them, but the SO2 mixing-ratios remain low in October–November and only in December we see a slight increase. Unlike, NO2 which has distributed sources like vehicular traffic, the major sources of SO2 are power plants and it takes time before the source region significantly overlaps with the power plant locations as shown in figure 7. More onto this is discussed in the next section.

Figure 5
figure 5

Monthly mean variation of SO2, NO2 and BC observed at Gadanki for the period January 2010–April 2012.

Full size image

Monthly averages of OMI SO2 for entire India for the period 2008–2012 along with power plant locations are shown in figure 6. The symbols in the figure show power plant locations and emission capacity as – blue diamond for 50 kt, red triangles for 100 kt and magenta squares for 200 kt of SO2 emissions (Garg et al. 2002). High amounts of SO2 is found near and around the major power plant locations in eastern India (Chattisgarh, Odissa, and Jharkhand). A significant amount of SO2 is emitted from the power plants located in Tamil Nadu in South India. High concentrations are observed during January–May and October–December months in OMI SO2. The observed low concentration of SO2 during the monsoon period is due to removal of SO2 by reaction with OH radical.

Figure 6
figure 6

(a–l) Monthly mean OMI SO2 for the period 2008–2012 along with power plant locations and their annual emission fluxes.

Full size image

6.2 Model results

FLEXPART model-runs in backward mode give the source–receptor relationship (also known as potential emission sensitivity – PES) of given species. Monthly average PES maps of SO2 obtained from FLEXPART for the year 2011, over-plotted with power plant locations are shown in figure 7. PES value of a grid-box is the average time spent by the virtual air parcel in that grid box during the simulations. When the PES value is multiplied by emission flux in a corresponding grid box, it represents the contribution of the sources in that grid box to the observed concentrations at the receptor. Integrating the contribution at all the grid points represents expected concentration at the receptor site. The more information of PES is available in Seibert and Frank (2004).

Figure 7
figure 7

(a–l) FLEXPART simulated potential emission sensitivity (PES) of SO2 at Gadanki for different months of the year 2011. (Unit of the colour-bars is number of seconds).

Full size image

Comparison of daily mean modelled and observed SO2 values for the year 2011 are shown in figure 8(a). Though, all the three, viz., the in-situ observations, CAMS and FLEXPART models show seasonal variation, high SO2 values during winter months which is seen in the models is absent in the observations. Same is observed for the year 2010 (figure not shown). As mentioned before, levels of SO2 mixing ratios observed at Gadanki have two distinct seasons or phases – one from January to May where relatively high SO2 mixing ratios are observed with high day-to-day variability. The other phase is from June to December when the values and day-to-day variability are low. CAMS and FLEXPART models capture day-to-day variability very well until August, but then the values start rising in models, but they remain low in the observations. Also, there is a large difference in the observed and modelled SO2 mixing ratios. CAMS SO2 mixing ratios are significantly higher than both observations and FLEXPART. This is surprising since the FLEXPART does not include many of the chemical loss processes responsible for SO2 removal which are present in CAMS. The difference between modelled and observed SO2 can be due to two reasons, one possibility is higher emission fluxes than actual and the other possibility is incorrect vertical distribution of SO2 in the models. We have used ECLIPSE-v5 emissions of SO2 for FLEXPART simulations, whereas CAMS is using MACCcity-RCP8.5 emissions of SO2. Venkataraman et al. (2018) have showed that ECLIPSE-v5 emission estimates of SO2 from India are about 18% higher compared to recently developed regional emission estimates of SO2. Also, the difference between RCP8.5 emission inventory of India and ECLIPSE-v5 inventory of India is less than 4% with RCP8.5 emission fluxes being smaller. Hence, if the incorrect inventory was the primary reason, then one would expect FLEXPART SO2 comparable or higher than the CAMS output. However, FLEXPART SO2 values are comparable to observed values during February to May, which lead us to believe that error in emission inventory may not be the primary cause of positive bias in the model values. When the vertical transport or convection is not correctly accounted for in a model, it results in high concentration near the surface and low concentrations at higher altitude, resulting in a positive bias near the surface and a negative bias at higher altitude. Unlike stations in plains, Naja et al. (2014) observed that MOZART-4 model underestimated the SO2 mixing ratios at Nainital, which is a high-altitude site in the Himalayas. We believe that the incorrect vertical distribution is a primary cause of positive bias in the model SO2 values. There have been several observations showing concentrations higher than expected at higher altitude for various pollutants over India (Gadhavi and Jayaraman 2006; Babu et al. 2011; Vernier et al. 2018). This suggests that proper validation of models including emission inventories and vertical transport should be made for a better prediction of the vertical distribution of pollutant species.

Figure 8
figure 8

(a) Observed and modelled daily mean SO2 mixing-ratios at Gadanki for the year 2011. (b) Observed and modelled annual, Jan–May and Jun–Dec mean SO2 mixing-ratios for the year 2011.

Full size image

While the CAMS SO2 values are positively biased in all the seasons, the FLEXPART values are comparable to observations during pre-monsoon but have higher values in winter. This may be due to long range transport of SO2 and issues in vertical distribution as mentioned before. During monsoon months, source region is confined to a small nearby region in south India (more about it in next section), but during winter the source region is whole of north India. If there is a long-range transport of SO2, and the plume remains confined at higher altitudes, then it will not be observed in ground-based measurements, but models as mentioned before may incorrectly distribute them resulting in higher values near surface. Overall the annual mean observed SO2 mixing ratios are overestimated by a factor of 2 by FLEXPART with ECLIPSEv5 inventory and a factor of 7.8 by CAMS (figure 8b).

Figure 8(b) shows comparison of observed and modelled SO2 mixing ratios for annual averages and two periods Jan–May and June–December. The figure also shows FLEXPART values with and without OH chemistry. SO2 is converted to H2SO4 in the presence of OH. OH, chemistry in FLEXPART model is simulated using monthly OH fields obtained from GOES-Chem model (Bey et al. 2001). Overall, there is a 34% decrease in annual average of simulated SO2 mixing-ratios due to OH chemistry. Incorrect OH values in models could also be a reason for positive bias observed at Gadanki.

6.3 Source apportionment using observations

SO2 is emitted into the atmosphere from power plants, industries and transport-vehicles as well as biomass burning. The amount of SO2 released in the atmosphere depends on the type of fuel and temperature at which it is burnt. The ratio of SO2 and NO2 gives the information on sources of SO2. Previous studies have found that power plant and industrial emissions have the ratio of SO2 to NO2 greater than 0.5, whereas transport and biomass burning emissions have the ratio less than 0.5 (Naja et al. 2014; and references therein). The observed ratio of SO2 to NO2 at Gadanki for the year 2011 is shown in figure 9. The ratio varied between 0.1 and 2.5. The ratio is >= 0.5 for most of the days during January to May, which suggests that the sources of high SO2 values during this period are coal burning locations probably power plant/industrial emissions. This is consistent with the analysis using FLEXPART shown in figure 7. However, it should be remembered that the high SO2 values observed at Gadanki are still significantly lower than the safe limit (19 ppbv) prescribed under National Air Quality Standard (CPCB 2009).

Figure 9
figure 9

Daily mean ratios of SO2 to NO2 mixing-ratios at Gadanki for the year 2011.

Full size image

6.4 Source apportionment using model

We have divided the data-set into two parts – one from January to May when SO2 mixing ratios are relatively high and have higher day-to-day variability and two from June to December when SO2 mixing ratios are relatively low and have lower day-to-day variability to understand the cause of high SO2 mixing ratios during pre-monsoon season.

The monthly PES maps of SO2 for Gadanki as a receptor site are shown in figure 7. The high PES values during January to May are located in southern and eastern India. The same can be observed in the OMI SO2 maps shown in figure 6. This covers the down-wind regions of the power plants located in southern India and big cities like Chennai, Bangalore, Nellore, Vijayawada, Visakhapatnam, etc. The PES values are high around the power plants located in Tamil Nadu, Karnataka, Andhra Pradesh, Telangana, Odisha and parts of Chhattisgarh states which lead to high SO2 mixing-ratios at the site during this period. Whereas, power plants are mostly away from the high PES values during July–December. The change in source region as captured by shift in high PES value regions nearer to power plants in January to May is the primary reason of high SO2 values during January–May. Even though, high values of OMI SO2 can be seen in the downwind regions of power plants during October and November, the SO2 values at surface at Gadanki are low in this period. October and November months are seasons of north-east monsoon at Gadanki with high moisture levels and SO2 is removed by reaction with OH during the transport.

6.5 Long-range transport of SO2

Unlike black carbon particles (Gadhavi et al. 2015), SO2 burden at Gadanki is mostly from sources found in southern part of India. However, we came across an event of long-range transport of SO2. High OMI SO2 values were observed at Gadanki during June 2011 which is a monsoon month. Generally, the SO2 values are low during this month. However, on 20th June 2011, a very high SO2 concentration with a value of 7.6 DU was observed which gradually decreased in subsequent days (figure 4, lower panel). The vertically integrated PES values from the FLEXPART model for 19th and 20th June 2011 are shown in figure 10. On both the days, air masses are found to arrive from East African region. The Mount Nabro volcano located at latitude of 13°22′N and longitude of 41°42′E in Ethiopia in African region erupted on 13th June 2011 (shown with red dot in figure 10). The volcanic plume arrived at Gadanki on 20th June 2011. Similar plume evolution can also be seen in OMI SO2 values and is shown in figure 11. Figure 11(a) shows SO2 concentrations on 13th June 2011 – the day when Mt. Nabro erupted. Figure 11(b–i) depict subsequent plume evaluation as observed in the satellite data. Though the satellite values were high between 20th and 25th June, ground-based observations did not show similar increase possibly as the transported SO2 plume remained confined at higher altitudes without mixing with lower atmosphere over the observation site.

Figure 10
figure 10

Total column sensitivity maps calculated using FLEXPART for (a) 19th June 2011 and (b) 20th June 2011. Red dot represents the location of Mt. Nabro Volcano in Ethiopia.

Full size image
Figure 11
figure 11

(a–i) OMI SO2 values from the day of Mt. Nabro eruption to the day it reaches Gadanki, India.

Full size image

7 Conclusions

SO2 mixing-ratios observed over Gadanki during January 2010 to April 2012 are found to be very low. The daily mean values varied from 0.2 to 2.6 ppbv. Both ground-based and satellite-based observations of SO2 showed a clear seasonal cycle with high values during pre-monsoon months and low values during monsoon months. Though the values of SO2 are very low representing a typical rural environment, on some days, relatively higher values (>1 ppbv) are observed during January to May. During this period, SO2 to NO2 ratios are greater than 0.5 which suggest that the source is a high temperature source like power plants. Using FLEXPART model, it is shown that these high values are due to the transport of SO2 from the power plants located in Neyveli in Cuddalore district and Mettur in Salem district of Tamil Nadu State of India and occasionally from power plants located in Karimnagar district of Telangana and Krishna district of Andhra Pradesh. CAMS and FLEXPART models are overestimating the SO2 mixing-ratios for Gadanki. The overestimation is severe during winter months. Overall annual mean SO2 mixing ratios were overestimated by a factor of 2 in FLEXPART and 7.8 in CAMS. We suggest that the overestimation is possibly due to incorrect vertical distribution of SO2 in the models. In one instance, SO2 plume from the Nabro volcanic eruption in East Africa is found to reach Gadanki and caused a significant increase in columnar SO2 concentrations but near surface mixing-ratios were not found to be affected. This is another situation which suggests the importance of measurements of vertical distribution of pollutants over India.