The Kathmandu Valley is an oval-shaped urban basin located in the central Himalayan foothills between the IGP and the Tibetan Plateau (Fig. 1). The valley is surrounded by mountain peaks with an altitude ranging from 2000 to 2800 m above mean sea level (a.m.s.l.) and five mountain passes (Nagdhunga, Bhimdhunga and Mudkhu Bhanjyang in the west and Sanga and Nagarkot in the east) with an altitude ranging from 1500 to 1550 m a.m.s.l., and the outlet of the Bagmati River is in the southwest corner of the valley. The flat base area of the Kathmandu Valley is about 340 km2 with a mean elevation of about 1300 m a.m.s.l. There is no river inlet into the Kathmandu Valley and only one narrow river outlet (Bagmati River) on the southwestern side. The spatial extent of the valley is about 25 km east–west and around 20 km in the north–south direction. The measurement sites located in the Kathmandu Valley during the SusKat field campaign are depicted in Fig. 1b. In this study observations of O3, CO and meteorological parameters made at Bode (27.68∘ N, 85.39∘ E, 1344 m a.m.s.l.) and at two sites in India, namely, ARIES, Nainital, a high-altitude site located on a mountaintop (29.36∘ N, 79.45∘ E, 1958 m a.m.s.l.), and Pantnagar, located in the Himalayan foothills on the IGP (29.0∘ N, 79.5∘ E, 231 m a.m.s.l.), are discussed.

Two observations sites in India, i.e., Nainital and Pantnagar, represent cleaner Himalayan and polluted IGP environments, respectively. Surface ozone levels at Nainital are found to be driven mainly by transport of anthropogenic emissions from the IGP region (Kumar et al., 2010, 2011), while those at Pantnagar are mostly controlled by local emissions (Ojha et al., 2012). In the previous studies, the pollutant levels in the Kathmandu Valley were reported to be primarily influenced by the local emissions and the unique meteorology in the region (Panday et al., 2009; Putero et al., 2015).

Surface ozone measurements are conducted using two types of analyzers, i.e., Teledyne M400E (at Bode and Pantnagar) and Thermo model 49i (at Nainital). The observation principle of both the instruments is based on the commonly used technique of attenuation of UV radiation (∼254 nm) by ozone molecules. These instruments are regularly subjected to zero and span tests using an internal ozone generator, and ozone observations from both the instruments are also intercompared by running them side by side and using a common inlet. Further details of such intercomparisons are reported in Sarangi et al. (2014).

CO measurements are also conducted using two types of analyzers, i.e., Horiba APMA-370 (at Bode and Pantnagar) and Thermo 48i (at Nainital). The CO instrument, at Bode, was deployed for the first time in field after factory calibration from the manufacture. Nevertheless, both CO instruments were intercompared using a common inlet prior to the campaign. The correlation coefficient among CO mixing ratios measured by the instruments is estimated to be ∼0.9 with a slope of 1.09. CO instruments are based on the commonly used method of infrared (IR) absorption by the CO molecules at 4.6 µm. Regular zero checks and span checks for CO instruments are performed using a primary calibration mixture from Linde, UK (1150 ppbv; Sarangi et al., 2016) and secondary gas from Chemtron Science Laboratories (1790 ppbv). Multipoint calibrations (ultrapure gases) are also carried out in different observational ranges using a zero air generator (Thermo model 1160) and a dynamic gas calibrator (Thermo model 146i) (Sarangi et al., 2014). The meteorological measurements at Bode are performed using an automatic weather station (Campbell Scientific, UK).

Both the O3 and CO instruments were installed on the fourth floor of a building in Bode facing the eastern side of the Kathmandu Valley (Fig. 2; refer to Sarkar et al., 2016, for site description). The sampling inlet for these instruments was placed at the top of the building and Teflon (TFE) tubes were used for the air intake. The O3 and CO instruments at Nainital and Pantnagar were placed in the atmospheric science building at Manora Peak (refer to Sarangi et al., 2014, for site description) and at the College of Basic Sciences and Humanities (CBSH), G. B. Pant University of Agriculture and Technology (GBPUAT), at Pantnagar (refer to Ojha et al., 2012, for site description). Observations at Bode are in Nepalese standard time (NST), which is 05:45 h ahead of GMT, and observations at Nainital and Pantnagar are in Indian standard time, which is 05:30 h ahead of GMT.

A total of 16 air samples are collected at Bode from 30 December 2012 to 14 January 2013 with the frequency of one sample per day. These air samples are collected at 14:00 NST (two samples at 12:00 NST) when the boundary layer is fully evolved and the air is well mixed. Air samples are collected at a pressure of 1.5 bar and analyzed for ethane, ethene, propane, i-butane, n-butane, acetylene and i-pentane using a gas chromatograph (HP 5890 II) equipped with a flame ionization detector and a PLOT column of KCl / Al2O3. Helium is employed as the carrier gas and H2 and zero air are used for flame. Air samples are also analyzed for CH4 and CO using another gas chromatograph (Varian Vista, 6000, USA) and employing a molecular sieve 13 times with a packed column (4 m). CO is measured by converting in to CH4 using a Ni catalyst heated to about 325 ∘C. A standard mixture from Intergas (International Gases and Chemicals Ltd.), UK, traceable to National Physical Laboratory (NPL), UK, is employed for the calibration of non-methane hydrocarbons (NMHCs). Gases from NIST, USA, and Linde, UK, are used for the calibration of CH4 and CO. More details on sample pre-concentration and calibration can be seen in Lal et al. (2008), Mallik et al. (2014) and Sarangi et al. (2016).

In this study, the Ozone Monitoring Instrument (OMI) level 3 daily tropospheric column amount NO2 data product OMNO2d (cloud screened at 30 %) at 0.25∘ × 0.25∘ resolution is used to generate spatial maps during biomass burning periods. This product is based on the radiance measurements made by the OMI instrument in visible (VIS, 405–465 nm) channels. A detailed description of measurement principles can be found in Bucsela et al. (2013). The fire locations during the springtime biomass burning are used from monthly MODIS collection 6, level 2 (combined Aqua and Terra) global monthly fire product mcd14ml at 1 km resolution (Giglio, 2010). For this study fire locations with high detection confidence (> 80 %) are used and a detailed detection principle can be found in Giglio et al. (2003) and Justice et al. (2006). NASA's Atmospheric Infrared Sounder (AIRS) instrument suite measures atmospheric water vapor and temperature profiles on a global scale. At present the operational instruments in this suite consist of a hyperspectral infrared instrument (AIRS) and a multichannel microwave instrument, Advanced Microwave Sounding Unit-A (AMSU-A). Apart from meteorological datasets, these instruments also retrieve the vertical profiles of some of the trace gases such as ozone and CO, and a detailed description of retrieval algorithms is discussed in Susskind et al. (2003, 2006). In this study the vertical profiles of ozone, CO and relative humidity (RH) from the AIRS + AMSU joint data product (AIRX3STD v006) at 1∘ spatial resolution are used to understand stratosphere–troposphere transport phenomena during January–May 2013.

The biomass burning emissions of CO were estimated using Global Fire Emissions Database (GFED) version 4 datasets. These data use satellite information of fire hot spots and vegetation productivity to calculate gridded fire emissions. The data also included a fractional contribution of different types of vegetation to fire emissions. The version 4 dataset has a spatial resolution of 0.25∘, and a detailed description about the dataset and data access could be found from http://www.globalfiredata.org/data.html (last access: 15 October 2017). The Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2) is the reanalysis dataset produced by the Global Modeling and Assimilation Office (GMAO) of NASA. In this study Ertel potential vorticity data from this reanalysis are used and more details can be found in Gelaro et al. (2017).

To understand the history of air masses arriving at Bode, the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Hess, 1998) is used. The high-resolution (0.5∘ × 0.5∘) meteorological fields from the Global Data Assimilation System (GDAS) are used to generate 5-day backward trajectories over Bode. These trajectories are initiated as an ensemble of nine points separated by 0.25∘ around Bode at 1km above ground level for four different months.

WRF-Chem version 3.5.1 with two nested domains was used for this simulation. The coarse domain that encompasses the area between 16–43∘ N and 68–107∘ E was used with 15 km grid spacing, the nested domain that covers the central part of Nepal and the Kathmandu Valley with 3 km grid spacing, and 31 vertically stretched layers from the ground up to 10 hPa (Mues et al., 2018). The physics options used for this study include the Lin microphysics scheme (Lin et al., 1983), the Grell cumulus parameterization (Grell and Dévényi, 2002), the Rapid Radiative Transfer Model (Iacono et al., 2008) for the longwave and Goddard shortwave schemes (Chou and Suarez, 1994), the Yonsei University boundary layer parameterization (Hong et al., 2006), and the MM5 scheme for the surface layer (Jimenez et al., 2012). The initial and boundary conditions for meteorological fields use the ERA-Interim data. Anthropogenic emissions were obtained from the HTAP v2 inventory (http://edgar.jrc.ec.europa.eu/htap_v2, last access: 18 January 2017). Emission maps for CO and NOx are shown in the Supplement (Figs. S1 and S2). The RADM2-SORGAM chemical mechanism is used to represent the gas-phase and aerosol chemistry. The photolysis rates were computed using the Fast Tropospheric Ultraviolet and Visible (FTUV) Radiation Model (Tie et al., 2003; Li et al., 2005). The dry deposition was calculated following the Wesely (1989) resistance method. Biogenic emissions were computed on-line using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) (Guenther et al., 2006). The biomass burning emissions are based on the Fire INventory from NCAR (FINN) (Wiedinmyer et al., 2011).

The Kathmandu Valley is located in the central Himalayan region due north of the IGP and south of the Tibetan Plateau. The valley is influenced by the south Asian monsoon and in general receives most of its precipitation during summer (June–September). The remaining seasons are relatively dry with spring or pre-monsoon seasons (March–May) being the hottest (Panday and Prinn, 2009). Figure 3 shows the diurnal variations in temperature, RH, solar radiation, wind speed and wind direction at Bode during January and April 2013, which are chosen as representative months for winter (January–February) and spring (March–May), respectively. The average daily RH during January (67 %) was higher than in April (62 %), with the diurnal maximum during early morning hours. During January, high RH in early morning hours (92 % during 05:00–07:00) was associated with the foggy conditions during morning hours.

The average temperature and solar radiation were higher in April (20.7 ∘C and ∼800 W m-2) than January (11.4 ∘C and ∼600 W m-2). The diurnal variations in temperature showed the highest values during late afternoons, which was a few hours after the peak in diurnal solar radiation. Solar radiation strongly influences the diurnal variations in temperature and other meteorological parameters by surface heating, causing thermals to rise. The wind speeds were the slowest (< 1 m s-1) during the night and early morning hours, primarily for the easterlies, and the highest wind speeds were observed during middle to late afternoon, primarily for westerlies (4–6 m s-1). These wind patterns were similar in both the months, with April having a slightly longer duration of daytime westerly flow. These wind flows, together with boundary layer dynamics, are responsible for the dispersion and accumulation of pollutants during the course of a day.

The other sites in the Indian region, namely, Pantnagar and Nainital, also show high values of solar radiation and temperature during spring months and the lowest values during the winter season (Ojha et al., 2012; Sarangi et al., 2014; Naja et al., 2016). The Nainital site, which is at a remote mountaintop, experiences moderate northwesterly winds (∼2–3 m s-1) during most of the year, with a prevalence of southeasterly winds during the summer monsoon period. The Pantnagar site is situated in the vicinity of the Himalayan foothills in the IGP and similar seasonal changes in temperature, RH and solar radiation are observed (Ojha et al., 2012). In the summer monsoon season, the lowest levels of O3 and CO are observed due to the arrival of cleaner marine air masses. During winter months, slow winds, poor ventilation and the lowest boundary layer heights are observed, and during the first week of January several cases of widespread fog are also observed.

To understand the wind patterns over Bode, the HYSPLIT model is used to generate backward air trajectories. Figure 4 shows the monthly averaged nine-point trajectories (in different colors) for January, March, May and June around Bode. During the winter and early spring months of January and March, a strong westerly flow is observed, with air masses passing through IGP prior to entering the Nepal region. During late spring (May) these westerly air masses slow down and a slight change in direction is observed as some of the air masses pass through the central part of India. The air masses were observed to be mostly arriving from higher altitudes during winter than during spring (not shown in Fig. 4). Similar behavior of air masses is also observed over Nainital (Kumar et al., 2010). However, during June, the monsoon circulations take over and reversal in wind directions is observed (Fig. 4d). The above features for Bode are similar to those seen over Nainital and Pantnagar. The backward air trajectory analysis for Nainital and Pantnagar shows that long-range transport (westerly wind) is a vital factor during winter. However, air masses mostly circulate over the continental northern Indian region at low altitudes during spring and autumn seasons when local pollution plays an important role (Kumar et al., 2010; Ojha et al., 2012; Naja et al., 2016).

The monthly (January to June) average diurnal variations in ozone at Bode are shown in Fig. 5. The diurnal variations in ozone show higher levels during the daytime. This daytime buildup in ozone is consistently observed throughout the observation period, with relatively less buildup during June due to prevailing cloudy–rainy conditions. Additionally, there are only a few days of observations in June, when the campaign ended. The daytime increment in surface ozone is a typical feature of polluted sites and can be associated with daytime photochemical production of ozone from its precursors in the presence of sunlight (e.g., Kleinman et al., 1994). Model-simulated normalized average diurnal variations also show a clear daytime buildup in ozone during February and May (Fig. S3).

The very low levels of ozone (during winter months) were observed during nighttime, which can be attributed to titration of O3 by NO. The boundary layer measurements made using a ceilometer during the campaign suggest lower heights during nighttime (Fig. 6) with very low ventilation coefficients (VCs) during winter (Mues et al., 2018). The sampling inlet for the gaseous measurements was on the rooftop about 20 m from the ground level; thus the possibility of loss of ozone due to surface deposition in the highly stratified nighttime boundary layer should be much less. Unfortunately, observations of NO and NO2 are not available during the campaign but some information about the levels of these species can be obtained from the previous studies. Pudasainee et al. (2006) measured the NO and NO2 during winter of 2003–2004. Yu et al. (2009) also measured NO, NO2 and HONO during a similar period and showed two peaks in the morning (07:00–08:00) and evening (19:00–20:00) in NO, with maximum levels reaching as high as 60 ppb. Although the results are almost a decade old and the difference in O3 magnitude is also observed (Fig. S4), we expect the increase in NO levels over the valley because satellite-retrieved tropospheric column NO2 shows an increase over this region (Fig. S4).

Just around sunrise, a dip in ozone levels was observed. A similar dip in ozone was observed at an urban site in India (Lal et al., 2000), which is suggested to be due to its reaction with NO and NO2. Since observations of NO, NO2, NO3 and N2O5 are not made during the campaign, we have employed model results to explain this feature. Figure 7 shows model-simulated average diurnal variations in NO, NO2, NO3 and N2O5 during February and May 2013. NO mixing ratios are close to zero during the nighttime because it rapidly reacts with O3 to form NO2, which also explains higher NO2 levels during nighttime. NO3 and N2O5 also show higher levels during nighttime because of the reactions of NO2 with O3 and NO2 with NO3, respectively. The sharp morning increase in NO mixing ratios correlates strongly with the sharp decrease in NO3 and N2O5 mixing ratios, especially during February, indicating that photodissociation of NO3 (λ < 670 nm) and N2O5 (280 < λ < 380 nm) releases NO back into the atmosphere. It is evident that NO3 and N2O5 peak prior to NO and NO2 and later they are removed.

At Bode, higher springtime levels of ozone with a broader peak were observed when compared with winter months; this can be attributed to the increase in incoming solar radiation, which in turn increases the photochemical production of ozone. The role of air masses (local vs. regional contribution) will be discussed later in subsequent sections in which differences in winter and spring variations will be clearer. The nighttime ozone mixing ratios during the campaign are the lowest during the winter season. During this period, lower boundary layer height was also observed (Fig. 6) (Mues et al., 2018). The VC is a measure of transport and/or mixing of pollutants in the boundary layer. Due to very low wind speeds (< 1 m s-1) and shallow boundary layer height during the night, low VC (< 100 m2 s-1) were observed during this season (Fig. S5).

The monthly average diurnal variations in CO showed two peaks, one during the morning and the other in the evening hours (Fig. 5). The monthly mean values of O3 and CO with standard deviation, maximum and minimum are given in Table 1. Averages of both these gases in four time periods are also given in Table 2. Diurnal variation in CO with two peaks is a typical pattern over a polluted site and such variations have been reported at different south Asian urban sites, e.g., Ahmedabad (Lal et al., 2000), Kanpur (Gaur et al., 2014), Pune (Beig et al., 2007), Santiago de Chile (Rappenglück et al., 2005) and Chicago (Pun et al., 2003). The major sources for CO in the valley are vehicular emissions, brick kiln emissions, domestic burning of biofuels for cooking and heating, and garbage burning. Out of these, vehicular emissions, cooking and heating occur largely two times a day, morning and evening, but with some time differences. Additionally, the nighttime and early morning hours are characterized by very slow wind speeds and a shallower boundary layer along with poor mixing.

Note that CO levels during morning peaks are greater than those during the evening. Such higher peaks during morning times than those during evening times have also been observed by Panday and Prinn (2009) and have been explained on the basis of overnight accumulation of CO emissions. During daytime the emissions are countered by the boundary layer evolution and dynamic processes such as flushing of CO and other pollutants by westerly winds (Figs. 1 and 3) blowing eastwards across the mountain passes in the valley throughout the afternoon. The chemical loss of CO via reaction with OH radicals could also contribute to lower daytime CO levels. However, OH measurements are not made during the campaign and it is not possible to confirm this aspect.

The mixing layer starts evolving after sunrise and reaches its peak median values of about 900–1200 m during the winter and spring seasons (Mues et al., 2018). The VC estimated (with 15 m wind data) during the daytime suggested vertical mixing could occur from mid-afternoon until early evening hours when consistently high VC (> 1000 m2 s-1) was observed (Fig. S3). During this period the daily maximum wind speeds (4–5 m s-1 at 15:00–17:00), which were mostly westerly, were observed, and further decreases in the CO levels with daily minimum levels (148–218 ppbv) were also observed during this period. Due to relatively higher wind speeds and associated advection before the evening hours, the CO mixing ratios at 19:00–20:00 are about 30–50 % lower than for the morning hours.

In addition to CO, observations of VOCs made at Bode (Sarkar et al., 2016) also showed similar diurnal variations with two peaks and levels up to about 15 ppbv. Similar to CO, many VOCs (e.g., acetonitrile, benzene, furan) showed higher levels during the morning, when compared to the evening peaks. We also collected one air sample almost every day during 30 December 2012–14 January 2013 and analyzed them for light NMHCs (C2-C5) and for CH4-CO. Average values along with standard deviation, minimum, maximum and number of samples are given in Table 3. Average methane levels for the measurement period at Bode are 2.55±0.12 ppmv, which is much higher than the global average and about 28 % and 27 % higher than measured the Northern Hemisphere background site at Mauna Loa (1.84 ppmv) and at the remote site Mt. Waliguan, China (1.87 ppmv), respectively (www.esrl.noaa.gov, last access: 10 March 2018). The average values of other hydrocarbons varied from about 1 to about 4.4 ppbv. The maximum value is observed for propane (15.48 ppbv) and acetylene (14.35 ppbv). These highest mixing ratios are observed on 7 January 2013.

The mixing ratios of eight light NMHCs at Bode are relatively higher (ppbv) than those observed at Nainital (0.8–2.2 ppbv) and comparable with those at Pantnagar–Haldwani (0.8–3.7 ppbv and 1.6–4.2 ppbv, respectively). Additionally, unlike Bode, none of these sites in India showed NMHC mixing ratios exceeding 10 ppbv. The methane mixing ratio is also higher at Bode (2.55 ppmv) than that observed at Nainital (1.89 ppmv). Figure 8 shows a comparison of the contribution of eight light NMHCs at Bode, Nainital, Pantnagar (including another town, Haldwani) (Sarangi et al., 2016) and Kanpur (Lal et al., 2008). Bode data are for December–January, while data from the rest of the three sites are from December. Composition at Bode shows a difference from that at Indian sites. Propane (20 %) and n-butane (13.5 %) show a greater contribution, while the contribution of i-pentane (4.2 %) is significantly lower at Bode when compared with India sites. A greater contribution of propane and n-butane indicates leakages of LPG in the Kathmandu Valley.

It has been discussed in the previous section that O3 and CO show some contrasting diurnal variations. Here we discuss the correlation between O3 and CO during different months (Fig. 9). The highest negative correlation is seen in the winter period (r2=0.82 in January and r2=0.71 in February) and this negative correlation reduces gradually, with the lowest value in May (r2=0.12). The reduction in the tendency of the negative correlation from January to May could be due to changes in emission patterns and the boundary layer mixing. This is discussed further in the subsequent paragraphs. Hourly average CO levels show a systematic decrease from ∼2300 ppbv in January to about 680 ppbv in June, whereas ozone levels show an increasing tendency. The daytime ozone levels during the spring season are slightly higher (∼62 ppbv) when compared to winter (∼54 ppbv). This springtime increase in ozone levels is also reported by several other studies in the northern part of the Indian subcontinent (Kumar et al., 2010; Ojha et al., 2012; Kumar et al., 2013; Gaur et al., 2014). The higher values of CO during the winter season can be attributed to an increase in its emissions (domestic and garbage burning to keep warm in the winter season) and their inefficient dilution due to poor mixing and a shallower boundary layer (Mues et al., 2018). However, during the spring season, a greater and relatively well-mixed daytime boundary layer leads to lower CO levels. As mentioned previously, the role of OH chemistry could be an additional contributor in lower CO levels in spring. Model-simulated OH levels are found be higher in May, when compared with February. This would suggest great chemical loss of CO in spring (May).

Figure 10 shows daily variations in O3 and CO during four different times, i.e., 03:00–05:00, 07:30–08:30, 13:00–15:00 and 22:00–23:00 h. We consider that 03:00–05:00 and 22:00–23:00 h would provide information for the periods when photochemical production of ozone is absent, while 13:00–15:00 h can be used to understand the behavior during the periods of high photochemical activity and a fully evolved daytime mixed layer. Variations during 07:30–08:30 h will provide information during the morning period. The stable nocturnal boundary layer starts evolving during morning hours and air masses close to the surface begin to mix with air at higher heights. In general, CO levels (blue line) show a decrease from January to June during 03:00–05:00, 07:30–08:30 and 22:00–23:00 h, while they do not show significant changes during 13:00–15:00 h. In contrast, ozone levels (red lines in Fig. 10) increase from January to May–June during all four time periods. The highest noontime ozone level is observed to be about 80 ppbv during January to March, which increases to about 102 ppbv during April–May. The noontime ozone level comes down to about 46 ppbv in June, which is mainly due to the beginning of the monsoon season, which is characterized by the arrival of cleaner air from the oceanic regions of the Bay of Bengal, the Arabian Sea and the Indian Ocean.

The increase in ozone from January to May is rather more during nighttime or early morning hours, when photochemical production of ozone is absent. This suggests an enhancement in the background ozone levels. Hence, the average ozone value (13.1±1.2 ppbv) during the early morning period of 03:00–05:00 h would most likely represent background ozone levels for this region. However, the emissions directly influence the CO levels. Hence, the estimated CO mixing ratio (325.4±98.3 ppbv) during noontime (13:00–15:00 h) could be considered to be background levels for the Bode region. The noontime boundary layer at Bode is assumed to be well mixed and the fast westerly flows across the valley reduce the direct sampling of air masses at Bode from its immediate emission sources scattered in the valley.

Figure 10 also shows correlation between O3 and CO for different time periods. Weak negative correlation is seen during the early morning (03:00–05:00 and 07:30–08:30 h) or nighttime hours (22:00–23:00 h). During nighttime and early morning hours, the lowest boundary layer height (150–200 m) was observed. In the urban or semi-urban environment of Kathmandu Valley, where NOx levels are not lower (Pudasainee et al., 2006), ozone titration takes place throughout the nighttime. Figure 10 also indicates the increase in overnight CO levels, which peaked during morning hours. These contrasting variations tend to show the negative correlations. However, a slight positive correlation is observed around noon (13:00–15:00 h), which is similar to what is generally observed at high-altitude sites (Kajii et al., 1998; Tsutsumi and Matsueda, 2000; Naja et al., 2003; Sarangi et al., 2014) and cleaner sites (e.g., island sites; Pochanart et al., 1999).

Apart from O3 and CO observations at Bode, simultaneous observations of these two gases were also made at the central Himalayan site in India (Nainital) and at a site in the Himalayan foothills in the IGP region (Pantnagar) and are discussed in this section. The wind patterns over Pantnagar and Nainital, despite the difference in altitude, are mostly northerly or northwesterly during winter (Kumar et al., 2010; Ojha et al., 2012; Sarangi et al., 2014). Further, both sides receive polluted air masses from the IGP during spring. Whereas the cleaner marine air masses arrive at these sites during the summer monsoon season. Average diurnal patterns in O3 and CO mixing ratios are similar at Bode and Pantnagar, with twin peaks in CO and high daytime levels of O3. However, different variations (Fig. 11) are observed at Nainital (green line), which being a remote high-altitude site does not show any daytime photochemical buildup or nighttime loss in ozone. Further, the daytime ozone levels at Bode are higher than those at Pantnagar during winter while they are comparable during spring. Additionally, CO levels are also higher at Bode than those at Pantnagar during winter (Fig. 11). A comparison of surface ozone measurements at Kanpur (India) showed a relatively better agreements with Pantnagar (India) while measurement at Paknajol (Nepal) showed better agreement with Bode (Nepal) during both seasons (winter and spring) (Fig. 11), indicating the differences in emission sources of the Kathmandu Valley and the IGP.

The changes in ozone increase and decrease rates (ppbv h-1) are analyzed for all five sites. Generally, the ozone increase and decrease rates are nearly symmetric during morning and evening at an urban site. However, it is asymmetric, with slower changes occurring during the afternoon–evening at rural or semi-urban sites (Naja and Lal, 2002). Ozone production is strongly dependent on the amount of precursor gas and available sunlight. Conversely, evening ozone loss depends mainly upon its titration with NO, in addition to surface deposition. This rate of change of ozone during morning and evening hours has been used as an indicator of chemical environment (rural or urban) over a site (e.g., Naja and Lal, 2002). Below, we discuss calculated ozone increase and decrease rates.

Figure 12 shows that the wintertime rate of ozone increase in the morning hours is much higher at Bode (about 17 ppbv h-1) when compared to Pantnagar (about 9 ppbv h-1). This suggests a more rapid ozone buildup at Bode than at Pantnagar. In contrast, the ozone decrease rate is lower at Bode (5–6 ppbv h-1) when compared to the decrease rate at Pantnagar (about 14 ppbv h-1) during spring. This suggests rather slower ozone loss at Bode via NO titration, indicating a somewhat less polluted environment in Bode during spring. However, this does not necessarily mean that NOx emissions are lower in the Kathmandu Valley. NO levels are reported to be as high as 60 ppbv (Yu et al., 2009). Another process driven by diurnal variations in winds could lead to slower evening ozone loss rates. Faster afternoon westerly winds flush the pollutants out of the valley every day, leaving less NOx to titrate ozone during evening hours. In contrast, slower winds at night allow overnight accumulation of precursor gases in the nocturnal boundary layer of the Kathmandu Valley that in turn can potentially contribute to next-morning ozone buildup. Similar to the diurnal variations in average ozone, diurnal patterns in ozone change rates are similar at Pantnagar and Kanpur.

Analysis of ozone observations including backward air trajectories at Nainital in the central Himalayas shows that the major role of regional and local pollution is in spring when regionally polluted ozone levels are estimated to be 47.1±16.7 ppbv (Kumar et al., 2010). During spring, net ozone production over the northern Indian subcontinent is estimated to be 3.2 ppbv day-1 in regionally polluted air masses in spring but no clear buildup is seen at other times of year. The role of long-range transport is shown to be important in winter, with a contribution of about 8–11 ppbv of ozone.

Shorter duration of solar heating during winter leads to weaker dynamical processes including convective mixing of pollutants, which in turn confines the pollutants close to the surface. Additionally, the Kathmandu Valley is isolated inside the Himalayas and the only way for pollutants to reach the area is via either upslope flow of polluted air masses through the mountain valleys or arrival of polluted regional air masses from the air aloft. Thus, the wintertime O3 and CO mixing ratios at Bode are unlikely to be influenced by IGP pollution. Further similar trapping of pollutants during winter is also reported by previous studies performed over this region (Panday and Prinn, 2009). However, intense heating and stronger convective mixing could induce the IGP outflow to influence this valley region during spring. Springtime ozone enhancement, due to IGP outflow, in the tropical marine region surrounding India has already been observed (Lal et al., 2013).

Every year northern Indian biomass burning emits large quantities of trace gases and aerosols and significantly affects the regional distribution of several trace species (Kumar et al., 2011; Sinha et al., 2014; Putero et al., 2014; Bhardwaj et al., 2016; Kumar et al., 2016). These studies showed the enhancement in O3 and CO levels due to crop residue and forest fire burning in the IGP during pre- and post-monsoon seasons. To investigate the influences of biomass burning on the surface ozone and CO mixing ratios over Bode, a time series of high-confidence (confidence > 80 %) fire counts over the northern Indian subcontinent (25–35∘ N, 70–95∘ E) were analyzed. Based on MODIS fire counts, the fire activity period is chosen to be 3 April to 31 May 2013.

A high-fire-activity period (HFAP) is defined when a 3-day running mean of fire counts exceeds the median fire counts during the fire activity period (Kumar et al., 2011). The low-fire-activity period (LFAP) is defined as 1–31 March when very low fire counts are observed over the northern Indian subcontinent. The changes in surface ozone and CO mixing ratios during these two periods (HFAP and LFAP) are shown in Fig. 13d, f. The average fire emissions over the Kathmandu region (27–28∘ N, 85–86∘ E) using the GFED v4.0 emission inventory are also calculated and shown in Fig. 13b. Since the HFAP is almost 4 weeks long, two different peaks in ozone and CO mixing ratios during April and May are selected (see highlighted peaks in Fig. 13). During both the periods a prior increase in fire counts was observed, which is followed by higher levels of ozone and CO mixing ratios. Another reason to separately study these cases is fire locations during these two periods. A high CO and O3 cases in April are shown to be induced by the fires that are mostly located in the Nepal region (dark red), while fire activity was very low in the northern part of the IGP in April (Fig. S6). Conversely, during the May event (orange), a high increase in fire counts (∼2.5-fold) is observed over the northern Indian subcontinent during the HFAP (Fig. S6) and very little fire was seen in the Nepal region. Therefore, these two cases could be studied separately to assess impacts of biomass burning over Bode and other sites. The changes during HFAP, LFAP and the above two event cases in April and May are shown in Table 4.

During the first half of April daily averaged CO mixing ratios of 861 ppbv were recorded, which were about 200 ppbv higher than the daily averaged CO levels for April. Ozone mixing ratios during this period were about 9 ppbv higher than the daily averaged ozone levels for April. The spatial maps of MODIS fire counts during the period were also analyzed (Fig. 14, top panel) and an increase in fire counts during the period was observed. Based on MODIS fire counts and O3 and CO time series at Bode, we classified 20–25 March as a pre-event period and 3–6 and 11–16 April as HFAPs. During the first HFAP, increases of 55 % and 15 % in surface ozone and CO mixing ratios were observed. The fire maps during this period also indicate that the majority of the fires were occurring near Bode in the Nepal region and are shown in orange in the spatial map (Fig. 14, top panel). The AIRS-retrieved CO mixing ratios at 925 and 850 hPa also indicate an increase in CO levels near Bode (indicated by black star); however no similar increase is observed in CO levels near Pantnagar (or IGP). The majority of air masses at 1 km a.g.l. (above ground level) prior to arriving to Bode (or Nainital) are westerly and pass through the IGP and northern Indian subcontinent. Since there is almost no fire activity near the Punjab region, we argue that these nearby fires and their associated polluted plumes around the Kathmandu Valley could affect the levels at Bode. Further, GFED biomass burning emissions were used to differentiate different types of biomass burning emissions during this period. During both of the events, the forest fire emissions dominated the total biomass burning emissions over the Kathmandu region. The major land use types in black boxes indicated in Fig. 14 are agricultural land or forests. The meteorological data observed during this period did not show any noticeable differences and boundary layer height during these events was close to monthly average values. Therefore, we can conclude that forest fires occurring in the nearby Kathmandu Valley were responsible for high levels of pollutants during this period.

The second event took place in the first week of May when simultaneous increases in surface O3 and CO levels at Bode, O3 levels at Nainital, and CO levels at Pantnagar were observed. During this event, O3 mixing ratios at Bode and Nainital increased by 73 % and 67 %, respectively, and CO mixing ratios at Bode and Pantnagar increased by about 24 % and 58 %, respectively. The MODIS fire maps during these events show a large increase in MODIS fire counts in the Punjab region (blue box in top panel of Fig. 15) in the northwestern Indian subcontinent. During this period, the total increase in fire counts is ∼350 % that of 25–30 April. The spatial distribution of AIRS-retrieved CO mixing ratios at 925 hPa and 850 hPa also indicates an overall increase in CO levels around the northern part of the Indian subcontinent. A similar increase is also observed in model (MOZART4/GEOS5) simulations in which a 2-fold increase in CO levels at 992 hPa during this period is observed (Fig. S7). Further, OMI tropospheric column NO2 (30 % cloud screened) also showed a small enhancement during the same period (Fig. S7). The fire emissions associated with this period indicate that the majority of emissions are from crop residue burning, unlike the previous case in April. Late April–early May is the harvest period for crops in northern India, and widespread crop residue burning is common during this period. This leads to the release of massive amounts of pollutants over this region (Bhardwaj et al., 2016). The HYSPLIT backward air trajectories also indicate that air masses during the period were arriving from these source regions. During this period, nothing noticeable in wind speeds, directions, temperature solar radiation and rainfall is observed; thus we could conclude that this increase in surface mixing ratios of O3 and CO at these sites could be related to biomass burning in the northern Indian subcontinent. During this period, the influences are not only limited to the surface level but also to higher altitudes at which satellite-retrieved vertical profiles of CO show high levels during this period. Rupakheti et al. (2017) also reported a similar increase in the ambient concentrations of black carbon, particulate matter, CO and ozone at Lumbini (regional site of SusKat campaign) during these two episodes (7–9 April and 3–4 May). This site is located south of the Kathmandu Valley Himalayan foothills near the IGP. The notable changes in surface O3 and CO mixing ratios and GFED biomass burning emissions over the Kathmandu region are shown in Table 4.

Similar influences of widespread biomass burning have also been observed in the vertical distribution of ozone over the central Himalayas. Details of balloon-borne observations of ozone (Ojha et al., 2014) and meteorological parameters along with an intercomparison of two kinds of meteorological sensors (iMet and Vaisala) are given in Naja et al. (2016). The weekly balloon-borne ozone profiles from Nainital (on 9 May) also confirmed an enhancement in ozone (∼16 ppb) in the lower troposphere (2–4 km) when compared with the ozone profile on 1 May (Fig. 16). The enhancement is about 14 ppbv in the 4–6 km region. Such events are generally observed during the spring season, when the influence of regionally polluted air masses from the IGP could travel over long distances.

The tropopause folding events and influences of stratosphere–troposphere exchange (STE) over the Indian subcontinent and Tibetan Plateau are more frequent during winter and early spring (Cristofanelli et al., 2010; Chen et al., 2011; Phanikumar et al., 2017) and a few attempts in the past were made to understand the role of STE over the Indian subcontinent (Mandal et al., 1998; Ganguly and Tzanis, 2011). Here, to understand the role of STE in surface ozone levels in Kathmandu vertical distribution of Ertel potential vorticity (EPV) using MERRA v2.0 reanalysis, AIRS-retrieved ozone, RH and CO were observed during January–May 2013 (Fig. 17). The EPV distribution is represented by the potential vorticity unit (PVU) threshold (EPV > 1.6 PVU = 1.6×10-6 K m2 Kg-1 s-1) defined by Cristofanelli et al. (2006). Since the EPV near the extratropical tropopause is at about 2 PVU and EPV in the stratosphere is about 1–2 orders of magnitude higher, any values of EPV greater than 1.6 are suggested to be associated with the downward transport of ozone-rich air masses from above (Cristofanelli et al., 2006). Overall, EPV distribution suggests downward transport associated with STE is limited to the upper and middle parts of the troposphere (Fig. 17a). The ozone distribution also exhibits similar behavior for which increases in ozone levels (downward transport) reaching the surface are negligible. However, a seasonal increase in ozone mixing ratios at the lower and middle troposphere can be observed during January to May. To investigate the role of STE during April and May, vertical profiles of EPV, O3, CO and RH were studied for these periods (Fig. 18). During the April event (Fig. 18a) EPV during 11–16 April was found to be increased compared to the previous period (March), but it was reduced during the 3–6 April period. During both the high fire periods, ozone and CO were higher during the event, which we propose is due to biomass burning since during STE reductions in CO are also observed. Therefore, the April event does not show any clear evidence of STE based on our analysis. Further, during May (Fig. 18b) an increase in EPV at higher altitudes and O3 at lower altitudes is observed, but again CO did not show any decrease during this period. Therefore, we conclude that during both of these high-pollution periods, no signs of downward transport were observed.