One year of hydrogen sulfide measurements in an urban traffic site were used to assess the importance of traffic in the overall high ambient air burden of H2S (mean annual concentration of 8 μg/m3). During calm nighttime hours, hourly values of H2S were correlated positively with those of CO (R2 = 0.75) and SO2 (R2 = 0.70), suggesting a common source from traffic. For the driving conditions of the measurement site, H2S automotive emissions are around 24% w/w those of SO2, while European inventories calculate these at only 2.5%. H2S annual traffic emissions, based on the existing emission inventories for CO and SO2 are calculated at 0.199–0.311 × 10−3 Tg a−1 for Thessaloniki and are extrapolated to 0.030–0.0485 Tg a−1 for the EU15 countries and 2.6 Tg a−1 for Europe and Eurasia. It appears that traffic emissions make up a significant contribution to the global H2S budget.
 Unmodified three-way catalysts in which PGM (platinum-group metals) are supported on CeO2-Al2O3, are known to effectively store sulfur during a fuel-lean phase and then release it in the form of a large H2S spike when the exhaust gas becomes rich [e.g., Diwell et al., 1991]. Hence, traffic could be a potentially significant global source of H2S. However, there are very sparse data on H2S in urban areas, and no urban monitoring network makes measurements routinely to allow for an assessment of traffic's contribution to the global budget under real-world driving conditions. Only non-systematic measurements of ambient air H2S concentrations at roadsides or road roundabout sites in UK have been made in an effort to assess whether traffic emissions can impact air quality [Watts and Roberts, 1999]. The levels found in London were 700–800 pptv, and up to 2 ppbv, although transient, few-seconds peaks up to 80 ppbv have been observed [Deuchar et al., 1999]. Much lower levels were reported from two freeway sites in Los Angeles, with monthly means from below 20 pptv to 269 pptv (Oxford Brookes University and the Getty Conservation Institute, cited by Watts and Roberts ). This may reflect the absence of Nickel in US catalysts.
 Presently, in the sketchy estimates available for the various components of the H2S budget, automotive emissions are thought to play a minor role, contributing, according to Watts  0.3 Tg a−1 to a total anthropogenic source of 3.3 Tg a−1 [Moeller, 1984; Watts, 2000] and a total global source of 7.7–60 Tg a−1.
 The data used in the study were acquired by the Environmental Department of Municipality of Thessaloniki [Petrakakis et al., 1995]. Thessaloniki has a population of approximately one million, and is a coastal city. The measurements were made during the full year of 1995 (January and April had limited coverage), at the Venizelou str. station (station coordinates: 40°38′15″N, 22°56′30″E, 11.6 m ASL) of the air pollution monitoring network of the Municipality of Thessaloniki, operated by the Environmental Department of the Municipality. The station is located at the centre of the city, at the crossroads of Egnatia str. and Venizelou str., two main roads. The traffic is relatively heavy at this cross road, with up to 5,000 vehicles hr−1 at Egnatia str. and up to 1,400 vehicles hr−1 at Venizelou str. During a typical day, around 4,000 vehicles hr−1 pass Egnatia str., their numbers remaining more or less stable between 7:00–23:00 hrs, as this street is one of the major axes of the city. The traffic data are from the measurements of the responsible authority, the 3rd Regional Public Works Agency (3rd PYDE). Other parameters measured at the station and used in the present study are CO and SO2 mixing ratios, temperature, wind velocity and wind direction.
 Hydrogen sulfide was measured by the commercial H2S analyser SH20M from Environnement S.A., which operates on the principle of ultra-violet fluorescence (UVF). Briefly, ambient air was sampled at a height of 3.5 m above street level and via a Teflon tube entered the analyser at a flow rate of 0.35 l/min, after passing through a Teflon filter for particle removal. After passing through an SO2 filter (PTFE tube, filled with sodium carbonate 5%) and a HC filter/H2S to SO2 converter (Pyrex glass, filled with Vanadium pentoxide on silica substrate, at 400°C) the sample enters the measurement chamber of the analyser (T = 40°C, P = 60 mm Hg) where it is illuminated by a UV lamp with maximum at 214 nm. At this wavelength, SO2 absorbs incoming radiation and re-emits it as fluorescence. For zeroing, the sample passes through an active carbon filter. The detection limit of the analyser was 1 ppb, the response time from 0 to 90% was 30 sec and the reproducibility 1%. The analyser was calibrated at least once a week either with the in-build H2S permeation tube or with gas flasks of known H2S concentrations and a dilution system.
 CO was measured by the CO11M model of Environnement S.A. (NDIR method), while SO2 was measured with the AF21M model of Environnement S.A. (UVF method). The analysers were calibrated at least once weekly with gas flasks of known concentrations and a dilution system.
 Temperature was measured by a platinum sensor (R.M. Young Company, model 43372VC), while wind was measured by a Gill UVW vector anemometer (R.M. Young Company, model 27005). Wind was measured at a height of 31 m above street level, while temperature was measured at a height of 28 m above street level.
 Data are archived as hourly and daily mean values. In this study, both the 1-hr and the daily mean mixing ratios were used. While daily means are used for the whole 1995, 1-hr values were used for February and March (wintertime) and July and August (summertime).
3. Results and Discussion
 Although not the scope of this paper, below some features of the H2S dataset are given briefly. The mixing ratios of H2S were at their lowest during the summer months as a result of a combination of factors, such as increased vertical mixing due to higher mixing height, shorter lifetime due to higher OH levels and, perhaps, lower emissions during summertime because of the holiday season. A surprising result was the relatively high mixing ratios, especially during the winter, where monthly mean values of 20 μg m−3 (about 12.9 ppb at the measured temperatures, January) were observed. The diurnal variation of H2S during wintertime closely follows that of the other primary pollutants CO and SO2; however, H2S ratios to these two species are higher during the early morning rush hours. During summertime, the diurnal variation of H2S resembles that CO and SO2; however, the peak in CO is broader (i.e. lasts more hours) than the SO2 and H2S peaks and H2S ratios to these two species follow a more complex pattern than winter (figure not shown).
 To obtain information on the possible sources of the high levels observed, two bimonthly subsets of the 1995 data were chosen for further analysis, to represent wintertime and summertime conditions: February–March, and July–August, respectively. This choice is justified by the temperature measurements. Although we used February–March values for wintertime due to the low January coverage in the data, the mean February temperature was 10.6 ± 2.6°C while the mean March temperature was 10.7 ± 2.9°C, hence both months can be considered typical of the cold part of the year. Hourly values were used for this analysis.
 The correlation between H2S and CO (or SO2) hourly values during summertime is very low (R2 = 0.08 for CO and R2 = 0.09 for SO2). In summertime, generally, correlations between species with common sources may be low because of different photochemical lifetimes and more intense vertical mixing [e.g., Kourtidis et al., 1999, 2002]. Indeed, the correlation between H2S and CO (or SO2) hourly values during wintertime is quite good (R2 = 0.57 for CO and R2 = 0.41 for SO2, Figure 1) and improves if only conditions with reduced mixing are chosen, as are periods of wind speed <0.5 m/s (Figure 2). These periods for more than 95% of the time occurred during nighttime. For these periods, R2 = 0.75 for the correlation with CO and R2 = 0.7 for the correlation with SO2 (Figure 2). Given that the main sources of CO in the close vicinity of the station for the given conditions are traffic, the observed relationships make us conclude that traffic is also the source of H2S at the measurement site. In such cases, the slope of the correlation of two species with sufficiently long lifetimes can be used to obtain relative emission rates [see Kourtidis et al., 1999]. The slope of the correlation, suggests relative emission rates near 2.68 μg H2S/mg CO (i.e. 2.68 × 10−3 g H2S/g CO). Fried et al.  reported for another sulfur species, COS, also at the −2 oxidation state, US fleet exhaust measurements slopes of 5.8 × 10−6 g COS/g CO for gasoline vehicles and 200 × 10−6 g COS/g CO for diesel vehicles, three orders of magnitude and one order of magnitude, respectively, lower than the H2S relative emission rates inferred in the present study. However, because of the different catalyst metals used in US and Europe, these results are not directly comparable [e.g., Diwell et al., 1991]. Our results here indicate also that the Fried et al.  calculations on the relative importance of global automotive emissions on the global COS budget might be revised after measurements have been made for different types of catalysts, as those of the European fleet, since it appears that different catalysts behave differently regarding the release of reduced sulfur.
 The emission inventory for Thessaloniki of Ziomas et al.  calculates annual traffic CO emissions of 0.076 Tg a−1, an estimate that was recently revised upward to 0.116 Tg a−1 by Symeonidis . Using the slope of 2.68 μg H2S/mg CO of the linear regression curve of the observed tight relationship between H2S and CO during winter low wind speed periods (<0.5 m/s) and the estimate of annual traffic CO emissions of Symeonidis , an estimate for the annual traffic H2S emissions of Thessaloniki can be derived. This is calculated at 0.116 × 2.68 × 10−3 Tg a−1 = 0.311 × 10−3 Tg a−1.
 The slope of 0.24 μg H2S/μg SO2 of the linear regression curve (R2 = 0.7) of the observed tight relationship between H2S and SO2 during winter under low wind speed suggests that about 19% of the annual sulfur (i.e., SO2 + H2S) traffic emissions in Thessaloniki are in the form of H2S. This is substantially (factor-of-ten) higher than the 2.5% quoted by Watts  for automotive UK, Europe and Eurasia emissions. From this slope and annual traffic SO2 emissions for Thessaloniki of 0.831 × 10−3 Tg a−1 (from the inventory of Symeonidis ), we derive 0.199 × 10−3 Tg a−1 annual traffic H2S from Thessaloniki, which is around 30% lower than the value calculated from the H2S/CO relationship. Given the uncertainties in emission inventory methodologies, this is a reasonable agreement and indicates that the estimates of Symeonidis  are either somewhat high for CO, somewhat low for SO2, or both.
 Of the major S-containing atmospheric gases, hydrogen sulfide budget estimates are considered the “most tenuous” [Watts, 2000]. Watts  reviews sparse available data and estimates and presents estimates of a total source to the atmosphere of 7.72 ± 1.25 Tg a−1, with an anthropogenic component of 3.3 ± 0.33 Tg a−1. The latter stems from the estimate of Moeller , with the addition of a 0.3 Tg a−1 source from car exhaust based on Fried et al.  results.
 Given that the annual traffic European Union (EU15) SO2 and CO emissions for 1994 are 0.505 Tg a−1 and 27.99 Tg a−1 respectively [European Topic Centre on Air Emissions, 1998], the slopes of the respective ratios with H2S from the present study, and assuming that about 40% of these emissions occur under urban traffic conditions similar to those at the measurement site, a European Union source of 0.030 Tg a−1 H2S (from CO) - 0.0485 Tg a−1 H2S (from SO2) is estimated.
Watts  arrived at a 0.3 Tg a−1 H2S automotive source from Europe and Eurasia starting from the assumption that 2.5% of emitted SO2 is in the H2S form and a 13.7 Tg a−1 estimate for 2000 SO2 traffic emissions for the same area (no quoted source is given). Our 19% estimate for the H2S contribution to traffic S emissions revises the Watts  0.3 Tg a−1 H2S automotive source from Europe and Eurasia to 2.6 Tg a−1 H2S (using the Watts  13.7 Tg a−1 estimate for 2000 SO2 traffic emissions), which is a quite substantial contribution to the H2S budget (comparable to the combined natural emissions from the ocean, salt-marshes, estuaries and wetlands and twice as the volcanic-geothermal emissions).
 Although the uncertainties associated with extrapolations from a single site to the regional level are big, the H2S budget is currently so sketchy that our results from a large database from real city traffic conditions represent an improvement of current understanding and clearly point out the importance of traffic emissions. Until now, the only experimental work relevant to H2S anthropogenic emission estimates was a by-product of the Fried et al.  work on COS in car exhaust. Hence, the present work may provide some momentum to more widespread monitoring of H2S in urban areas, given also the emerging concerns for potential health effects of reduced sulfur compounds. The results may also have some relevance for the budget of photooxidants in cities, since it requires more oxidation steps for the conversion of H2S from oxidation state −2 to the +6 oxidation state of sulfate than for SO2, which has an oxidation state of +4.
 We presented here an analysis of H2S/CO and H2S/SO2 ratio measurements in the center of the city of Thessaloniki, at the crossroads of two major roads with the aim of gaining some insight into the magnitude of traffic H2S emissions.
 Based on observed H2S/CO and H2S/SO2 relationships during meteorological periods of reduced dispersion, we conclude that the main source of H2S at the measurement location is automotive emissions. Based on observed H2S/SO2 relationships during meteorological periods of reduced dispersion, we conclude that around 19% of the emitted SO2 from traffic is in the form of H2S. Taking into account a recent estimate for traffic CO emissions in Thessaloniki, we estimate that the annual H2S traffic emissions from the city might lie in the order of 0.119–0.311 × 10−3 Tg a−1. European (EU15 countries) emissions are calculated at 0.03–0.0485 Tg a−1, while Europe + Eurasia emissions are calculated at 2.6 Tg a−1, a substantial contribution to the total global source of H2S.
 Our results here give also some indication that the Fried et al.  calculations on the relative importance of global automotive emissions on the global COS budget might be revised after measurements have been made for the European fleet.