OCS in the atmosphere is primarily generated from ocean and anthropogenic sources. A portion of these sources are indirect, emitted as CS2 which can be oxidized to OCS (Zeng et al., 2016). Within the atmosphere, major sinks of OCS are OH oxidation in the troposphere and photolysis in the stratosphere. Besides large volcanic eruptions, OCS is a significant source of sulfur to the stratosphere and was briefly entertained as a geoengineering approach to promote global dimming (Crutzen, 2006). However, the global warming potential of OCS roughly balances whatever global cooling effect it might have (Brühl et al., 2012). Abiotic hydrolysis in the atmosphere plays a small role: while snow and rain were observed to be supersaturated with OCS (Belviso et al., 1989; Mu et al., 2004), even in the densest supersaturated clouds the OCS in the air would represent 99.99 % of the OCS present (Campbell et al., 2017b). Multiple lines of evidence support uptake by plants as the dominant removal mechanism of atmospheric OCS (e.g., Asaf et al., 2013; Berry et al., 2013; Billesbach et al., 2014; Campbell et al., 2008; Glatthor et al., 2017; Hilton et al., 2017; Launois et al., 2015b; Mihalopoulos et al., 1989; Montkza et al., 2007; Protoschill-Krebs and Kesselemeier, 1992; Sandoval-Soto et al., 2005; Stimler, 2010b; Suntharalingham et al., 2008).

The observed atmospheric OCS distribution suggests that seasonality is driven by terrestrial uptake in the Northern Hemisphere and oceanic fluxes in the Southern Hemisphere (Montzka et al., 2007). Improvements in the OCS budget were derived through inverse modeling of NOAA tower and airborne observations on a global scale (Berry et al., 2013; Launois et al., 2015b; Suntharalingam et al., 2008). Lower concentrations were generally found in the terrestrial atmospheric boundary layer compared to the free troposphere during the growing season, and amplitudes of seasonal variability were enhanced at low-altitude stations, particularly those situated mid-continent.

Total column measurements of OCS from ground-based FTS show trends in OCS concentrations coincident with the rise and fall of global rayon production, which creates OCS indirectly (Campbell et al., 2015). Kremser et al. (2015) found an overall positive tropospheric rise of 0.43–0.73 %yr-1 at three sites in the Southern Hemisphere from 2001 to 2014. The trend was interrupted by a sharply decreasing interval from 2008 to 2010, also observed in the global surface flask measurements (Fig. S2; Campbell et al., 2017a). A similar but smaller dip was observed in the stratosphere, indicating that the trends are driven by processes within the troposphere. Over Jungfraujoch, Switzerland, Lejeune et al. (2017) observed a decrease in tropospheric OCS from 1995 to 2002 and an increase from 2002 to 2008; after 2008 no significant trend was observed. An increase in OCS concentrations from the mid-20th-century with a decline around the 1980s was also recorded in firn air (Montzka et al., 2004), following historic rayon production trends.

Changes in terrestrial OCS uptake and possibly the ocean OCS source can be observed from the 54 000-year record from ice cores. Global OCS concentrations dropped 45 to 50 % between the Last Glacial Maximum and the start of the Holocene (Aydin et al., 2016). By the late Holocene, concentrations had risen, and the highest levels were recorded in the 1980s (Campbell et al., 2017a).

Recommendations. Modern seasonal and annual variability of OCS can be validated with smaller vertical profile datasets, e.g., Kato et al. (2011), and data from flights, e.g., Wofsy et al. (2011). Interhemispheric variability on millennia timescales requires ice core data from the Northern Hemisphere: all current ice core data are from the Antarctic (Aydin et al., 2016).

OCS uptake by terrestrial vegetation is governed mechanistically by the series of diffusive conductances of OCS into the leaf and the reaction rate coefficient for OCS destruction by carbonic anhydrase (CA) (Wohlfahrt et al., 2012), though it can also be destroyed by other photosynthetic enzymes, e.g., RuBisCo (Lorimer and Pierce, 1989). CA is present both in plant leaves and soils, although soil uptake tends to be proportionally much lower than plant uptake. In soil systems, OCS uptake provides information about CA activities within diverse microbial communities. OCS uptake over plants integrates information about the sequential components of the diffusive conductance (the leaf boundary layer, stomatal, and mesophyll conductances) and about CA activity, all important aspects of plant and ecosystem function. Stomatal conductance in particular is a prominent research focus in its own right, as it couples the carbon and water cycles via transpiration and photosynthesis.

Terrestrial plant OCS uptake has typically been derived by scaling estimates of the plant CO2 uptake with proportionality coefficients, such as the empirically derived leaf relative uptake rate ratio (LRU; Sandoval-Soto et al., 2005): FOCS=FCO2[OCS][CO2]-1LRU, where FOCS is the uptake of OCS into plant leaves; FCO2 is CO2 uptake; [OCS] and [CO2] are the ambient concentrations of OCS and CO2; and LRU is the ratio of the OCS to CO2 uptake, which is a function of plant type and water and light conditions. The concept of the LRU is a simplification of the leaf CO2 and OCS uptake process. The FCO2-to-FOCS relationship depends on the leaf conductance to each gas as it changes with the difference between concentrations inside and outside of the leaf. This requires further modeling to anticipate within-leaf concentrations of OCS and CO2, which cannot be observed directly. To keep the simplicity of the approach, especially when using OCS to evaluate models with many other built-in assumptions, the data-based LRU approximation is sufficient in many cases. We have compiled LRU data (n=53) from an earlier review and merged them with more recent published studies (Berkelhammer et al., 2014; Stimler et al., 2010b, 2011, 2012). The LRUs compiled in Sandoval-Soto et al. (2005) were partly re-calculated in Seibt et al. (2010) to account for the lower gas concentrations in the sample cuvettes. For C3 plants, OCS uptake behavior is attributed to CA activity (Yonemura et al., 2005). As shown in Fig. 2, LRU estimates for C3 species under well-illuminated conditions are positively skewed, with 95 % of the data between 0.7 to 6.2, which coincides with the theoretically expected range of 0.6 to 4.3 (Wohlfahrt et al., 2012). The median, 1.68, is quite close to values reported and used in earlier studies and provides a solid “anchor ratio” for linking C3 plant OCS uptake and photosynthesis in high light. LRU data are fewer for C4 species (n=4), converging to a median of 1.21, reflecting more efficient CO2 uptake rates compared to C3 species (Stimler et al., 2011).

LRU remains fairly constant with changes in boundary layer and stomatal conductance but is expected to deviate due to changes in internal OCS conductance and CA activity (Seibt et al., 2010; Wohlfahrt et al., 2012). The primary environmental driver of LRU is light, and an increase in LRU with decreasing photosynthetically active radiation has been observed at both the leaf (Stimler et al., 2010b, 2011) and ecosystem scale (Maseyk et al., 2014; Commane et al., 2015; Wehr et al., 2017; Yang et al., 2018). This behavior arises because photosynthetic CO2 assimilation is reduced in low light, whereas OCS uptake continues since the reaction with CA is not light dependent (Stimler et al., 2011). Note that since low light reduces CO2 uptake, the flux-weighted effect of the variations in LRU on estimating FCO2 (or GPP) is also reduced on daily or longer timescales (Yang et al., 2018).

An additional complication is introduced by soil and non-vascular plant processes that both emit and consume OCS, with a few studies reporting net OCS emission under certain conditions comparable in magnitude to net uptake rates during peak growth. Generally, soil OCS fluxes are low compared to plant uptake with a few exceptions (Fig. 3). In non-vascular plants, OCS uptake continues in the dark even when photosynthesis ceases (Gries et al., 1994; Kuhn et al., 1999; Kuhn and Kesselmeier, 2000; Gimeno et al., 2017; Rastogi et al., 2018). Unlike other plants, bryophytes and lichens lack responsive stomata and protective cuticles to control water losses. OCS emissions from these organisms seem to be primarily driven by temperature (Gimeno et al., 2017).

The yearly average land OCS flux rate in recent modeling studies of global budgets (i.e., plant and soil uptake minus soil emissions) ranges from -2.5 to -12.9 pmolm-2s-1 (Fig. 3). The only study reporting year-round OCS flux measurements is from a mixed temperate forest, which was a sink for OCS with a net flux of -4.7 pmolm-2s-1 during the observation period (Commane et al., 2015). Daily average OCS fluxes during the peak growing season are available from a larger selection of studies and cover the range from -8 to -23 pmolm-2s-1, excluding that of Xu et al. (2002), which found a surprisingly high uptake (–97±11.7 pmolm-2s-1) from the relaxed eddy accumulation method (Fig. 3). Despite the limited temporal and spatial coverage, these data suggest that some of the larger global land net sink estimates may be too high (Launois et al., 2015b).

The following subsections explore a few aspects of ecosystem OCS exchange in greater detail. Observations and conclusions about forests, grasslands, wetlands, and freshwater ecosystems are explored. Then we examine OCS interactions reported for components of ecosystems: soils, microbial communities, and abiotic hydrolysis and sorption.

Recommendations. Available observations are limited in time and do not cover tropical ecosystems, which contribute almost 60 % of global GPP (Beer et al., 2010). More year-round measurements from a larger number of biomes, in particular those presently underrepresented, are required to provide reliable bottom-up estimates of the total net land OCS flux. The causes for the observed variability in Fig. 2 require more investigation because they hamper the specification of defensible plant-functional-type-specific LRUs (Sandoval-Soto et al., 2005) and the development of models with non-constant LRU (Wohlfahrt et al., 2012). Relatively little is known regarding using OCS to estimate CA activity (Wehr et al., 2017), which is a promising new avenue of OCS research. Within this context, plant physiological and enzymatic adaptations to increasing CO2 and their effects on the exchange of OCS are of special interest.

The oceans are known to contribute to the atmospheric budget of OCS directly via OCS and indirectly via CS2 (Fig. 8) (Chin and Davis, 1993; Watts, 2000; Kettle et al., 2002). Large uncertainties are still associated with current estimates of marine fluxes (Launois et al., 2015a; Lennartz et al., 2017, and references therein) and have led to diverging conclusions regarding the magnitude of their global role.

The range of observed OCS concentrations in surface waters informs how the magnitude of direct oceanic emissions is calculated. Observations of OCS in the surface water of the Atlantic, Pacific, Indian, and Southern oceans revealed a consistent daily concentration range of ∼10–100 pmolL-1 in the surface mixed layer on average, across different methods. Largest differences are found between coastal and estuaries (range: nanomoles per liter) and open oceans (range: picomoles per liter) (Table 3).

Anthropogenic OCS sources include direct emissions of OCS and indirect sources from emissions of CS2. The dominant source is from rayon production (Campbell et al., 2015), while other large sources include coal combustion, aluminum smelting, pigment production, shipping, tire wear, vehicle emissions, and coke production (Blake et al., 2008; Chin and Davis, 1993; Du et al., 2016; Lee and Brimblecombe, 2016; Watts, 2000; Zumkehr et al., 2017).

All recent global atmospheric modeling studies have used the low estimate of 180 GgSyr-1 from Kettle et al. (2002), which did not capture significant emissions from China. Updated globally gridded inventories are considerably higher: a bottom-up estimate of 223–586 GgSyr-1 for 2012 (Zumkehr et al., 2018) and a top-down assessment of 230 to 350 GgSyr-1 for 2011 to 2013 (Campbell et al., 2015). One reason for the gap between the two recent inventories is that the top-down study used an optimization approach in which the result was limited to the a priori range, 150 to 364 GgSyr-1. Both datasets indicate that most anthropogenic sources are in Asia.

Biomass burning is generally accounted for as a category separate from anthropogenic emissions. Several airborne campaigns have observed increases in OCS concentrations in air masses from nearby burning events (Blake et al., 2008). The most recent estimate suggests that biofuels, open burning, and agriculture residue are 63, 26, and 11 % of the total OCS biomass burning emissions, respectively (Campbell et al., 2015).

Recommendations. Anthropogenic OCS emissions experience large year-to-year variation (Campbell et al., 2017a). Ambient OCS monitoring and on-site industry observations in Asia could observe the anthropogenic contribution over time. In particular, modern viscose-rayon factory emissions are necessary to capture the variability of emissions factors used to scale rayon production to OCS emissions using economic data.

OCS is emitted into the atmosphere by degassing magma, volcanic fumaroles, and geothermal fluids. OCS can be released at room temperature by volcanic ash (Rasmussen et al., 1982) and has been observed to be conservative in the atmospheric plume emitted by the Mount Erebus volcano up to tens of kilometers downwind of the volcanic source (Oppenheimer et al., 2010).

Using the linear relationship between the logarithm of the OCS / CO2 ratio in volcanic gases and temperature, the volcanic OCS contribution was determined from estimated CO2 emissions (Belviso et al., 1986). Here we calculate a revised temperature dependence of log[OCS / CO2] with additional data (Chiodini et al., 1991; Notsu and Toshiya, 2010; Sawyer et al., 2008; Symonds et al., 1992), as shown in Fig. 10. The compilation of measurements from 14 volcanoes shows that the former relationship from Belviso et al. (1986) overestimated the OCS / CO2 ratio of volcanic gases with emission temperatures from 110 to 400 ∘C, typical of extra-eruptive volcanoes. Even with this improved estimate, OCS emissions of extra-eruptive volcanoes are negligibly small and can definitely be discarded from the inventory of volcanic OCS emissions. Eruptive and post-eruptive volcanoes contribute almost all OCS emissions from volcanism.

Recommendations. An updated inventory of eruptive volcanoes and a better assessment of their CO2 emissions will refine our understanding at a regional scale of the contribution of OCS from volcanoes. Special attention should be paid to the Ring of Fire off the Asian continent where satellites have observed significant atmospheric OCS enhancements.

We calculate a “bottom-up” global balance of OCS with several approaches, as presented in Table 4. Within the atmosphere, the tropospheric sink owing to oxidation by OH is estimated to be in the range 82–130 GgSyr-1 (Berry et al., 2013; Kettle et al., 2002; Watts, 2000), and the stratospheric sink is in the range 30–80 GgSyr-1, or 50±15 GgSyr-1 (Barkley et al., 2008; Chin and Davis, 1995; Crutzen, 1976; Engel and Schmidt, 1994; Krysztofiak et al., 2015; Turco et al., 1980; Weisenstein et al., 1997). OCS concentrations are increasing roughly 0.5–1 pptyear-1 averaged over the last 10 years (Campbell et al., 2017a), suggesting approximately 2 to 5 GgSyr-1 remains in the troposphere.

We build a budget for terrestrial biomes that relies on observations where available, and on estimates of carbon uptake where no data exists, as has been done previously (Campbell et al., 2008; Kettle et al., 2002; Suntharalingam et al., 2008). In Table 5, the estimated OCS uptake is first calculated from a GPP estimate and Eq. (1); then the net OCS flux is appraised by taking into account observed or estimated soil fluxes for each biome. The [CO2] and [OCS] are assumed to be 400 ppm and 500 ppt, respectively, and LRU is 1.16±0.2 for C4 plants (Stimler et al., 2010b) and 1.99±1.44 for C3 plants (Fig. 2). We further assume a 150-day growing season with 12 h of light per day for the purposes of converting between annual estimates of GPP and field measurements calculated in per-second units, though this obviously does not represent the diversity of biomes' carbon assimilation patterns. Additionally, we assume that plants in desert biomes photosynthesize using the C4 pathway. Converting annual estimated FCOS from an annual estimate to a per-second estimate is sensitive to our growing season assumption. The lack of soil OCS flux time series datasets makes a more sophisticated upscaling approach ineffective. Anticipated fluxes from soils and plants are therefore combined in this purposely simple method, scaled to the area of the biome extent, and presented in Table 4 as annual contributions to the atmospheric OCS budget.

We use a range of OCS flux observations in picomoles of OCS per square meter per second for fresh and saline wetlands: -15 (de Mello and Hines, 1994) to +27 (Liu and Li, 2008) for freshwater wetlands and -9.5 (Li et al., 2016) to +60 (Whelan et al., 2013) for saltwater wetlands (Fig. 4). Marine and inland wetlands cover 660 and 9200×103 km2, respectively (Lehner and Döll, 2004). Performing a simple scaling exercise results in contributions of -140 to 250 and -6 to 40 GgSyr-1 for fresh and saltwater wetlands, respectively, yielding a total range of -150 to 290 GgSyr-1 (Table 4).

To determine the role of non-vascular plant communities to the atmospheric OCS loading, we leverage Eq. (1) and work that has already been done on their carbon balance. According to Elbert et al. (2012), the annual contribution is 3.9 PgCyr-1. A [OCS] of 500 ppt, a [CO2] of 400 ppm, and a LRU of 1.1±0.5 (Gimeno et al., 2017) yield -8 to -21 GgSyr-1.

To estimate the maximum possible source of lakes to the atmospheric OCS burden, we perform a simple estimation of the global OCS flux following the approach in MacIntyre et al. (1995) as FOCS=k(caq-ceq), where gas transfer coefficient, k, is assumed to be constant at 0.54 md-1 (Read et al., 2012); OCS concentration in the water, caq, is 90 pmolL-1 to 1.1 nmolL-1 (Richards et al., 1991); and OCS concentration in the surface water if it was in equilibrium with the above air, ceq, is calculated using Henry's law at global average temperature of 15 ∘C and global atmospheric OCS mixing ratio of 500 ppt. Accounting for the number of ice-free days in a year and total lake surface area per latitude (Downing et al., 2006), the range of possible COS burden from lakes to the atmosphere is reported here as 0.8 to 12 GgSyr-1.

Lennartz et al. (2017) generated a direct estimate of direct OCS emissions from oceans as 130±80 GgSyr-1. A molar yield of CS2 to OCS of 0.81–0.93 was established by Stickel et al. (1993) and Chin and Davis (1993), resulting in ocean OCS emissions from CS2 with an uncertainty of 20–80 GgSyr-1. This uncertainty is from the emissions of CS2, not the molar yield, for which a globally constant factor is used. The global DMS oxidation source of OCS was estimated by Barnes et al. (1994) as 50.1–140.3 GgSyr-1, and subsequent budgets contain only revisions according to updated DMS emissions (Kettle et al., 2002; Watts, 2000). We suggest that the uncertainty in the production of OCS from DMS is underestimated. Until these issues are resolved, we recommend that this term be removed as a source from future budgets, but retained as an uncertainty.

Bottom-up analysis of the global anthropogenic inventory indicates a source of 500±220 GgSyr-1 for the year 2012 (Zumkehr et al., 2018). The large uncertainty is primarily due to limited observations of emission factors, particularly for the rayon, pulp, and paper industries. The most recent estimate of the biomass burning sources is 116±52 GgSyr-1 (Campbell et al., 2015).

To calculate global volcanic OCS emissions, we first consider the range of global volcanic CO2 emission estimates of the five studies reviewed by Gerlach (2011) of 0.15–0.26 Pgyr-1, or 0.205±0.055 Pgyr-1. Assuming that the mean OCS / CO2 molar ratio of gases emitted by eruptive and post-eruptive volcanoes is 2.3×10-4 (for emission temperatures in the range 525–1130 ∘C, see Fig. 10), the revised annual volcanic input of OCS into the troposphere is estimated to be in the range 25–43 GgSyr-1.

Examining Table 4, we find large uncertainties in many global estimates, and some biome observations are completely absent. It has been suggested that ocean OCS production has been underestimated (Berry et al., 2013), and some research points to unaccounted-for anthropogenic sources (Zumkehr et al., 2018). The uncertainty in our ocean OCS production and/or the industry inventories does not necessarily capture the true range of OCS fluxes. Despite the large uncertainties of the global OCS budget, many applications of the OCS tracer have been attempted with success.

Recommendations. More observations in the ocean OCS source region and from industrial processes, particularly in Asia, are needed to further assess their actual magnitude and variation (Suntharalingam et al., 2008). Current leaf-based investigations need to be expanded to include water- or nutrient-stressed plants. Measurements from biomes with a complete lack of data, such as deserts and the entirety of the tropics, are desperately needed.

Top-down estimates use observed spatial and temporal gradients of OCS in the atmosphere to adjust independent surface fluxes, called the prior estimate. Constraints can be introduced to the results; e.g., Launois et al. (2015b) used flask measurement observations to optimize surface OCS flux components to obtain a closed global OCS budget. Other top-down estimates without this restriction found a missing source of about 600–800 GgSyr-1 in the atmospheric budget of OCS (Berry et al., 2013; Glatthor et al., 2015; Kuai et al., 2015; Suntharalingam et al., 2008; Wang et al., 2016). This could be the result of missing oceanic sources, missing anthropogenic OCS sources from Asia, overestimated plant uptake, or a combination of factors.

Kuai et al. (2015) implied a large ocean OCS source over the Indo-Pacific region with the total ocean source budget consistent with the global budget proposed by Berry et al. (2013). The observations in Kuai et al. (2015) were estimated OCS surface fluxes from NASA's Tropospheric Emission Spectrometer (TES) ocean-only observations. A similar conclusion was obtained by Glatthor et al. (2015), who showed that the OCS global seasonal cycle observed by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) was more consistent with the seasonal cycles modeled using the Berry et al. (2013) global budget than using the global budget proposed earlier by Kettle et al. (2002).

Most of the anthropogenic source is located in China, while most of the atmospheric OCS monitoring is located in North America (Campbell et al., 2015). The spatial separation allows regional applications of OCS to North America to control for most of the anthropogenic influence through observed boundary conditions (Campbell et al., 2008; Hilton et al., 2015, 2017). The anthropogenic source has large interannual variations (Campbell et al., 2015), which suggest that applications of the OCS tracer to inter-annual carbon cycle analysis will require careful consideration of anthropogenic variability.

Recommendations. The accuracy of OCS surface flux inversions can be improved by using simultaneous OCS observations from multiple satellites, e.g., TES and MIPAS, to provide more constraints on the OCS distribution in different parts of the atmosphere. Satellite products need to be compared to observations to determine how well the upper troposphere can reflect surface fluxes, e.g., long-term tower measurements, airborne eddy flux covariance, and atmospheric profiles. This effort is furthered by better estimates of surface fluxes, in particular observations of OCS emissions from the oceans where we assume a large source region might exist (Kuai et al., 2015) and where poorly described anthropogenic sources are located in Asia (Zumkehr et al., 2018).

Here we describe work using OCS observations to assemble more information about ecosystem functioning on different scales. Estimates disagree in their diagnoses of global (Piao et al., 2013) and regional (Parazoo et al., 2015) GPP magnitude and spatial distribution in North America (Huntzinger et al., 2012), the Amazon (Restrepo-Coupe et al., 2017), and Southeast Asia (Ichii et al., 2013). Feeding observations of OCS uptake over land into transport models informs the spatial distribution and magnitude of GPP. With the suite of OCS flask and satellite data available, we describe studies that examine OCS fluxes with the top-down approach. Finally, we examine GPP estimates on very long temporal scales using the OCS ice core record.

OCS and CO2 uptake within plant leaves is partly regulated by the opening of stomata on leaf surfaces. Stomatal conductance is typically determined from combined estimates of transpiration, water vapor concentration, and leaf temperature. That approach can be particularly challenging at the canopy scale, where transpiration is difficult to distinguish from non-stomatal water fluxes (i.e., evaporation from soil and canopy surfaces) and to upscale from sap flux measurements (Wilson et al., 2001). Use of OCS uptake involves the similar but more tractable challenge of distinguishing the canopy OCS uptake from soil OCS uptake or emission, as in Wehr et al. (2017). OCS data can also look at changes in uptake activity when plants are grown in elevated CO2 environments (White et al., 2010; Sandoval-Soto et al., 2012). Use of OCS uptake may also be less sensitive to errors in leaf temperature, which is difficult to define and quantify at the canopy scale but may be improved by OCS measurements (Yang et al., 2018). However, leaf temperature may still enter the problem via estimation of mesophyll conductance and CA activity.

The use of OCS to study canopy and stomatal conductance is therefore promising, but it is so far represented mostly by very few studies (Wehr et al., 2017; Yang et al., 2018). Wehr et al. (2017) used OCS uptake to derive canopy stomatal conductance and hence transpiration in a temperate forest. Stomatal conductance was the rate-limiting diffusive step, and so its diel and seasonal patterns were retrievable from the canopy OCS uptake to within 6 % of independent estimates based on sensible and latent heat flux measurements (Fig. 12). OCS would be especially useful in humid environments or at night, when transpiration is too small to use other methods that rely on sap flow or heat flux (Campbell et al., 2017b). However, an independent estimate of CA activity and mesophyll conductance would be required.

Recommendations. OCS observations should be used to link plant physiological variables to one another. OCS fluxes are related to GPP via all three diffusive conductances, CA activity, transpiration, and the 18O isotope compositions of CO2 and H2O. The 18O connection results from the fact that CA promotes the exchange of oxygen isotopes between CO2 and liquid water in the leaves. Solar-induced fluorescence measurements could also be synergistic, as they relate to the photochemical aspect of photosynthesis, while OCS uptake relates to the gas transport aspect. So far, few research schemes have taken advantage of these relationships.

Global OCS concentrations have been retrieved from several satellite instruments, including NASA's TES (Kuai et al., 2014), the Canadian Space Agency's Atmospheric Chemistry Experiment–Fourier Transform Spectrometer (ACE-FTS) (Boone et al., 2005), and the European Space Agency's MIPAS (von Clarmann et al., 2003; Glatthor et al., 2017) and Infrared Atmospheric Sounding Interferometer (IASI) (Camy-Peyret et al., 2017; Vincent and Dudhia, 2017). Among these instruments, TES and IASI are nadir-viewing instruments (i.e., looking downwards from space towards the surface), while ACE-FTS and MIPAS are limb scanners (i.e., looking through the atmosphere tangentially). Nadir measurements are less prone to cloud interference and provide good horizontal spatial resolution but coarse vertical resolution. Limb measurements provide better vertical resolution and higher sensitivity to tracer concentrations, but they are subject to a higher probability of cloud interference and poorer line-of-sight spatial resolution. Currently there are no satellite measurements that are strongly sensitive to OCS concentrations near the surface, where they are most needed to evaluate surface fluxes.

The standard TES OCS product is an average between 200 and 900 hPa, with maximum sensitivity to the mid-tropospheric value (Kuai et al., 2014; Fig. 13a). Currently, the TES OCS retrievals are available over ocean only for latitudes below 40∘, where the signal-to-noise ratio is higher (due to larger thermal contrasts) and the surface spectral emissivity can be easily specified. Comparisons with collocated airborne and ground measurements show that the current TES OCS data have an accuracy of 50–80 ppt, and the accuracy is improved to ∼7 ppt when averaged over 1 month (Kuai et al., 2014).

MIPAS retrievals from 7 to 25 km characterize the average OCS concentration in a thin layer (a few kilometers thick) around the corresponding tangent height. Currently, the MIPAS OCS product (Fig. 13b) provides pole-to-pole OCS concentrations at multiple levels in the upper troposphere and the stratosphere, which show an accuracy of ∼50 ppt against balloon-borne measurements. Figure 14 shows the summertime (June–August) latitudinal distribution of OCS observed by MIPAS (Glatthor et al., 2017).

IASI retrieves a single value for the total column OCS (Fig. 13c). Recently, Vincent and Dudhia (2017) reported the pole-to-pole global OCS retrieved from the IASI measurements. Their preliminary test showed that the seasonally averaged IASI OCS data vary consistently with ground measurements. The IASI OCS observations over land generally agree with the MIPAS observations, showing large sinks over South America and Africa. The high spatial resolution also reveals more clearly the land OCS sources over Asia, which are not seen in TES or MIPAS observations. Furthermore, the relatively low OCS abundance over the Intertropical Convergence Zone is only apparent in IASI data.

The ACE-FTS-reported OCS concentrations in the lower stratosphere are known to be 15 % lower than the balloon-borne measurements (Velazco et al., 2011) and ∼100 ppt lower than MIPAS OCS (Glatthor et al., 2017).

Ground-based FTIR retrievals of OCS are sensitive to the altitudes between the surface and 30 km, and can therefore more directly capture the variations near the surface compared to satellite data. There are two networks of FTIR spectrometers: the Network for the Detection of Atmospheric Composition Change (NDACC), recording the mid-infrared spectra including the OCS bands, and the Total Carbon Column Observing Network (TCCON), mainly focusing on the near infrared with only some sites including the OCS bands. The FTIR remote-sensing measurement is an indirect measurement and therefore needs to be calibrated to in situ observations to have the same scale when combining the datasets. For example, Wang et al. (2016) added an offset when comparing FTIR retrievals and HIAPER Pole-to-Pole Observations (HIPPO) to the same model. Published datasets exist for the periods 1993–1997 (Griffith et al., 1998), 1978–2002 (Rinsland et al., 2002), 2001–2014 (Kremser et al., 2015), 2005–2012 (Wang et al., 2016), and 1995–2015 (Lejeune et al., 2017) and by an airborne Fourier spectrometer for the period 1978–2005 (Coffey and Hannigan, 2010). Balloon-borne FTIR data are available starting in 1985 (Toon et al., 2018).

Data are available from two kinds of airborne sampling: survey flights and atmospheric chemistry projects. OCS measurements from aircraft began in the late 1980s, using both in situ and flask collection with subsequent analysis by GC-MS (e.g., Bandy et al., 1992, 1993; Hoell et al., 1993; Thornton et al., 1996; Blake et al., 2008, etc). The airborne survey flight data are designed to sample background air at set locations on a regular basis over long time periods and are part of the NOAA/ESRL/GMD Global Greenhouse Gas Reference Network's aircraft program (http://www.esrl.noaa.gov/gmd/ccgg/aircraft/index.html, last access: 6 June 2018; an update of results published in Montzka et al., 2007). This data collection started in 1999 at a range of locations and has been used extensively in analysis of the continental US carbon budget (e.g., Campbell et al., 2008; Hilton et al., 2017). OCS has been measured at 10 globally distributed sites in the AGAGE network using the MEDUSA GC-MS. The data for the Jungfraujoch site are presented in Lejeune et al. (2017). Larger-spatial-scale, shorter-time-interval survey flights include the HIPPO (2009–2011) and ATom (2016–2018) airborne programs, which predominantly sample OCS over remote marine locations. Atmospheric chemistry flights are designed to understand chemical processing and pollution transport and include sampling as part of pollution transport across the Pacific (e.g., Pacific Exploratory Mission–West A (PEM-A); Thornton et al., 1996) or Transport and Chemical Evolution over the Pacific experiment (TRACE-P), which sampled Asian outflow dominated by anthropogenic OCS emissions in 2001 (Blake et al., 2004). Other projects included sampling of OCS over continents (e.g., over the US in 2004; Blake et al., 2008).

OCS measurements have been made from tall towers using flasks and subsequent analysis by GC-MS. Most long-term tall-tower observations have been conducted as part of the NOAA/ESRL/GMD tower network (Montzka et al., 2007). These data from 11–12 sites include continuous sampling from 2000 onward at a daily or twice-daily time basis for most of the record.

Three approaches have been used to quantify ecosystem fluxes of OCS: chamber measurements, gradient measurements, and eddy flux covariance measurements. While researchers have been quantifying OCS measurements with chambers for decades, most field outings prior to 1990 used dynamic chambers with sulfur-free sweep air, artificially inducing high emissions (Castro and Galloway, 1991).

Measurements from towers have been made in a variety of ecosystems. An OCS analyzer capable of determining ambient OCS and CO2 concentrations at 10 Hz is commercially available (Kooijmans et al., 2016; Commane et al., 2013; Stimler et al., 2010a), allowing for eddy flux covariance measurements (Asaf et al., 2013; Billesbach et al., 2014; Commane et al., 2015; Wehr et al., 2017). With this powerful new tool, traditional methods of partitioning carbon fluxes over ecosystems can be directly compared to using OCS data as a proxy for GPP in situ. A few studies have made use of the gradient method (Berresheim and Vulcan, 1992; Blonquist et al., 2011; Rastogi et al., 2018).

OCS measurements in the surface ocean comprise about 6000 ship-based measurements. These samples are usually taken at a depth of 0–5 m below the ocean surface and analyzed by gas chromatography with various detectors or off-axis integrated cavity output spectrometry. Table 3 gives an overview on available measurements. A central database for ship-based OCS measurements is desired to derive global patterns and facilitate model comparison. Measurements of the precursor gas CS2 are scarcer than OCS measurements. Samples for CS2 are taken usually in a similar way to OCS samples from the same depth range and analyzed using gas chromatography and mass spectrometry detection.

Different hydrolysis rates apply for OCS trapped in bubbly ice vs. clathrate (bubble-free) ice. Some ice core material is not suitable for OCS analysis because the environment was too warm for long periods and OCS was hydrolyzed at high rates for thousands of years. Aydin et al. (2014, 2016) developed the necessary corrections to take into account OCS hydrolysis within the ice core bubbles. Corrected data are published for Taylor Dome, the West Antarctic Ice Sheet Divide, and Siple Dome (Aydin et al., 2016). Firn data are available for more recent time periods (Montzka et al., 2004; Sturges et al., 2001).