Various current and upcoming satellite sensors aim at measuring the atmospheric abundances of the carbon compounds carbon dioxide (CO₂), methane (CH₄) and carbon monoxide (CO). Most of the space-borne instruments such as the GOSAT series (Greenhouse Gas Observing Satellite) (Kuze et al., 2009; Suto et al., 2021), the OCO series (Orbiting Carbon Observatory) (Eldering et al., 2017), the S5P (Sentinel-5 Precursor) (Hu et al., 2018) and S5 series (Sentinel-5), the TanSat program (Yang et al., 2021) and others rely on spectrometric measurements of sunlight reflected by the Earth in the shortwave-infrared (SWIR) spectral range which lends sensitivity to the column-average dry-air mole fractions of the gases (commonly denoted by XCO₂, XCH₄, and XCO). In particular, the upcoming European Copernicus Carbon Monitoring mission (CO2M) (Kuhlmann et al., 2019; Sierk et al., 2019) and various missions that focus on localized targets and specific emission sectors (Strandgren et al., 2020; Jervis et al., 2021) will rely on the same measurement principle.

Measuring the most abundant atmospheric carbon compounds, the general goal of these missions is to contribute to a better understanding of the Earth’s contemporary carbon cycle. In terms of the required precision and accuracy, there are two main categories: Category 1 aims at constraining the natural variability of the carbon cycle on regional-to-continental spatial and on monthly-to-yearly temporal scales [e.g. (Guerlet et al., 2013; Liu et al., 2017; Crowell et al., 2019; Palmer et al., 2019; Western et al., 2021)]. Typically, these variabilities are driven by the world’s large ecosystems through variability in photosynthetic and respiratory activity, wetland dynamics, fire occurrences and other carbon cycle mechanisms. Category 2 targets at budgeting anthropogenic emissions on regional-to-local and monthly-to-instantaneous scales, where the instantaneous scale relates to the usage of individual satellite overpasses [e.g. (Hakkarainen et al., 2016; Nassar et al., 2017; Borsdorff et al., 2018; Pandey et al., 2019; Varon et al., 2019)]. Due to the larger spatial and temporal perspective, category 1 requires that satellite measurements do not show any spurious gradients or trends on the targeted larger scales. Category 2, in contrast, targets at the local-to-regional excess of the compound concentrations above the adjacent background and on direct instantaneous emission rate estimates. Thus, category 2 requires eliminating residual gradients on the local-to-regional scale e.g., within the satellite’s swath and it requires high single-shot precision since individual soundings are to be used.

In order to meet these accuracy and precision requirements – for XCO₂ and XCH₄ on the order of a few permille, for XCO on the order of a few percent – the satellite measurements need careful validation by ground-based instruments that are able to measure XCO₂, XCH₄, and XCO with similar column-average sensitivity as the satellite spectrometers. Currently, the TCCON (Total Carbon Column Observing Network) (Wunch et al., 2011) and COCCON (Collaborative Carbon Column Observing Network) (Frey et al., 2019) form the backbone of routine validation of the satellite-derived carbon compound concentrations. These networks consist of a few dozens of ground-based Fourier Transform Spectrometer (FTS) stations that are equipped with a solar tracker enabling direct-sun absorption spectroscopy. The networks’ coverage mostly serves the category 1 applications since most of these FTS are distributed throughout the world’s continents. There is a gap of validation capacities for direct verification of local-to-regional scale gradients and the respective emission estimates (category 2) and for measurements over the oceans (part of category 1). Addressing the former gap, urban networks and episodic coordinated deployments of the portable COCCON spectrometers are emerging [e.g. (Hase et al., 2015; Luther et al., 2019; Dietrich et al., 2021)] but are far from covering a representative number of cities and hotspot regions. Validation over the oceans is currently mostly limited to stations on islands or at coastal sites which implies that the satellite coincidences are not direct and that differences in topographic height often need to be carefully accounted for. Validation over the oceans should be warranted since typically the satellites observe ocean scenes under glint-view i.e., deliberately pointing towards the specular reflection at the ocean surface in order to gain sufficient signal while, over land, nadir-viewing is preferred. Basu et al. (Basu et al., 2013) reported on land-ocean biases with detrimental impacts for flux inversions that were on the order of a few tenths of a ppm for XCO₂. Such land-ocean biases might be due to scattering by atmospheric particles affecting the lightpath quite differently above ocean and land (Butz et al., 2013).

Here, we review and refine plans how portable FTS of the COCCON type, further developed for mobile deployments, can be valuable assets to the existing XCO₂, XCH₄ and XCO validation capacities, in particular contributing directly to validating local-to-regional scale gradients and measurements over the oceans. The key innovation for making the off-the-shelf FTS suitable for deployment on mobile platforms is a solar tracker system that 1) can find the position of the Sun independent of knowledge on the instrument’s location and orientation and that, once the Sun is captured, 2) is sufficiently agile to track the Sun under (potentially fast) platform motion. Another requirement for routine and autonomous operations is a weather-proof housing. We have developed such a system and demonstrated its performance for continuous operations on ships and for stop-and-go operations on small trucks (Klappenbach et al., 2015; Butz et al., 2017; Luther et al., 2019; Knapp et al., 2021). Section 2 reviews previous campaign deployments, Section 3 summarizes the key instrument parameters and reports on the processing recipe. In Section 4, we focus our discussion on a ship deployment in March 2021 in the Western North Pacific on the research vessel (RV) Mirai which we have not reported before and which serves as an illustrative assessment of the instrument’s capabilities. Finally, Section 5 relates to potential use cases in the context of validating future satellite missions such as CO2M.

In the past years, our mobile FTS has been deployed during five large scale field campaigns listed in Table 1. Three deployments were on ships demonstrating the performance for category 1 applications i.e., for validating the large scale constituent gradients over the oceans. Two deployments were on road vehicles in the vicinity of local emission hotspots evaluating the usefulness for category 2 applications focusing on local-scale concentration gradients and emission rates of localized sources.

Since the first deployment in spring 2014 on RV Polarstern, the setup has substantially matured in terms of solar tracker performance and its ability to be remotely operated and to resist harsh weather. Klappenbach et al. (Klappenbach et al., 2015) and Knapp et al. (Knapp et al., 2021) report on ship deployments in the Atlantic and Pacific oceans which aimed at measuring meridional and zonal cross sections of background XCO₂, XCH₄ and, for the latest instrument upgrade (Hase et al., 2016), XCO over the large oceans. Beside validating the satellite records from GOSAT, OCO-2, and S5P, our concentration records were also used for validating and improving a model used in the Copernicus Atmosphere Monitoring Service (CAMS) (Agusti-Panareda et al., 2017; Knapp et al., 2021). The latest ship deployment took place in the Western Pacific and along the coast of Japan from February 13 to March 24, 2021 when the mobile EM27/SUN was operated onboard the Japanese RV Mirai (framed by a collaboration of Heidelberg University and the National Institute for Environmental Studies (NIES) of Japan). Figure 1 shows the trajectory of RV Mirai and denotes the locations where weather conditions were sufficiently fair to conduct direct-sun measurements. The first part of the cruise, starting out from Shimizu port heading north-east towards the Kamtschatka peninsula, suffered from rough seas with average wave heights of up to 10 m challenging equipment and operating personnel. Weather conditions were better for the second part in the south-east and south of Japan. Therefore, the discussion here concentrates on the latter part with a particular focus on the last 3 days where we collected measurements along the southern coast of Japan observing the outflow from the island before entering Shimizu port.

On land, Butz et al. (Butz et al., 2017) and Luther et al. (Luther et al., 2019) measured emission plumes of CO₂ emitted by the Mt. Etna volcano and plumes of CH₄ from coal mine ventilation in the Upper Silesian Coal Basin (USCB), Poland, respectively. To this end, the mobile EM27/SUN was mounted on a van operated in stop-and-go patterns underneath the emission plumes of the local sources. The instrument collected cross sections of the column enhancements inside the plumes with respect to the background. Luther et al. (Luther et al., 2019) show the van setup and typical plume enhancements for XCH₄ observed downwind of coal mine ventilation shafts. Together with estimates of the local wind conditions and a mass balance method, we succeeded in estimating the instantaneous CH₄ emission rates from individual ventilation facilities.

Following the procedure outlined in Section 3.2, we obtain the XCO₂, XCH₄, and XCO time series for our recent deployment in the Western Pacific and along the coast of Japan on RV Mirai, for which we use the calibration derived from the TCCON side-by-side deployment at Karlsruhe on Oct. 31, 2020 (cf. Table 3). Figure 4 compares our mobile EM27/SUN records to XCO₂, XCH₄, and XCO from the CAMS near-real-time analysis product (Massart et al., 2014; Inness et al., 2015; Massart et al., 2016), and to XCO retrieved from S5P/TROPOMI measurements. Data from OCO-2 and GOSAT were too sparse to allow for a meaningful comparison. Retrievals of XCH₄ over the oceans from TROPOMI are currently not available. For CAMS, we have interpolated the model fields to the locations and sampling times of our ship-borne measurements, and for TROPOMI, we have collected coincidences within a radius of 0.2° and a period of 4 h around our measurement instances. Since we use the CAMS data as a priori in our ground-based retrievals, there is no residual smoothing effects when comparing them. For comparing our data with TROPOMI XCO, we neglected smoothing effects since we consider them to be small. The ground-based and satellite XCO retrievals are performed on the same spectral bands with similar spectral resolution implying similar column sensitivities. Further, the satellite and ground-based retrievals use up-to-data CO prior profiles (from TM5 for TROPOMI) that represent the actual dynamic state of atmosphere.

Our data record is quite sparse in the first half of the campaign due to bad weather conditions but reasonably dense for the later half. We find that the campaign-average hourly standard deviations are 0.33 ppm for XCO₂, 1.5 ppb for XCH₄, and 1.0 ppb for XCO which we take as a measure of the precision and repeatability of the reported hourly means. These precision estimates are consistent with the previous deployments (cf. Table 1 and references therein). While the mobile EM27/SUN sampled mostly background airmasses until Mar. 18, 2021, there is a clear enhancement of a few ppm in XCO₂ (ΔXCO₂) and a few 10 ppb in XCO (ΔXCO) for the last days of the cruise (Mar. 19 to Mar. 22, 2021) where, on the last day, the ship’s trajectory was along the southern coast of the main island of Japan (cf. Figure 1) and the lower tropospheric winds came from North-West directions (as modelled by the HySplit online tool (Stein et al., 2015), not shown). For XCH₄, variations during the campaign are on the order of 20 ppb but there is no clear enhancement pattern for the last days.

Comparing the ship records to CAMS, we find differences of CAMS XCO₂, XCH₄, and XCO amounting on average to 4.1 ± 0.5 ppm (mean difference±standard deviation of differences), −32.1 ± 9.7 ppb, and 0.5 ± 5.8 ppb, respectively. The CAMS data follow the observed variability well, i.e., they show an enhancement of XCO₂ and XCO for Mar. 19 to Mar. 22, 2021, and they generally follow the observed day-to-day variations. But, CAMS XCO₂ is overall high-biased, and CAMS XCH₄ is overall low-biased while CAMS XCO fits well on average. For our previous assessment (Knapp et al., 2021) with measurements across the subtropical Pacific in 2019, we did not find these biases for XCO₂ and XCH₄. However, comparisons of CAMS to TCCON indicate that there might be a time dependent error in CAMS (Schulz et al., 2021) which might explain the larger differences observed here than in (Knapp et al., 2021). These time dependent differences might point to deficiencies in the anthropogenic emissions and natural fluxes (and their temporal variation) that drive the respective concentrations in the CAMS model. Comparing the mobile EM27/SUN data to TROPOMI XCO, we find a mean differences of 9.0 ± 6.6 ppb with TROPOMI showing overall greater scatter than the ship records and confirming the XCO enhancements on Mar. 19 and 22, 2021.

The period Mar. 19 to Mar. 22, 2021 (green box in Figure 4) at the end of the cruise shows correlated enhancements for XCO₂ and XCO. To calculate the enhancements, we define the period Mar. 11 to Mar. 18 as a background period (red box in Figure 4). Averaging the records over the background period yields background concentrations amounting to 415.6 ± 0.4 ppm (mean ± standard deviation over period) for XCO₂ and 110.4 ± 4.0 ppb for XCO. The choice of the background period causes uncertainties of less than 0.3 ppm and 2 ppb, respectively, as checked by repeating the calculations for other periods. Subtracting the background from the measurements during the enhancement period yields the enhancements ΔXCO₂ and ΔXCO. Since the CAMS simulations largely follow the same pattern, we apply the same calculations for ΔXCO₂ and ΔXCO and compare the enhancement ratio ΔXCO/ΔXCO₂ to our measurements (see Figure 5). The enhancement ratios vary between 30 and 10 ppb/ppm with quite some intra-day and day-to-day variability. CAMS agrees reasonably well with EM27/SUN measurements pointing at a reasonable model representation of both, the transport of emission signatures toward our measurement instances and the relative partitioning of the upwind CO₂ and CO emission sources contributing to the enhancement.

Overall, the deployment of our mobile EM27/SUN on RV Mirai confirms the fitness of the instrument for providing reliable XCO₂, XCH₄, and XCO validation measurements over the oceans. On top of sampling background airmasses, we were also able to collect a few days of measurements with XCO₂ and XCO enhancements that point to local-to-regional gradients and source signatures although a confirmed attribution to specific regions using backward trajectories was not undertaken here.

The mobile EM27/SUN is a variant of the FTS operated within the COCCON, supplemented with a fast and flexible solar tracker. If put into operational network-like infrastructures, such versatile and flexible mobile instruments will add substantial value to the validation system for space-borne measurements of XCO₂, XCH₄, and XCO.

Ship deployments of the mobile EM27/SUN have been demonstrated through three campaigns over the past years (cf. Table 1). The implementation of an operational ship-based EM27/SUN could be realized in the near-term. Our system withstands harsh weather conditions and only minor technical updates are required to make it operate semi-automatically under remote control. The system requires access to electric power and network and it must be placed such that it can observe direct sunlight without obstructions from platform structures. Envisaging deployment on a cargo ship, our experience from previous campaigns suggests that the system would need some basic maintenance of 1–2 days duration on a bimonthly basis in the harbor i.e. collecting data, cleaning optical surfaces, conducting ILS measurements, and, occasionally under sunny conditions, gauging XCO₂, XCH₄, and XCO towards the TCCON and COCCON by side-by-side deployments of a transfer standard (e.g., another EM27/SUN). The operations on the ship would require some basic, low-bandwidth network interface to check whether the system works nominally and to allow for altering configuration files if needed. While we do not expect that expert personnel needs to travel along, it would be required to train a person on the ship how to clean the outside surfaces and how to operate a standard measurement e.g., in case of anomalous behaviour after power outages or stormy weather. In the context of CO2M and other satellites, we envision the following general use cases for such ship-based EM27/SUN:

1. A few of such ship-borne spectrometers could be arranged in a travelling network that covers the major ship-routes to serve category 1 validation purposes, i.e., validating the continental-to-regional scale concentration gradients across the world’s oceans which are currently practically devoid of validation systems. The meridional transects across the equator might prove particularly useful to evaluate not just satellite data but also the inter-hemispheric transport of long-lived gases in global models (Agusti-Panareda et al., 2017). Targeting at the validation of the minute background gradients, the periodic side-by-side calibration activities to gauge the records to TCCON and COCCON standards are particularly important to guarantee that the data are traceable the common scale of the World Meteorological Organization (WMO).

The installation and operations of EM27/SUN on road vehicles on land are less straight-forward than the ship-based variant. Our previous campaigns show that the solar tracker supports quick stop-and-go patterns, but it does not support measurements while driving. Typically, for any reasonable driving speed, shocks due to uneven road cover cause disturbances of the solar tracking and the mirror retardation of the interferometer inside the FTS. We expect that facilitating operations while driving requires a gimbal mount of the entire spectrometer system (mass roughly 30 kg) which is a technical development step and investment ahead. Further, measurements from a road vehicle are difficult to conduct in an automatic or remote way implying the need for personnel driving the vehicle and operating the instrument. We see the potential use cases as follows:

3. Mobile EM27/SUN operated on road vehicles can add a versatile and flexible component to stationary regional and local networks of COCCON spectrometers. Operated within urban observatory networks (Dietrich et al., 2021), for example, the mobile spectrometers would allow for measuring column curtains between the nodes of the network to better constrain the unsampled airmasses in-between the stationary instruments and to guarantee the relative calibration of the network by frequent side-by-side measurements with the stationary nodes. Thus, this use case would support category 2 applications by helping validate the XCO₂ (and XCH₄, XCO) gradients that occur on scales of urban agglomerations and by contributing to the validation of emission estimates for the respective region. One might envision a future development of the spectrometer system toward deployment on light-rail trains e.g., sampling urban domes along ring rail tracks. We expect that the current solar tracker would support measurements on driving trains right away in terms of compensating for shocks and vibrations, but the effect of viewing obstructions from the overhead wiring, tunnels, and buildings would need to be evaluated for any particular location. Likewise, deployment on small ships going along channels and rivers might be thinkable for some cities.

The use cases 1 through 4 exemplified above imply some further technical developments and some customizations that depend on the particular platform and field of application but they do not require modifications of the key hardware, the spectrometer and the solar tracker, and software. Developing the mobile system towards the ability to measure more molecular species can enhance the versatility of the setup even further:

• During the deployment at Mt. Etna, we operated a UV/visible DOAS (Differential Optical Absorption Spectroscopy) instrument together with the EM27/SUN FTS (Butz et al., 2017). The two instruments shared the solar tracker and a large part of the operational infrastructures. The UV/visible spectral range gives access to species such as sulfur dioxide (SO₂) and nitrogen dioxide (NO₂) co-emitted with CO₂ from volcanoes and combustion processes, respectively. Due to the small background concentrations, the SO₂ and NO₂ plumes are easier to identify than those of CO₂ and thus, they can serve as tracers for defining plume extent, shape and direction, and the constituent ratios are indicative of the emissions processes. In this spirit, CO2M will carry an NO₂ spectrometer to better contour combustion plumes and to support the CO₂ emission estimates from the respective sources. Thus, supplementing the mobile EM27/SUN with a UV/visible DOAS spectrometer would add the ability to validate NO₂ and the NO₂/CO₂ ratios relevant for the category 2 use cases 2, 3, and 4. Typically, DOAS spectrometers can provide real-time information on the sampled airmasses which is particularly useful for use case 4 to inform on whether one measures inside or outside a plume.

• The spectral range of the mobile EM27/SUN covers the shortwave infrared from roughly 4,000 to 11,000 cm⁻¹ (1.1–2.5 μm) via two detector channels. At Mt. Etna, we demonstrated that this large spectral range also allows for measuring the HCl and HF plumes of the volcano. Like, SO₂ measured by the DOAS spectrometer, HCl and HF are co-emitted with CO₂ and the respective concentration ratios are indicative for volcano-interior processes. Thus, for dedicated applications beyond satellite validation, developments towards standardizing the retrieval of constituents other than the carbon compounds will be useful.

The mobile EM27/SUN is a sun-viewing spectrometer that is able to measure XCO₂, XCH₄, and XCO from ships and road vehicles, the latter operated in stop-and-go patterns. Throughout our deployments, we find a typical precision of few tenths of a ppm for XCO₂, a few ppb for XCH₄ and XCO. Regular side-by-side measurements with TCCON and COCCON spectrometers enable diagnosing drifts and ensure traceability with respect to the networks’ calibration. Thus, the mobile EM27/SUN provides a versatile validation tool for CO2M and other sensors. It can be deployed on moving platforms, it can be relocated quickly, and thus, it can be used for targeted validation campaigns in the context of verifying emission estimates for localized sources. The identified uses cases for ship deployments relate to validating background concentrations over the open oceans and enhancement episodes along coast lines due to continental outflow. Through deployments on road vehicles, the mobile EM27/SUN can validate emission estimates for point sources and, it can contribute a mobile component to regional validation networks by sampling the airmasses between the network nodes and providing an internal calibration tool. Further developments will aim at covering UV/visible absorbing gases such as NO₂ by co-mounting a DOAS spectrometer and exploring the retrieval of other gases such as HCl and HF that are potentially useful for attribution of emission processes. Technically, the system is ready to be operationalized as part of a satellite validation activity.