Total ozone columns are retrieved from GOME-2 (ir)radiance spectra using the DOAS method in the wavelength region 325–335 nm (Huggins absorption band). The fitting includes an effective temperature for the ozone absorption, an NO2 absorption cross-section, and scaling factors for interference due to the Ring effect. The retrieved ozone slant column densities are converted to vertical columns using an iterative air mass factor. Furthermore, a cloud correction is applied in the vertical column retrieval for (partially) cloudy scenes. Cloud parameters are retrieved from GOME-2 using the OCRA and ROCINN algorithms . OCRA provides the cloud fraction using the PMD measurements and ROCINN provides cloud height and cloud albedo using the oxygen A-band measurements. The cloud parameters retrieved from OCRA and ROCINN have been validated and verified by comparing with other measurements and other cloud algorithms . To reduce the impact of the instrument degradation, the degradation correction factors have been implemented in both the OCRA and ROCINN algorithms . Note that the OCRA and ROCINN cloud parameters are also used for the cloud correction of the other minor trace gas column products (NO2, BrO, HCHO, SO2 and water vapour) as well as for the determination of the tropical tropospheric ozone column product.

A detailed description of the GOME-2 total ozone column algorithm can be found in , and . Figure  shows one example of the total ozone product: The Antarctic ozone hole for 2 October 2014 measured from GOME-2A and GOME-2B when the minimum total ozone column reached was around 120 DU.

The total ozone NRT and offline products from GOME-2A and GOME-2B are periodically validated and routinely monitored for their quality. For this purpose, a dedicated web-portal is available at http://lap3.physics.auth.gr/eumetsat. In this portal, online comparisons with quality assured ground-based total ozone measurements are available. The reference data for the offline total ozone comparisons are archived Brewer, Dobson and M-124 total ozone data; these are deposited at the World Ozone and UV Data Centre (http://www.woudc.org) and are employed after being thoroughly controlled for their quality. These online comparisons are routinely and automatically updated every month. For the near-real-time total ozone comparisons, near-real time Brewer and Dobson total ozone data, deposited in the World Meteorological Organisation Ozone Mapping Center (http://lap.physics.auth.gr/ozonemaps2/), are downloaded and compared to the GOME-2A and GOME-2B near-real-time observations on a weekly basis.

Comparison results for the GOME-2A offline data for the period 2007–2011, processed with GDP4.4 , have been previously presented by , indicating that GOME-2 total ozone data agree at the 1 % level with the ground-based measurements as well as other satellite instrument data sets, showing a small dependency on solar zenith angle for angles less than 75∘ and almost no dependency on cloud fraction and cloud top pressure. Results from an initial validation of the GOME-2B offline data, processed with GDP4.7 for the year 2013, have been presented in . These results show excellent consistency with coincident GOME-2A total ozone measurements. In Fig.  we present an updated time series of the differences between GOME-2A and GOME-2B vs. Dobson and Brewer total ozone data, averaged over the Northern Hemisphere for the period 2007–2014. This confirms the consistency between the two satellite data sets as well as their long-term stability. The use of BDM (Brion, Daumont and Malicet) absorption cross sections in the GDP4.7 version of the algorithm explains the overestimation of GOME-2 data, compared with the results presented earlier by . Both satellites are consistent over the Northern Hemisphere with negligible latitudinal dependence, while over the Southern Hemisphere there is a systematic difference of 1 % between the two satellite instruments.

A wide community is using the GOME-2 total ozone product provided by the O3M SAF. Since October 2013, total ozone column data have been assimilated in the Copernicus atmosphere core service project CAMS (Monitoring Atmospheric Composition and Climate) NRT system. The GOME-2 ozone columns are also used in the Integrated Forecast System (IFS) of the European Centre for Medium-Range Weather Forecasts (ECMWF). Furthermore, the GOME-2 total ozone data were used in the WMO/UNEP Scientific Assessment of Ozone Depletion reports of 2010 and 2014. Other users are WMO World Ozone and Ultraviolet Radiation Data Centre, WMO Ozone Mapping Centre, DLR World Data Center for Remote Sensing of the Atmosphere as well as ESA Tropospheric Emission Monitoring Internet Service (TEMIS).

The NRT and offline vertical ozone profile product uses reflectance measurements from 260 to 330 nm and iteratively finds the ozone profile best matching the original radiance measurements using optimal estimation. The ozone profiles are given as partial ozone columns in DU in 40 logarithmically spaced layers from the surface up to 0.001 hPa. For (partially) cloudy scenes, the retrieved cloud pressure replaces the nearest pressure level. The cloud pressure is retrieved by the FRESCO algorithm from the O2 A-band measurement of GOME-2 . GOME-2 has nominal spatial resolution of 80 km × 40 km, but for the shortest UV wavelengths the integration time is 8 times longer. The initial O3M SAF ozone profile product was provided at a spatial resolution of 640 km × 40 km (coarse resolution, CR) but has since been improved to the 80 km × 40 km resolution using radiance measurements from a mix of integration times. As an example of the ozone profile product from GOME-2B, Fig.  shows a vertical cross section along an orbit from north to south. The measurements end in a location over Antarctica where the Southern Hemisphere spring ozone hole is present, indicated by low ozone values in the stratosphere.

The ozone profile products are validated with ozone sonde data as well as with lidar and microwave measurements. The ozone sonde data are used in particular for the tropospheric and the lower stratospheric part up to about 30 km, whereas lidar and microwave validation focus on the stratosphere from 15 to 55 km. The ozone sonde data come from the World Ozone and Ultraviolet Radiation Data Center (WOUDC), the SHADOZ (Southern Hemisphere ADditional OZonesondes) network, NDACC (Network for the Detection of Atmospheric Composition Change) and the NILU's Atmospheric Database for Interactive Retrieval (NADIR) at Norsk Institutt for Luftforskning (NILU) (http://www.nilu.no/nadir/). The NDACC network also provides the data of four lidar and five microwave stations. These are the only NDACC stations which deliver ozone profile data regularly and with a comparatively small delay. A comprehensive description of the ground-based validation methods and results for the upper stratosphere will soon be published elsewhere.

The validation and quality monitoring is done by trending of monthly mean values and by direct comparison of satellite and ground-based data. The regular validation uses ground-based ozone profile measurements and considers the dependence of the deviations between satellite and ground-based profiles on total ozone, scan angle, solar zenith angle, cloud fraction and distance in space and time. All comparisons are performed for the high- and coarse-resolution satellite profiles, respectively. The complete set of results is regularly available at the O3M SAF validation internet site.

Overall target values are met in the troposphere (30 %) and the lower stratosphere (15 %), not taking into account the UTLS zone, which shows more elevated relative differences, which cannot be assigned to the troposphere or to the stratosphere. However, GOME-2 shows a clear underestimation of ozone concentrations above about 30 km (Fig. ). The difference to ground-based data increases with altitude, amounting up to about 25 % at 55 km for GOME-2B and up to about 55 % for GOME-2A, depending to some degree on ground-based station and instrument. Furthermore, a satellite to satellite comparison with the Global Ozone Monitoring by Occultation of Stars (GOMOS), Optical Spectrograph and Infrared Imager System (OSIRIS) and Microwave Limb Sounder (MLS) confirms these conclusions ().

Degradation of the GOME-2A instrument is clearly visible in the ozone profile products as a decrease in retrieved ozone concentrations at most altitudes over the years of the mission (Fig. ). Especially for the higher stratospheric part (above 30 km), such a decrease could be identified. For the coarse-resolution profiles, this trend can be as large as 75 % per decade in the upper stratosphere above 30 km. GOME-2B data (red lines in Fig. ) do not yet show this effect.

Using a comparison between measured and expected/simulated spectra, a multiplicative bias factor can be established with which the measurement needs to be corrected. Figure  is an example of the bias factor for 267 and 329 nm for GOME-2A. Clearly visible is that there was an initial bias at 267 nm of roughly 8 %, increasing to 60 % in 2013 and that while the bias at 329 nm was initially minimal it has increased to about 10 % in 2013. The implementation of this degradation correction in the operational retrieval will be an important part of the algorithm development in the near future.

NO2 plays a key role in air quality and atmospheric chemistry. It is an air pollutant affecting human health and ecosystems, and an important ozone precursor. In addition, NO2 is involved in ozone depletion processes in the stratosphere and it is important for climate change studies, because of the indirect effect on the global climate. Total NO2 columns, including a tropospheric and stratospheric component, are retrieved with the DOAS method in the VIS wavelength region 425–450 nm . Tropospheric NO2 columns are obtained from the total columns by estimating the stratospheric content and removing it from the total amount. Several methods exist for the stratosphere estimation: , and . A spatial filtering approach is used by masking potentially polluted areas and then applying a low-pass filter in the zonal direction . After the stratosphere–troposphere separation, the tropospheric NO2 column is calculated using tropospheric air mass factors based on monthly average NO2 profiles from the MOZART-2 chemistry transport model, and a cloud correction for partially cloudy conditions is applied (see also Sect. 5.1). Figure  shows the average tropospheric NO2 columns over East Asia measured by GOME-2A for the period 2007–2013. The world's largest area with high NO2 pollution is found above eastern China, which is a result of China's spectacular economic growth during the last decade, accompanied by a strong increase in emissions of air pollutants. Another remarkable feature visible in Fig.  is the enhanced tropospheric NO2 along shipping lanes in the Bay of Bengal and the South China Sea, see also .

The near-real-time and offline total and tropospheric NO2 products are regularly validated and monitored by comparing to (1) data sets acquired by other satellites (GOME, SCIAMACHY, OMI) and GOME-2 retrievals performed with other processors (such as the TEMIS system, http://www.temis.nl) and (2) ground-based reference measurements acquired by UV-visible DOAS zenith sky looking spectrometers performing network operation in the framework of the Network for the Detection of Atmospheric Composition Change (NDACC, http://ndacc.org) and zenith-sky looking spectrometers for the NO2 stratospheric column and multi-axis (MAX-DOAS) spectrometers for the NO2 tropospheric column. Detailed validation results for GOME-2A can be found in . The results for the validation of GOME-2A and B NO2 columns can be found in and on the O3M SAF BIRA validation pages (http://cdop.aeronomie.be/validation/valid-results).

The step by-step verification of the operational GOME-2B product against GOME-2A, and GOME-2A and B alternative retrievals performed in , has highlighted a global underestimation by GOME- 2B slant columns of 1 × 1015 molec cm-2 with respect to GOME-2A, which translates into a bias of 0.15 × 1015 and 0.4 × 1015 molec cm-2 in total and stratospheric columns. GOME-2B tropospheric column data underestimate GOME-2A by less than 0.5 × 1015 molec cm-2 in moderately polluted conditions, while larger differences (up to 8 × 1015 molec cm-2) occur in polluted regions. The latter can be explained by the local time difference between GOME-2A and GOME-2B overpasses and the associated impact of the variability in NO2 content and cloud cover on the comparison results. Figure  presents, from pole to pole, the median bias and spread for total NO2 between GOME-2A/B total NO2 and correlative ground-based measurements acquired by 25 zenith-sky DOAS instruments. Prior to the comparisons, GOME-2/NDACC co-located data are filtered to avoid GOME-2 data contaminated by tropospheric pollution, and corrected for photochemical diurnal effects (arising from solar local time differences between NDACC and GOME-2 observations). Only the latest, consolidated versions of the NDACC data are considered for Fig.  (no NRT data), and only the GOME-2 forward pixels (no backward pixels) are considered. Figure  illustrates the good agreement between the different stratospheric NO2 data sets, usually within 0.1–0.5 × 1014 molec cm-2, a value close to the combined uncertainty of the comparison method and within the target accuracies for total NO2 of 3–5 × 1014 molec cm-2 in unpolluted conditions (Table 1). Nevertheless, the slight bias between GOME-2A and GOME-2B data is also seen in this figure with GOME-2B, underestimating GOME-2A by about 1–3 × 1014 molec cm-2 (bias propagated from the retrieval of slant column densities), as well as another slight bias of 2–5 × 1014 molec cm-2 between comparison results averaged over the Northern Hemisphere and those averaged over the Southern Hemisphere.

Figure  presents some examples of validation results for tropospheric NO2 GOME-2 columns based on BIRA-IASB MAXDOAS stations. As can be seen in Fig. a, very different conditions of tropospheric NO2 are sampled at the four stations, from clean/remote region (OHP, south of France), city (Uccle, Belgium) and heavily polluted region in Beijing and Xianghe in China (just outside the city, at 60 km south-east of Beijing). Figure b presents the scatter plots of the monthly means at the four stations. Good correlations between GOME-2A and the ground-based MAXDOAS data are obtained, both in terms of correlation coefficients R (ranging from 0.65 in Uccle to 0.92 in Beijing) and slopes of the regression analysis S, with values ∼ 0.8 at Xianghe, slightly smaller in Uccle and OHP (∼ 0.6) and with larger differences (S = 0.47) in Beijing. The impact of the location of the ground-based instrument can be seen when comparing the results between Beijing and Xianghe, that are only 60 km apart: in the first case the MAXDOAS is situated in the centre of the Beijing megacity and then it has been moved to the Xianghe site outside the city, in a zone more representative of what is seen by a satellite pixel. This large differences in Beijing compared to results in Xianghe are due to a representation error for stations in urban locations, affected by local pollution episodes, not seen in the averaged GOME-2 pixel. This representation error has been also highlighted in larger scale comparisons started in for total, stratospheric and tropospheric GOME-2 NO2 columns compared to DirectSun, ZenithSky and MAXDOAS data sets. Figure c presents an example of the time series of GOME-2A and B above OHP and its comparisons to ground-based MAXDOAS tropospheric NO2 data, extending the results presented in . The pollution episodes are captured well by both GOME-2 instruments and the comparisons of monthly averaged columns show consistent seasonal variations, with high NO2 in winter and low NO2 in summer.

The GOME-2 tropospheric NO2 columns are used in the CAMS NRT system and in support of regional (e.g. European) air quality models, for example. The NO2 products from the O3M-SAF are also used by a number of research institutes for various applications such as verification of emissions, investigation of regional and global trends, and assessment of chemistry-transport models.

Atmospheric formaldehyde (HCHO) is an intermediate product common to the degradation of many volatile organic compounds, and therefore it is a central component of tropospheric chemistry. While the global formaldehyde background is due to methane oxidation, emissions of non-methane volatile organic compounds (NMVOCs) from biogenic, biomass burning and anthropogenic continental sources result in important and localised enhancements of the HCHO concentration. Improving our knowledge of NMVOC emissions is essential for a better understanding of the processes that control the production and the evolution of tropospheric ozone, a key actor in air quality and climate change, but also of the hydroxyl radical OH, the main cleansing agent of our troposphere. As the lifetime of HCHO is typically a few hours, enhanced concentrations serve as tracers for reactive NMVOC emissions (Fig. ). Moreover, HCHO satellite observations are used in combination with tropospheric chemistry transport models to constrain NMVOC emission inventories in so-called top-down inversion approaches. Satellite HCHO and NO2 observations are also combined to estimate VOC / NOx ratios, and thereby the local chemical regime of tropospheric ozone production. The HCHO retrieval is based on the DOAS technique and follows the developments published in . Common settings are used for GOME-2A and GOME-2B retrievals. The HCHO slant column inversion is performed in the UV wavelength range 328.5–346 nm, and includes a post-treatment for systematic biases based on a latitude dependent offset correction. Tropospheric HCHO vertical columns are calculated using air mass factors that depend on the solar and viewing geometries, the surface albedo, cloud fraction and cloud top height (OCRA/ROCINN product), and the HCHO profile shape which is derived from three-dimensional chemical-transport model simulations obtained using the IMAGES model .

For the validation of the HCHO columns, a step-by-step verification of the operational product against the reference scientific GOME-2 HCHO retrieval algorithm has been performed in and confirms the validity of the target accuracy of 50 % in polluted conditions (Table 1). An example of this verification is given in Fig.  for two different emissions regions: northern China and Amazonia. GOME-2B HCHO slant columns, fit residuals, and scatter are comparable to those obtained from GOME-2A spectra at the beginning of the mission in 2007, both for the operational and scientific products. The effect of Metop-A degradation (increased noise caused by decreased throughput, see ) is clearly visible in Fig. , with the DOAS fit RMS increase with time for both operational and scientific products. Differences between operational and scientific data sets are within 40 and 28 % for GOME-2A since 2007 and within 20 and 13 % for GOME-2B since December 2012, for Amazonia and northern China, respectively. GOME-2 A and B GDP 4.7 HCHO retrievals are in very good agreement with the reference scientific retrievals when using the same settings (not shown here). The reduced HCHO slant column noise level in the scientific algorithm presented here (v14) is obtained by pre-fitting BrO in a separate larger interval (328.5–359 nm) which allows de-correlation of HCHO and BrO differential absorption structures . Remaining differences between the operational and scientific HCHO vertical columns are mainly related to the different input parameters used for the air mass factor calculation, namely the cloud product and the surface albedo. An improved version of the HCHO operational product addressing both GOME-2 A and B is under preparation based on recent scientific algorithm developments and will be released with the next version of the GDP data product (GDP 4.8). In addition to the verification against the scientific prototype, comparisons with other HCHO satellite data sets, such as GOME, SCIAMACHY and OMI instruments as well as direct comparisons with ground-based MAXDOAS and/or FTIR instruments are planned to be performed in the near future.

The main user of the O3M SAF NRT and reprocessed GOME-2 H2CO products is CAMS (Copernicus Atmosphere Monitoring Service).

The main sources for SO2 in the Earth's atmosphere are anthropogenic emissions and volcanic eruptions. During volcanic eruptions, SO2, ash and various gases are emitted into the atmosphere. SO2 is thus a robust indicator for volcanic activity which can serve as a proxy for the potential emission of volcanic ash that is initially collocated with the emitted volcanic SO2. The timely and global measurement of volcanic SO2 can thus provide critical information for aviation hazard mitigation when it is used by the Volcanic Ash Advisory Centers (VAACs) . Volcanic SO2 has an impact on the local air quality affecting human health and ecosystems and threatening aviation safety since SO2 causes sulfidation in the aircraft engines which might lead to a total engine failure if exposed over a long period of time. Volcanic SO2 also has an effect on the global climate. When released into the atmosphere it is subject to wet and dry decomposition as well as oxidisation to sulfate aerosols .

In the lower troposphere SO2 and sulfate aerosols have a lifetime of a few days. When injected into the stratosphere by, e.g. explosive volcanic eruptions, the SO2 lifetime is several weeks, whereas sulfate aerosols can remain for over a year , affecting the Earth's radiative forcing by reflecting the solar irradiation and by changing the albedo and lifetime of clouds. For example the 1991 Pinatubo eruption caused an average global cooling of about 0.3–0.5 ∘C for years (see e.g. ). Global SO2 emissions averaged over the year 2015 are shown in Fig. .

The DOAS slant column retrieval of SO2 is performed in the UV wavelength range 315–326 nm . The retrieved SO2 slant columns are corrected for any systematic bias by applying a latitude and surface-elevation dependent offset correction. Total vertical SO2 columns are calculated by applying an air mass factor which depends on the viewing geometry surface conditions and the SO2 profile shape. Since at the time of the measurement the SO2 plume height is unknown, the total vertical column is calculated for three scenarios: the passive degassing of volcanoes and anthropogenic pollution (plume height: 2.5 km); passive degassing of high altitude volcanoes and weak eruptions (6 km); and explosive volcanic eruptions (15 km). Furthermore, in order to detect volcanic eruptions, pixels with elevated SO2 columns are automatically flagged when they fulfill certain threshold criteria . This is especially important for the assimilation of the GOME-2 SO2 columns in forecast models, such as in the CAMS system. Note that this flag will be provided to the users with the release of the next version of the SO2 product (GDP4.8), which will be released in early 2016.

The validation of GOME-2 SO2 columns currently relies on inter-comparisons with other satellite data sets, such as SCIAMACHY/Envisat and OMI/Aura as well as inter-evaluation between the two GOME-2 instruments. The current comparisons show that the GOME-2 SO2 data are within the 50 % target accuracy (Table 1). An example of an inter-comparison of GOME-2A and GOME-2B for the Copahue volcanic eruption on the 23 December 2012 is given in Fig. , left, showing very consistent results between the two sensors for their co-located measurements. Figure , right, shows an example of anthropogenic SO2 comparisons, for a region in Eastern China centralised over Beijing. The OMI/Aura PBL (planetary boundary layer) product (Version 003, ) which assumes an SO2 loading with a centre of mass altitude of about 0.9 km , is compared to both GOME-2A and GOME-2B observations considering an SO2 plume height at 2.5 km. A very good correlation between GOME-2A and B results is shown in the scatter plot (blue line) with r-squared of 0.74, to be expected since the same algorithm is applied to both sensors. However, the GOME-2B retrievals result in lower SO2 columns than the OMI retrievals for loadings over 0.5 DU as shown in the equivalent scatter plot (red line), even though the absolute correlation is very satisfactory at 0.80. The reason for this difference may be attributed to the different assumptions on the a priori SO2 profile used by the two different algorithms (see , for more details). In the future, comparisons with ground-based MAXDOAS instruments at different global locations are envisaged for strong anthropogenic and volcanic SO2 cases as this type of comparison allows the dependence of the retrievals on the SO2 profile shape to be taken into account.

The main users of the GOME-2 SO2 product are the volcanic hazard and air quality communities. The GOME-2 SO2 product is used within the Support to Aviation Control Service (SACS) (Brenot et al., 2014), that provides volcanic plume information to the Volcanic Ash Advisory Centers (VAACs) and other users. The GOME-2 SO2 columns are also used within the GMES service EVOSS, which monitors volcanic activity and relevant hazards at a global scale using Earth observation data products. Furthermore, the GOME-2 SO2 columns have been used within the mobile volcano fast response system Exupéy .

BrO is present in the lower stratosphere, where it depletes ozone through very efficient catalytic reactions. BrO is also released in the troposphere from sea ice, at the interface of ocean, salt lakes and volcanoes.

The retrieval of total BrO is achieved by a DOAS analysis in the wavelength region 332–359 nm , including five BrO absorption bands. It is followed by an equatorial offset correction applied to the retrieved slant columns, in order to minimize the impact of the long-term instrumental degradation. The conversion into total BrO vertical columns is made using air mass factors assuming representative stratospheric BrO profiles. The main interest is on bromine emissions in the polar boundary layer. GOME-2 measurements allow springtime BrO columns to be studied in both the Arctic and Antarctic regions (as illustrated in Fig. ). BrO production is related to heterogeneous photochemistry due to the presence of sea salt, but the mechanisms are currently not fully characterized, especially the link with meterological variables (e.g. wind) and the importance of bromine for the deposition of mercury which impacts the polar biosphere. Owing to its wide coverage, GOME-2 measurements enable long-range transport of polar tropospheric BrO to be studied, and also emissions from volcanic eruptions . In addition there is an interest in the study of stratospheric BrO and its long-term trend. However, strictly speaking, scientific studies of BrO from GOME-2 measurements are only possible by separating the retrieved total BrO column into its tropospheric and stratospheric components (see Sect. 6).

The validation of BrO columns relies on the comparison of the total columns to other satellite data sets, such as the SCIAMACHY and GOME-2 scientific products and on ground-based zenith-sky DOAS instruments from NDACC stations, such as Harestua . The results for the validation of GOME-2A and B BrO total columns can be found in and on the O3M SAF BIRA validation pages (http://cdop.aeronomie.be/validation/valid-results). Examples are shown in Fig. . One can see that the satellite columns are very consistent with each other at all latitudes and seasons. The comparison of GOME-2A and GOME-2B with ground-based BrO columns at Harestua is also good for the absolute values and for the short-term variation of the BrO total column. The conclusion is that the GOME-2A and -B total BrO columns have an accuracy better than 30 % most of the time and hence the product reaches its target accuracy.

Atmospheric water vapour plays a major role for meteorology as well as for climate because it is an important greenhouse gas and via its influence on the formation of clouds and precipitation and the growth of aerosols . Hence, advancing in understanding of variability and long-term changes in water vapour is vital, especially considering that atmospheric water vapour exhibits a highly variable spatial and temporal distribution. Total column water vapour (TCWV) from GOME-2 is the only data product with the following combination of features: global coverage over land and over sea, good sensitivity down to the surface (where most of the water vapour column resides), independent of model input (no model-dependent information in climatological data record), and retrievals insensitive to instrument drift. Thus, the GOME-2 water vapour product is especially valuable for long-term series and climatological studies . The example of the retrieved monthly mean maps of total column water vapour in Fig.  shows the global distribution of TCWV in February and August 2008 retrieved by GOME-2A. The algorithm used for the retrieval of GOME-2 TCWV is based on a classical differential optical absorption spectroscopy (DOAS) performed in the wavelength interval 614.0–683.2 nm and does not include explicit numerical modelling of the atmospheric radiative transfer . It consists of three basic steps: in the first step, the spectral DOAS fitting is carried out, taking into account absorption by O2 and O4, in addition to that of water vapour. In the second step, a correction for absorption non-linearity effects is applied because the highly fine-structured H2O (and O2) absorption bands cannot be spectroscopically resolved by the GOME-2 instrument. In the last step, the corrected water vapour slant columns determined with the DOAS fitting are converted to geometry-independent vertical column densities (VCDs) through division by an appropriate air mass factor (AMF), which is derived from the measured O2 absorption.

To assess the quality of the GOME-2 water vapour columns, the TCWV products were compared with corresponding model data from the European Centre for Medium Range Weather Forecast (ECMWF) ERA-Interim reanalysis and with SSMIS satellite F16 measurements during the full period January 2007–June 2014 . A comparison between the GOME-2A product and the combined SSM/I + MERIS GlobVapour data set in 2007 and 2008 was also carried out. Good general agreement was reported between all three data sets with mean bias within 0.035 g cm2, although some seasonal and regional biases have been identified. Figure  shows a time series of globally averaged total bias in the TCWV distribution between GOME-2A and the data sets described above. Since January 2013, the bias between the most recent GOME-2B results and the ERA-Interim and SSMIS retrievals was also computed. It was found that the combined SSM/I - MERIS sample and the ERA-Interim data set are typically drier than the GOME-2 retrievals (0.032 and 0.035 g cm-2, respectively), while on average GOME-2 data overestimate the SSMIS measurements by only 0.006 g cm-2. However, the size of these biases is seasonally dependent. Monthly average differences as large as 0.1 g cm-2 were observed in the analysis against SSMIS measurements, but were not so evident in the comparison with ERA/Interim and SSM/I + MERIS data, because of the compensating effect of having land and ocean retrievals. Pronounced negative biases over land areas were identified in regions with high humidity or a relatively large surface albedo (0.3–0.5).

The water vapour product has also been validated against ground-based data . In that study, the radiosonde data were from the Integrated Global Radiosonde Archive (IGRA) maintained by National Climatic Data Center (NCDC) and screened for soundings with incomplete tropospheric column, whereas the ground-based GPS observations from COSMIC/SuomiNet network were used as the second independent data source. The outcome described in the paper was that the general agreement between GOME-2 and the ground-based observations is good. The median relative difference to radiosonde observations was found to be -2.7 % for GOME-2A and -0.3 % for GOME-2B. Against GPS observations, the median relative differences were 4.9 and 3.2 % for GOME-2A and B, respectively. For water vapour total columns below 10 kg m-2, large wet biases were observed, especially against GPS observations. Conversely, at values above 50 kg m-2, GOME-2 generally underestimates both ground-based observations.

The Absorbing Aerosol Index (AAI) product is derived from the reflectances measured by GOME-2 at 340 and 380 nm . The AAI may be used to detect absorbing aerosols over both land and sea surfaces, even in the presence of clouds. The AAI retrieval algorithm compares the measured reflectances with simulated clear-sky reflectances and calculates a residue from these which represents the difference between measurement and (clear-sky) simulation and takes the form of a single index.

A high radiometric stability of the Level-2 product is extremely important for the AAI data record as a whole, because the AAI is very sensitive to changes in the radiometric calibration of the instrument. Before calculating the AAI from the Earth reflectances, the reflectances are first corrected for the effects of instrument degradation. For this pre-processing correction we make use of the method introduced in for the SCIAMACHY instrument that was later applied to the GOME-2 instrument . This correction for instrument degradation is a function of time, wavelength and scan mirror position, and it is instrument-specific. After application of the correction, the errors caused by instrument degradation are removed to within 0.1 index point over the entire time period covered by the satellite.

In the algorithm, the surface albedo is assumed to be constant over the spectral range of the two selected wavelengths, and when the total ozone column is taken into account in the simulations, deviations from the Rayleigh scattering atmosphere are caused purely by clouds and/or aerosols. Negative values for the AAI are caused by the presence of clouds and/or scattering aerosol in the scene. However, a positive value for the AAI can only be explained by the presence of absorbing aerosols. This makes the AAI ideally suited for aerosol masking and the detection of smoke from forest fires or volcanic eruptions. An example of the detection of a volcanic plume is shown in Fig. , where the ash plume originating from the Puyehue volcano can be seen to travel from its source eastward over the south Atlantic Ocean. This event has been analysed in .

For the validation of the GOME-2 AAI product we rely on comparisons with AAI products derived from other satellite instruments, such as GOME-1, SCIAMACHY and OMI. The validation results show that the GOME-2 AAI has reached the target accuracy of 0.5 index point . Since the launch of Metop-B, the AAI products from GOME-2A and GOME-2B can be compared, and the global means of the two AAI products are nearly identical, with GOME-2B having AAI values on average 0.2 index points higher than GOME-2A. Both AAI products meet the target requirements. More detailed validation results can be found in the mentioned validation report.

The Lambertian-equivalent reflectivity (LER) of the Earth's surface is based on reflectance measurements taken by GOME-2A. It is the improved follow-up of earlier surface LER databases based on GOME-1 (on the ERS-2 satellite) and OMI (on the Aura satellite). As a result of this, the algorithm was influenced strongly by the approaches followed for the GOME-1 surface LER database and the OMI surface LER database . The surface LER is an essential input parameter for the retrieval of many trace gases, cloud properties and aerosols.

The GOME-2 surface LER is retrieved from the main science channels for 15 pre-defined 1 nm-wide wavelength bands ranging from the UV to the near-infrared. Additionally, we provide a surface LER product based on the PMD (polarisation measurement device) measurements. The definition of these PMD bands is fixed but the advantage is the much smaller footprint size (10 km × 40 km instead of 80 km × 40 km) which results in better statistics. The surface LER spectra are provided for each month of the year in a grid of 1∘×1∘ (MSC-LER) or 0.5∘ × 0.5∘ (PMD-LER).

The retrieval algorithm follows an approach in which the reflectance measurements from the entire mission are transformed for each of the months into scene LERs. The scene LERs are distributed into the grid cells of the Earth grid and their histograms are analysed to find the representative (cloud-free) scene LER, i.e. the spectral surface LER. For some surfaces the mode of this distribution is taken (e.g. snow, desert), for others the minimum value is taken (e.g. ocean). Cloud contamination over the oceans caused by persistent cloud presence is corrected by looking for near-by grid cells that can act as a surface LER donor. The surface LER is retrieved for land and sea surfaces, including those covered by snow and ice.

Artificial trends in the Earth reflectances due to instrument degradation are removed before the actual surface LER retrieval starts. The method that is used is described in Sect. 5.8. A more extensive description of the algorithm and the products can be found in . An example of the surface LER product is presented in Fig. .

Offline surface ultraviolet (UV) radiation products (OUVs) contain the most important quantities of the solar radiation that can be harmful to life and materials on the Earth. These quantities include daily doses and maximum dose rates of integrated UVB and UVA radiation together with values obtained by different biological weighting functions, solar noon UV index and quality control flags. Recent additions are the photolysis frequencies of two key reactions in the atmospheric chemistry of the troposphere: the photodissociation of ozone and nitrogen dioxide. The photolysis frequency of ozone refers to the rate constant jO(1D) of the following reaction forming atomic oxygen in its exited 1D state from ozone O3+hν(λ<320nm)→O(1D)+O2. This is an important photodissociation route of ozone leading to production of the hydroxyl radical, a key species in oxidation of hydrocarbons in the troposphere. Figure  shows, as an example, the daily maximum jO(1D) at ground level on 10 April 2015.

The OUV products are derived from the O3M SAF total ozone column product and the visible channel reflectances of the third Advanced Very High Resolution Radiometer (AVHRR/3), therefore combining data from two different instruments aboard the Metop satellites . Sampling of the diurnal cloud cycle is improved by complementing the morning orbit Metop AVHRR/3 data with corresponding afternoon orbit data from the Polar Orbiting Environmental Satellites (POES) of the National Oceanic and Atmospheric Administration (NOAA), available through the data exchange between EUMETSAT and NOAA in their Initial Joint Polar System (IJPS) programme. The offline UV products are currently used by research institutes in Europe and Australia and the main interest is in health and long-term applications.

The near-real-time UV product is a 1–5 day forecast of UV index on global scale. The NUV/CLEAR product is based on GOME-2 assimilated total ozone (ATO) from KNMI. The UV index is calculated from ATO applying climatologies for surface albedo, aerosol optical depth etc. The final NUV/CLEAR index is valid for clear-sky conditions at local noon, i.e. minimum solar zenith angle, thus reflecting the maximum UV index to be expected for the current day. The NUV/CLOUD product is the NUV/CLEAR product modified with an algorithm using the cloud cover fraction forecast from ECMWF to reflect the actual UV index. The UV fields are calculated on the same grid size as the input ATO fields, with a sub-pixel resolution of 0.25∘ × 0.25∘. The product may be highly customised for users, with regard to geographical coverage, forecast time etc. The users are National Meteorological Institutes as well as news groups for radiation warnings.

The NUV/CLEAR and NUV/CLOUD have been validated against ground-based measurements of UV index. The most recent validation was performed on data from 2011 for six locations with a total of 2064 measurements . In Fig.  the distribution of the difference between measured and forecasted NUV/CLOUD product is shown. The red line is the mean and the blue line the median of the distribution. The mean absolute relative difference is 22.6 % for the whole NUV/CLOUD comparison, while the NUV/CLEAR product shows a mean absolute relative difference of 8.5 %