During the last decade, passive MAX-DOAS (Multi-Axis Differential Optical Absorption Spectroscopy) instruments have been deployed worldwide, focusing on the monitoring of air quality tropospheric trace gas species (NO2, HCHO, SO2, CHOCHO) but also halogens (BrO, IO) and aerosols (through oxygen dimer (O4) measurements). Because they have similar spatial domains, MAX-DOAS is widely used to validate satellite nadir observations of pollutants like NO2, HCHO, and SO2 (see e.g. Hassinen et al. (2016) for the validation of the GOME-2 instruments on board of the METOP-A and B platforms). As for all UV-visible DOAS data products (see e.g. Platt and Stutz, 2008), the MAX-DOAS retrieval is based on a two-step approach: (1) a spectral inversion step using the differential optical absorption spectroscopy (DOAS) method and providing the slant column densities (SCD, which is the trace gas concentration integrated along the effective light path), and (2) a subsequent conversion step which ultimately provides the end products (tropospheric vertical columns and/or profiles).

Compared to other trace gases, tropospheric ozone retrievals are much more challenging since most of the ozone column (90%) is located in the stratosphere and therefore dominates the total ozone absorption, making the separation between the tropospheric and stratospheric ozone absorption signals difficult. Moreover, given the fact that for tropospheric ozone, the spectral fitting is usually done in the Huggins bands (i.e. 300-340 nm), the retrieval problem cannot be considered as linear as for other trace gases, because of the strong ozone absorption in this wavelength range. These difficulties explain why a limited effort has been made to date by the DOAS Community on this topic: so far only the exploratory studies of Liu et al. (2006) and Irie et al. (2011), both based on the Optimal Estimation Method (OEM; Rodgers et al., 2000), and of Gomez et al. (2014) have been reported in the literature.

In Liu et al. (2006), the atmosphere is modeled on an Umkehr-type grid with 22 layers from 0 to ~60 km, in steps of ~2.5 km for each of the bottom 20 layers and ~5 km for the top two layers. The total column ozone is also treated as one element of the measurement vector. The difference between the integrated total column from the ozone profile and the constrained total column estimated from zenith-sky or direct-sun observations is then minimized in the retrievals simultaneously with those between measured and simulated radiances at different elevation angles. The a priori ozone profile used in the retrievals and its standard deviations are extracted from the Total Ozone Mapping Spectrometer (TOMS) version-8 climatology. To extract more available information from the measurements, the a priori constraint is relaxed by increasing the original a priori standard deviations in the troposphere. A correlation length of 5 km is used to construct the a priori covariance matrix for the whole atmosphere. Tropospheric aerosols corresponding to a visibility of 50 km and background stratospheric aerosols from the LOWTRAN climatology are used. The temperature profile is taken from the US Standard Atmosphere.

In Irie et al. (2011), a simpler description of the troposphere is used and the state vector consists of VCD times a factor fclm VCD is defined as the vertical column density (VCD) for altitudes below 5 km. The ozone number density is fixed to 5.8×1011 molecules cm−3 at 5 km based on the US Standard Atmosphere and the vertical profile shape is assumed to be linear between 0 and 5 km. Then, the vertical profile of ozone below 5 km is determined depending on the VCD: a smaller VCD tends to yield a linearly increasing profile with altitude while a larger VCD produces a linearly decreasing profile. It is assumed that ozone concentrations are more variable in the Planetary Boundary Layer (PBL) than in the lower free troposphere, as the primary target of the Irie et al. (2006) study is to see variations in PBL concentrations. Above 5 km, the a priori profile has been set to the US Standard Atmosphere ozone profile. However, the profile above 5 km has been made multipliable by a factor, fclm, in the retrieval in order to ensure a smooth matching between the profile parts below and above 5 km. For each 30-min interval, the a priori VCD value and the corresponding error are set to 20% and 100% of the maximum ozone differential slant column density (DSCD) values. The a priori fclm (±error) is set to 1.0±1.0. Regarding the aerosols, a fixed AOD value (0.2) is assumed together with an exponentially decreasing with height profile shape.

In Gomez et al. (2014), a new approximation is proposed to estimate ozone mixing ratios from MAX-DOAS measurements at high-altitude sites. The proposed method uses O4 slant column densities (SCDs) at horizontal and near-zenith geometries to estimate a station-level differential path. This modified geometrical approach (MGA) takes advantage of a very long horizontal path to retrieve ozone mixing ratios in the range of a few pptv (parts per thousand by volume). Moreover, measurements and retrieval approaches should be thoroughly characterized in terms of uncertainty budget and information content (vertical sensitivity, horizontal representativeness, dependency on measurement and solar geometries, and atmospheric visibility).

Although there have been these exploratory studies discussed above, there is still a clear need for a significant research effort to be undertaken by the DOAS community in order to (1) develop reference methods/algorithms and recommendations for the retrieval of tropospheric ozone vertical profiles and columns from MAX-DOAS measurements, and (2) operationally apply these algorithms to all existing MAX-DOAS stations.

In particular, the following specific issues have been identified:

1. Lack of understanding of the information content of MAX-DOAS tropospheric ozone measurements. Although the studies discussed above have demonstrated the feasibility of tropospheric ozone measurements from UV-visible absorption measurements in both the Huggins and Chappuis bands (see Liu et al., 2006; Irie et al., 2011; Gomez et al., 2014), the information content of such measurements remains to be thoroughly explored in terms of vertical sensitivity, dependency on measurement geometry (in particular the number of viewing angles being sampled), dependency on atmospheric visibility (i.e. aerosol content), solar geometry, horizontal representativeness, etc. This current lack of knowledge of the information content of MAX-DOAS tropospheric ozone measurements restrains the usage of this technique for large scale ozone monitoring and satellite and model validation. A better characterization of this information content will contribute to the development of robust retrieval methods (see also Remedy #1).

2. Better characterization of the different MAX-DOAS tropospheric ozone retrieval methods needed. So far the retrieval methods applied are experimental and are either based on Optimal Estimation (OE) schemes (Liu et al., 2006; Irie et al., 2011) or on more simple approaches such as the modified geometrical approximation used in Gomez et al. (2014) to infer free-tropospheric ozone concentration from a high-altitude site. More work is necessary to better characterize the different approaches. Such characterization will, in turn, also contribute to a better understanding of the information content corresponding the MAX-DOAS tropospheric ozone measurements (see bullet 1) above and Remedy #1 below).

3. Lack of in-depth understanding of random and systematic uncertainties of MAX-DOAS tropospheric ozone measurements. A better characterization of these uncertainties will contribute to a more in-depth knowledge of the information content of the corresponding MAX-DOAS tropospheric ozone measurements. As for other trace gases, the main uncertainties are related to the estimation of the effective photon light path, which is dependent on the aerosol content and optical properties. Moreover, in the case of ozone, the interference with the strong ozone absorption taking place higher up in the atmosphere is potentially a significant source of systematic bias and a comprehensive error budget of tropospheric ozone retrieval from MAX-DOAS measurements is lacking. The lack of uncertainty characterization of tropospheric ozone measurements from MAX-DOAS instruments restrains the potential for network capabilities assessment and the usage of these data for satellite and model validation purpose.