The developed code was used for the simulation of MARSCHALS measurements and for subsequent retrievals from these simulations. An example of the expected measured spectra is reported in Figure 1 , where MARSCHALS measurements at mid latitude on the three bands have been simulated for three tangent altitudes (20, 16 and 10 km). As it can be seen in the figure, molecules whose emission features are present in MARSCHALS spectra are H2O, O3, HNO3, N2O, CO and O2.

For simulated retrievals, several case studies have been considered and provided useful theoretical guidelines for the preferred measurement and retrieval configurations as well as for the feasibility of the measurements.

As reported in table 4 it was found that from the simultaneous retrieval of a single scan made of 21 limb views at 1 km step in tangent altitude, in the three bands with a NET (Noise Equivalent Ratio) of 1 K from an altitude of 20 km at middle latitude it is possible to measure down to about 6 km and that are retrieved: 8.7 degrees of freedom for temperature, 16 degrees of freedom for H2O, 11.2 degrees of freedom for O3, 7.3 degrees of freedom for HNO3, 4.7 degrees of freedom for N2O, 4.5 degrees of freedom for CO. These degrees of freedom characterize the effective vertical resolution.

As far as the altitude range and the accuracy are concerned, water is retrieved from flight altitude down to 6 km with a total error that varies from 10 to 40 %, ozone down to 10 km with a total error that varies from 10 to 60 %, HNO3 down to 11 km with a total error that varies from 10 to 55 %, N2O down to 10 km with a total error that varies from 25 to 7 %, and CO with a total error that varies from 40% at 11 km to about 70 % at flight altitude.

Typically the altitude independent and band dependent offset can be determined with an error of about 0.2 K, the altitude independent and band dependent gain can be determined with an error of about 0.2% and the altitude independent and band independent pointing bias can be determined with an error of about 8 mdeg.

The absence of one of more bands does not prevent the retrieval of important information from MARSCHALS spectra; as shown in tables 5, 6 and 7 the retrievals performed on the measurements of one band at a time still produce acceptable results on most target quantities.

Scalar instrumental parameters can be determined with very good accuracy in a multi-target retrieval that includes the atmospheric species. It is found that horizontal gradients of ozone and water vapour can be retrieved together with the profile of the constituent when the gradient is very large. However, for the detection of more realistic gradients, the retrieval must be constrained with some external source of information.

As shown in Figure 2, retrieval tests performed using for the retrieval a full variance covariance matrix (VCM) that includes both the measurement errors and the Forward Model (FM) errors have highlighted a major reduction of the total retrieval error, with respect to an a-posteriori calculation of the total error budget, currently adopted in most retrieval codes.

This result is not surprising because when the FM errors are considered in the retrieval process, a reduced weight is applied to spectral channels affected by large systematic errors so that they contribute less to the total error budget.

Measurements in the case of a cloudy atmosphere have been modelled considering different types of clouds located between 10 and 12 km. We found that in the range of physically possible clouds (characterized by the combination of particle size and density) four regions exist. These regions are highlighted in Figure 3, where they are compared with the properties of observed clouds taken from Evans et al. [J. Appl. Met. 37, 184–205 (1998)]

The A region contains the clouds that are practically transparent at millimetre waves. The B region contains the clouds that can be detected from the measured spectra, but are adequately modelled in the retrieval by the fit of an atmospheric continuum absorption. The C region contains the clouds that are adequately modelled by the atmospheric continuum absorption, but can cause some geometrical errors if the retrieval cannot reproduce the sharp vertical discontinuity of the cloud. Finally, the D region contains the clouds that, still affected by the geometrical error and still transparent to millimetre waves, require the modelling of scattering effects. An important result of this study is the identification of the geometrical effect as an error that can be more important than the scattering correction.

Results of Field Measurements

Only one flight of the MARSCHALS instrument aboard of the stratospheric M55-Geophysica aircraft provided useful measurements. As a matter of fact, during the analysed flight scans were made for about 200 limb views in a single band. The three bands were acquired in subsequent scans: one scan for band B first, then band D and then band C. Band D was not yet operational at the time of this campaign. In band B good spectra were measured at the beginning of the flight, but the receivers proved to be thermally unstable and the quality of the recorded spectra rapidly deteriorated after takeoff in the harsh meteorological condition at high flight altitudes. Thus the results presented in this study are derived from analysing atmospheric spectra measured in band C on the remote sensing flight 9 of 5th Dec 2005 only. The reduced performances of the measurements have prevented the verification of the measurement performances and of the gradient retrieval capability that have been considered in the studied case. The use of the variance-covariance matrix of the forward model during the iterative procedure also turned out to have a small effect on the results of the retrieval because spectral error larger than planned reduced the effects of the systematic errors on the measurements. However other results have still been possible.

In particular, it was possible to verify with direct evidence the capability of millimetre wave measurements of making minor constituent measurements in presence of clouds that obscure middle infrared instruments. The flight took place inside a cloud that made impossible most of the limb measurement of MIPAS-STR, while MARSCHALS was practically unaffected by the clouds. The high concentration of water vapour in the tropical region where the flight was made, limited the altitude range of the retrieval to about 12 km. Furthermore, while the aircraft altitude ranged from 17 to 19 km, the tropopause was located at about 17 km. So species like ozone and HNO3, that have a very low concentration in the troposphere, had only in a small altitude range abundances above the detection limits of the instrument. Within the limits of the reduced sensitivity due to the single band measurements and to the larger spectral error, the retrieval performances were consistent with expectation and H2O, O3, and HNO3 have been measured in the UTLS (Upper Troposphere – Lower Stratosphere) in presence of clouds that are opaque in the middle infrared.

Figure 4 to Figure 8 show maps of the vertical distribution of the target quantities and of their biased error wrt the acquisition time. The figures highlight the fact that, despite the variable flight altitude and the cloud coverage, good and stable results have been obtained. These results have been validated with the help of measurements acquired by other instruments (both from remote sensing and in-situ) during the same flight.

 

 

 

 

 

In Table 8 the average retrieval performances of the analysis of band C scans during SCOUT-O3 flight 9 are reported.

A comparison of Table 6 with Table8 shows that despite the different location of the measurements and the different measurement scenario, the results obtained in the simulated retrievals are very close to the real performances of the instrument, validating therefore the whole theoretical exercise.

The retrieval of the scalar instrumental parameters confirmed the expected instrument specification. As shown in Table 9, the retrieved values for pointing bias, gain and offset are very close to the nominal ones, proving the quality of the calibration attained in Level 1, and the retrieval error is consistent with the value estimated in the retrieval study, also considering the different NET.

Analysis of the residuals obtained in the retrieval procedure highlighted that in the observed band C the spectroscopic database specifically developed for the MASTER instrument turned out to be adequate for the modelling of the observations.

The identification of a cloud correction procedure for clouds that introduce a distortion in the measured spectra was also an important objective. Unfortunately, the single band measurement and the missing calibration of the OCM instrument, devised to provide auxiliary information about the encountered clouds, have prevented most of the planned diagnostic. Nevertheless, a useful experience was made about cloud corrections. The retrieved continuum absorption indicates that during most of the flight the clouds were of either type A or B and, therefore, no further cloud correction procedure was needed for these measurements. During one scan (scan 12) a thick cloud of either type C or D was observed. The distortion introduced by the cloud is compensated by a value of the retrieved continuum that is significantly larger than the one obtained for the other scans. The saturation pressure of ice and water, the vertical profile of the retrieved continuum absorption (measured in terms of linear extinction coefficient) and thermodynamic considerations have been used to determine the phase and location of the cloud (that in our case was in the region between 15 and 12 km). The value of the linear extinction coefficient constrains the properties of the cloud (i.e. ice mass content and particle radius) to a limited set of possible combinations. With a few trials it has been possible to find a cloud that, when modelled with the MSSF code, leads to a reduction of the c2-test to the value of 1.39, even if the residual retrieved continuum absorption still shows values different from zero (indicating that a better determination of cloud thickness and location could be made). Despite the significant reduction of the c2-test the changes in the retrieved target parameters are not significant, given the large retrieval error caused by the reduced instrument performances. Therefore, it was not possible to observe the data improvement obtained with the cloud correction that was predicted in the theoretical retrieval study, but the possibility of performing this operation has been demonstrated and for an operational instrument an automated procedure can in future be considered.