description unit MARSS channel
16 17 18 19 20
center freq. GHz 88.99±1.1 157.08±2.6 183.25±1.0 183.25±3.0 183.25±7.0
IF bandwidth GHz 0.6 2.2 0.5 1.0 2.0
NE∆T K 0.5 0.7 0.6 0.4 0.3
cal. accuracy K 0.9 1.1 1.0 0.9 0.8
beam width deg. 11.8 11.0 6.2 6.2 6.2
Table 5.6: Description of the MARSS radiometer channels, valid for the POLEX campaign.
The beam width is given as FWHM of the antenna mainlobe. The integration time was 100 ms for all channels (Hewison, 2002;McGrath and Hewison, 2001).
Figure 5.13: Humidity and temperature profiles of the two selected MARSS flights within the POLEX campaign. The left plot shows the water vapor volume mixing ratio, the middle plot the relative humidity with respect to liquid water and the right plot shows the tem-perature profile. The measured atmospheric profiles are topped by a zonal mean profile of Kneizys et al.(1996) up to 90 km. The additionally shown profiles are from nearby radiosonde launches, these are for the flight A827.1: Ny-˚Alesund (78.55◦N/11.56◦E, 12:00 UTC), Dan-markshavn (76.77◦N/18.66◦W, 11:00 and 23:00 UTC), Scoresbysund (70.48◦N/21.95◦W, 11:00 and 23:00 UTC) and for the flight 829.7: Bodø (67.25◦N/14.37◦E, 12:00 and 23:00 UTC), Ny-˚Alesund (78.55◦N/11.56◦E, 12:00 UTC), Ørland (63.70◦N/9.60◦E, 11:00 UTC).
As can be seen from Figure 5.16, different atmospheric components contribute differently in channel 16 and 17. In channel 16 oxygen is the dominant absorber while in channel 17 the water vapor foreign continuum component is the main absorber. A change in the water vapor mixing ratio profile would hence only alter the results for channel 17.
However, some additional radiosonde profiles for temperature and humidity are shown in Figure 5.13. These radiosonde profiles from nearby stations show relatively good agreement with the profile used in the calculation. Hence the error due to incorrect atmospheric profiles in the radiative transfer calculations can not explain the discrepancies.
One way to eliminate a possible bias in the measured brightness temperature is to calculate from successive TB values the absorption (in units of Np/hPa) for an atmospheric layer (English et al., 1994):
α= 1
δ Ptot ·ln TB(P2)−T TB(P1)−T
!
(5.2) where T is the mean temperature of the atmospheric layer and δ Ptot = P1 −P2=200 hPa represents the chosen thickness of the atmospheric layers. The calculated and measured absorption is shown in Figures 5.17 and 5.18 for the two flights. For flight A827.1 the measured and calculated absorption shows relatively good agreement with the exception of channel 16. For flight A829.7 the agreement is sufficiently good for relatively low absorption in channel 17 to 20. With increasing absorption the difference increases, indicating some discrepancies in the lowermost atmospheric levels of this profile. Neglecting these boundary layer difficulties, the agreement in absorption between model calculations and MARSS are relatively good (for additional plots see Section F.3). This good agreement suggests a bias in the MARSS measurements during the POLEX campaign.
In both flights the agreement between calculation and MARRS measurement is relatively bad for channel 16 irrespective of the used water vapor absorption model. It is therefore assumed that the discrepancies in channel 16 are caused by difficulties on the instrumental side.
Figure 5.14: Comparison of the measured MARSS brightness temperatures (flight A827.1, filled circles) with those calculated with Arts (lines), using atmospheric data as input for the temperature and relative humidity profiles. The absorption models used in Arts are AAM02 (H2O), updated version of Rosenkranz (1993) (O2), and Liebe et al. (1993) (N2).
The nitrogen absorption is scaled by 1.29 according to Pardo et al. (2001b). The center frequencies of the channels are 89, 157, 183.3±1, 183.3±3, 183.3±7 GHz.
Figure 5.15: Comparison of the measured MARSS brightness temperatures (flight A829.7, filled circles) with those calculated with Arts (lines), using atmospheric data as input for the temperature and relative humidity profiles. The absorption models used in Arts are AAM02 (H2O), updated version of Rosenkranz (1993) (O2), and Liebe et al. (1993) (N2).
The nitrogen absorption is scaled by 1.29 according to Pardo et al. (2001b). The center frequencies of the channels are 89, 157, 183.3±1, 183.3±3, 183.3±7 GHz.
Figure 5.16: Brightness temperatures for the different components of the atmospheric ab-sorption models. TheTB calculation is for each absorber individually calculated, ignoring all other absorbers. For the three water vapor absorption components, (lines, self continuum, and foreign continuum) theAAM02model is used. For the oxygen and nitrogen absorption the updated version of Rosenkranz (1993) (O2) and Liebe et al. (1993) (N2) is used. The nitrogen absorption is scaled by 1.29 according to Pardo et al. (2001b). The atmospheric conditions are taken from the flight A829.7
Figure 5.17: Comparison of the measured and calculated absorption (flight A827.1), using atmospheric data as input for the temperature and relative humidity profiles. The absorption models used in Arts are AAM02 (H2O), updated version of Rosenkranz (1993) (O2), and Liebe et al.(1993) (N2). The nitrogen absorption is scaled by 1.29 according toPardo et al.
(2001b). The center frequencies of the channels 16 to 20 are 89, 157, 183.3±1, 183.3±3, 183.3±7 GHz.
Figure 5.18: Comparison of the measured and calculated absorption (flight A829.7), using atmospheric data as input for the temperature and relative humidity profiles. The absorption models used in Arts are AAM02 (H2O), updated version of Rosenkranz (1993) (O2), and Liebe et al.(1993) (N2). The nitrogen absorption is scaled by 1.29 according toPardo et al.
(2001b). The center frequencies of the channels 16 to 20 are 89, 157, 183.3±1, 183.3±3, 183.3±7 GHz.
Chapter 6
Conclusion
Different absorption models for oxygen, nitrogen, and water vapor for atmospheric conditions were outlined and compared to one another.
Furthermore, a new water vapor absorption model, H2O-AAM02, was built up for the 1-1000 GHz range. The line center frequencies and intensities of the 34 selected water vapor lines considered in H2O-AAM02 were taken over from the standard HITRAN line catalog. The line broadening parameters were merged from different published laboratory measurements and theoretical investigations. Additionally, the coefficients of the H2O-AAM02 empirical water vapor continuum parameterization were fitted to absorption coefficients determined from published laboratory measurements of pure water vapor and of water vapor mixed with molecular nitrogen. These measurements were performed in the window regions between strong resonant lines (150 GHz< ν <350 GHz).
The difference between the laboratory measurements and the model calculation was below 4 % in the window regions 150 GHz-160 GHz (H2O), and 200 GHz-350 GHz (H2O and H2 O-N2). For the H2O-N2 mixture the difference was between 8 % and 13 % in the 150-160 GHz window region.
Moreover the difference between measured and calculated absorption coefficients around the 183 GHz water vapor line (measurements not considered for the continuum parameter fit) was about 12-17 %.
Differences between a variety of absorption models were investigated using ECMWF at-mospheric profiles generated from the 40 year reanalysis (ERA-40) as well as radiosonde profiles in connection with radiometer measurements. The ECMWF profiles were especially useful to reveal qualitative differences between the absorption models while the comparisons between radiometer measurements and model calculations were used to validate the models.
The main findings of these variability and validity investigations were the following
• for the oxygen absorption models:
◦ based on calculations with the ECMWF atmospheric profiles a mean difference in brightness temperature between the different model calculations was about 0.4 K for AMSU-A frequencies. This is of the same order as the noise equivalent tem-perature of this instrument;
◦ the frequency range between the 60 GHz band and the 118 GHz line was the fre-quency range where the differences between the different oxygen absorption models were largest. Calculations with the ECMWF profiles revealed brightness temper-ature differences of several Kelvin between radiative transfer calculations using O2-MPM93 or O2-PWR98. Ground based radiometer measurements at these frequencies should therefore be valuable to validate these absorption models;
◦ the recent update of the O2-PWR model (O2-PWR98) included a modification of the pressure broadening of the 118 GHz line. This modification can lead to brightness temperature differences up to 4 K in the wing of the 118 GHz line in the lower stratosphere. It would therefore be of interest to compare these model predictions with Odin measurements, an operational MMW and SMMW limb sounder with a channel around this oxygen line;
◦ the analysis of the ARM measurements (ground based radiometers MWR and MIR) did not give a clear answer about the most favorable oxygen absorption model. However, best agreement was reached with O2-MPM93 or O2-PWR98 while older model versions were less favorable;
◦ An improved far wing line shape for the oxygen band at 60 GHz would be able to solve the unusual behavior of negative line absorption in the far wing region above 200 GHz. A possible approach could be to incorporate a correction term for the duration of the molecular collision as commonly done for the CO2 line mixing band in the infrared region. As was shown for the 650 GHz frequency range the uncertainty of the oxygen absorption was as large as the absorption of ClO. It was pointed out that the uncertainty in the nearly constant oxygen absorption at these frequencies can be included as an additional offset in trace gas retrieval schemes.
Hence this uncertainty is not crucial for trace gas retrievals even of weak species with low concentrations. Nevertheless, this uncertainty could be of importance for the determination of line broadening parameters from laboratory measurements with oxygen as buffer gas.
• for the nitrogen absorption models:
◦ a comparison of the fast parameterizations N2-PWR93 and N2-MPM93 with the most detailed absorption model of N2-BF86 revealed that N2-PWR93 is a good approximation up to frequencies of 200 GHz while N2-MPM93is a suitable approximation for the entire STHz range;
◦ the analysis of the ARM measurements (sub-arctic winter conditions) supported the findings of Pardo et al. (2001b). An increase of 29 % of the nitrogen absorp-tion yielded generally better agreement with the MIR radiometer measurements.
Nevertheless, one has to emphasize that this increase is not due to uncertainties in the nitrogen absorption model, but covers the collision induced absorption of other trace gases, which are not considered in the radiative transfer calculations. So far no collision induced absorption models for these trace gases are available for the STHz frequency range. Hence the scaling of the atmospheric nitrogen absorption has to be regarded as a first approach towards a more detailed model;
• for the water vapor absorption models:
◦ absorption calculations with different water vapor absorption models using the ECMWF profiles showed that the difference between the models was most pro-nounced in the window regions between the strong water vapor lines. Moreover, the difference between the models was most susceptible to the atmospheric state in the window regions from 22 GHz to 183 GHz and from 183 GHz to 325 GHz;
◦ comparing the different water vapor absorption models with the laboratory mea-surements ofBecker and Autler (1946) revealed that the absorption models which are not independent of this data set (i.e. H2O-MPM93and H2O-PWR98) differ by about 5 % in the line center of the 22 GHz water vapor line and up to 15 %
in the wing region. Similar differences were obtained for H2O-AAM02 which is independent of this data set. By decreasing the foreign continuum temperature coefficient from its original value of xf = 1.33±0.73 to about 0.6 increased the difference in the wing region significantly;
◦ comparing the different water vapor absorption models with radiometer data from MWR and MIR (22-350 GHz) for sub-arctic winter conditions demonstrated the validity of H2O-AAM02for such atmospheric conditions. H2O-AAM02matched the data at least as well as other models. Reviewing all channels simultaneously, the data was best matched by H2O-AAM02. In connection with H2O-AAM02it is of importance to note that the foreign continuum temperature coefficient deduced from the laboratory data (xf = 1.33) was too high compared with radiometer data. The MWR and MIR radiometer data suggested a value in the range of 0.0≤xf ≤0.6;
◦ the comparison between AMSU-B data over Lindenberg (Germany) with differ-ent water vapor absorption models did not clearly favor a single model. For the three channels around the 183 GHz line most model calculations were within the accuracy of the AMSU-B instrument. Nevertheless, based on the mean brightness temperature difference, ∆TB =TB(calc.)−TB(AMSU-B), H2O-MPM93agreed best with the 183±1 GHz channel while H2O-PWR98 and H2O-AAM02 showed the best agreement in the wing channels at 183±3 GHz and 183±7 GHz;
◦ the MARSS data set (89-190 GHz) from the POLEX campaign suffered from un-explained large differences in brightness temperatures compared to model pre-dictions. Although the MARSS radiometer performance was carefully checked before and during the campaign it was assumed that the differences were on the instrumental side. A check of the atmospheric temperature and humidity profiles revealed that this large discrepancy between the measurements and model calcula-tions could not be explained by using incorrect assumpcalcula-tions about the atmospheric state in the radiative transfer calculations. Further investigations are needed to uncover the causes of this discrepancy.
Comparing the absorption coefficients deduced from the measured and calculated brightness temperatures showed much better agreement (with the exception of the 89 GHz channel). This result suggested a bias in the MARSS TB measurements.
Rearranging the above conclusions, one can draw – nota bene within the restrictions of the present investigations – some preferable absorption model combinations for dedicated frequency bands in theMW and MMW range:
10-35 GHz: based on the ARM MWR radiometer data set the H2O-MPM93 and H2 O-AAM02 (xf = 0.6) are the most preferable water vapor absorption models.
50-118 GHz: in this frequency range the oxygen absorption is more crucial than water vapor absorption. No data have been analyzed in this frequency range. Since O2-PWR98 is the most recently updated model it is the most favorable model for oxygen. Furthermore it is on a sound physical basis without any artificial cutoff. The inaccuracies in the far wing of the 60 GHz band are a general line shape problem common to allMMWoxygen absorption models since all these models are intended to be used close to the 60 GHz band. The cutoff introduced in the O2-MPMis not an appropriate method to minimize the far wing inaccuracies.
150-200 GHz: taking the results for the MIR and AMSU-B 187±7 GHz channels into ac-count H2O-AAM02 (with xf = 0.6) together with H2O-MPM89 matches the data best. For the AMSU-B 187±1 GHz and 187±3 GHz channels, H2O-MPM93and H2 O-PWR98 are preferable. Hence in this frequency range the four absorption models H2O-MPM89, H2O-MPM93 and H2O-PWR98 and H2O-AAM02 are of the same validity.
200-350 GHz: H2O-AAM02matches best the MIR window channels (with 0.0≤xf≤0.6).
H2O-MPM93shows too strong an absorption in this frequency range and H2O-PWR98 underestimates these measurements significantly.
Appendix A
Physical Constants and Units
A.1 Constants
quantity abbreviation value unit
speed of light (vacuum) c 2.99792·108 m/s
29.9792 cm GHz
Planck constant h 6.626069·10−34 J s
Boltzmann constant kB 1.38066·10−23 J/K
0.69504 cm−1/K
kB/h 20.8366 GHz/K
h/kB 0.04799 K/GHz
Avogadro constant NA 6.02214·1023 mol−1
Loschmidt constant NL 2.68676·1025 m−3
gas constant R 8.31451 J/K/mol
specific gas constant of H2O RH2O 461.51 J/K/kg specific gas constant of dry air Rair 287.104 J/K/kg
H2O molecular weight MH2O 18.016 g/mol
O2 molecular weight MO2 32.01 g/mol
N2 molecular weight MN2 28.02 g/mol
dry air molecular weight Mda 28.97 g/mol
Bohr radius ao 0.529177·10−10 m
Table A.1: Frequently used physical constants taken fromMohr and Taylor (2000);Trenberth (1995).