TOMS Accuracy Over Cloud Regions
Newchurch, M. J. ; Liu, X. ; Kim, J. H. ; Bhartia, P. K., On the accuracy of Total Ozone Mapping Spectrometer retrievals over tropical cloudy regions, J. Geophys. Res., 106 (23) , 32,315-32327, 2001.[Full Text (PDF)]
Motivated by the desire to accurately derive tropospheric ozone from TOMS measurements using the clear/cloudy difference technique, we investigate the TOMS ozone retrieval over tropical highly reflecting clouds.
1. Cloud-height related errors
Because the TOMS instrument cannot detect ozone below optically thick clouds, the retrieval strategy is to include a climatological tropospheric ozone amount below observed clouds according to a simple monthly ISCCP cloud climatology. Therefore, incorrect cloud heights will introduce errors in the retrieved ozone. On the one hand, incorrect cloud-top height will affect the ozone added below clouds. On the other hand, the TOMS algorithm will mistreat the ozone absorption and Rayleigh scattering between the actual and assumed cloud pressure levels and affects the retrieved column ozone above clouds. In addition, in the radiance look-up table for the TOMS V7 algorithm, the radiance is computed only at 1.0 atm and 0.4 atm pressure levels. When the assumed cloud-top pressure is not at 0.4 atm, the TOMS algorithm would introduce error due to using interpolation to get the calculated radiances at that level. We call the above three errors associated with cloud-top height as radiation interpolation error (D O3, rad), ozone retrieval error above cloud (D O3, above), and ozone retrieval error below cloud (D O3, below), respectively.
The ozone retrieval error below cloud is straightforward. This error is simply the amount of climatological ozone between the assumed and actual cloud-top heights. The D O3, below depends on the cloud-height errors, the effective cloud fraction, and the actual vertical ozone profile. An error of 100 mb in cloud-top pressure at 40% reflectivity (effective cloud fraction of 0.4) has much less effect on the ozone retrieval than a similar error at 80% reflectivity (effective cloud fraction of 1 for reflectivities ?0.8).
Figure 1a shows the radiation interpolation error as a function of cloud height and viewing geometry. We can see this error is positive within 2 DU below 400 mb and negative above 400 mb, which could be as large as 8 DU. Figure 1b shows the ozone retrieval error above cloud as a function of assumed cloud top pressure for a actual cloud-top pressure of 400 mb. This error is usually positive within 2 DU above 400 mb and negative below 400 mb, which could be as large as 10 DU. For tropical high reflecting clouds, the assumed cloud is usually 400 mb, the actual might be at 100~200mb, both of these two errors are small. The total error is dominated by ozone retrievla error below cloud. The correction these three cloud-height induced error is called D P correction.
2. Total ozone difference between cloudy areas
Figure 2 shows a specific example of TOMS total ozone vs. reflectivity in the western Pacific Ocean. 5 DU more ozone is found over cloud areas (Figure 2a). Due to larger cloud height errors (Figure 2b), the D P correction leads to much higher ozone (~ 15 DU) over cloudy areas. The high spatial correlation between ozone and reflectivity after D P correction is evident in Figure 3. Figure 4 shows the time series average total ozone cloudy/clear difference in four selected regions (two in western Pacific Ocean and two in Equatorial Africa) in 1980 Nimbus-7 TOMS data. The D P correction increases the cloudy/clear difference by ~10 DU in the western Pacific Ocean and in both regions, leading to ~15 DU (~11 DU) more ozone over cloudy areas in the western Pacific Ocean (Equatorial Africa). Figure 5 indicates the uncorrected total column-ozone differences in the same regions in 1999 using Earth Probe (EP) TOMS v7 data. Compared to N7 TOMS data, the cloudy/clear difference is smaller by ~5 DU. This N7/EP bias is not clear, but most probably due to some non-linear calibration errors in N7 data.
3. Sources of excess ozone over cloudy areas
we investigated the following sources for those ozone excesses over clouds: incorrect tropospheric ozone climatology, dynamical and chemical influences, instrument calibration error, and the algorithm assumption of an opaque Lambertian scattering surface for cloudy scenes.
We see no evidence for either a dynamical or chemical mechanism producing a real, geophysical ozone excess over cloudy regions. Considering that EP instrument is the most recent, and probably best calibrated [Personal communication with S. Taylor and G. Jaross, 2000], we ascribe the bias of 6~7 DU between N7 and EP TOMS measurements to a N7 calibration error. We add the ozone below clouds based on TOMS standard tropospheric climatological ozone, however, it will overestimate or underestimate the total ozone depending the deviation of the actual ozone profile from the TOMS standard. The TOMS standard tropospheric ozone (L275) contains about 34 DU below ~100 mb, which is, in average, larger than that in the western Pacific Ocean and smaller than that in the Equatorial Africa. Considering the effects of incorrect tropospheric climatology based on ozonesonde measurements, the cloudy/clear difference reduces by ~5 DU in the western Pacific Ocean but exacerbates by ~ 5 DU in the Equatorial Africa.
After considering the above sources, there is still about ~4 DU more ozone in the western Pacific Ocean and ~9 DU in the equatorial Africa that remains unexplained. We speculate that a significant amount of this excess could be due to inaccurate treatment of clouds in the TOMS algorithm as opaque Lambertian surfaces rather than a multiple scattering medium. Because the actual clouds are not opaque surfaces, photons can penetrate into the clouds. Under cloudy conditions, the multiple scattering in clouds enhances the ozone absorption path length and therefore overestimates the ozone in TOMS Version-7 algorithm. The magnitude of overestimation depends on the optical thickness of clouds and the actual ozone distribution in clouds. In addition, high-altitude clouds near the tropopause consist of ice crystals. Neglecting the scattering properties by treating clouds as Lambertian surfaces might produce additional effects on ozone retrieval. Table 1 summarizes different types of error to the cloud/clear difference described above.
Table 1. Contribution of different types of errors to annual-mean total ozone clear/cloudy difference, D O3 (DU), over tropical high convective clouds and neighboring clear areas in 1980 NIMBUS-7 TOMS v7 measurements. Values in parenthesis indicate 1 standard deviation. The D P correction is defined in section 2.1 as the sum of the radiation interpolation error, retrieval error above clouds and below clouds.
Region |
D O3 before D P correction |
D O3 after D P correction |
Calibration error * |
Incorrect climatology |
Unknown sources |
Western Pacific Ocean |
2(3) | 15(3) | ~ 6 | ~ 5** | ~ 4 |
Equatorial Africa |
1(5) | 11(3) | ~ 7 | ~-5 *** | ~ 9 |
* Assume the calibration error is in only the N7 TOMS instrument.
** The annual average tropospheric ozone is assumed to be 22 DU in the western Pacific Ocean.
*** The annual average tropospheric ozone is assumed to be 40 DU in the Equatorial Africa.
4. Summary and Conclusions
We analyzed different types of errors associated with tropical high reflecting clouds in TOMS data: cloud-height related errors, calibration error, incorrect tropospheric climatology and treatment of optically thick clouds as Lambertian Surface. Of these errors, ozone retrieval errors below cloud and incorrect tropospheric climatology will only affect the added ozone below clouds, while other errors affect the retrieved ozone above clouds. Overall, the ozone retrieval error in total ozone (Table 1, column 2) is relatively small because the mainly cloud-high related negative error cancels (difference between column 3 and column 2) or partly cancels the other positive errors.
However, the excess ozone above clouds is critical to the tropospheric ozone derivation using clear/cloudy difference techniques. The overestimated ozone above clouds due to calibration error or unknown sources will tends to underestimate the derived tropospheric ozone by ~8 DU in the western Pacific Ocean or ~18 DU in the Equatorial Africa if using the clear/cloudy difference techniques. In the Convective Cloud Differential method [Ziemke et al., 1998], they assume zonal invariant stratospheric ozone and derive stratospheric ozone from the monthly minimum ozone above high clouds in the western Pacific Ocean. This special sampling cancels or partly cancels the excess ozone above high clouds because the minimum ozone is significantly smaller than the average.
We speculate that a significant amount of this excess could be due to inaccurate treatment of clouds in the TOMS algorithm as opaque Lambertian surfaces rather than a multiple scattering medium. Detailed radiative transfer calculations using a non-Lambertian model compared to TOMS Radiative transfer Code and TOMS V7 algorithm are still required to analyze the effects of both wavelength-dependent ozone absorption enhancement by in-cloud multiple scattering and scattering properties of ice crystals on ozone retrieval.
