Comparison between TOMS, TOVS and DOBSON Observations: Satellite and Surface Views of Total Column Ozone

Presented at the American Meteorological Society's Fifth Conference on Satellite Meteorology and Oceanography, in London, England, September 1990.

Abstract

TOVS infrared and microwave observations are now being used to estimate total column ozone over the globe by employing a new physical retrieval algorithm. The TOVS archives from 1979 to the present have been reprocessed, and the results during the great Antarctic "ozone hole" of 1987 are compared with the corresponding TOMS and DOBSON ultraviolet observations of total ozone. The TOVS total ozone patterns appear to be reliable on the synoptic scale and provide a unique opportunity to track ozone variations during polar night. However, there are times and places where each of the three ozone datasets contains errors that are significantly larger than a few percent.

Introduction

NASA's NIMBUS-7 satellite has been tracking total column ozone over the globe every day since late in 1978, using the ultraviolet Total Ozone Mapping Spectrometer (TOMS). Over the same period, NOAA's TIROS-N series of satellites have been collecting infrared ozone observations, using the TIROS-N Operational Vertical Sounder (TOVS), which has an ozone-sensitive band centered at 9.7 µm. To make TOVS ozone retrievals, a new physical algorithm has been devised and implemented in 1990, including a reprocessing of the TOVS ozone data from 1979 to the present.

This paper examines the reliability of the TOVS ozone retrievals. The quality of the new TOVS total ozone soundings are compared visually and statistically to the corresponding TOMS values and to values reported by the ground-based DOBSON network. The three-way TOMS, TOVS, and DOBSON comparison concentrates on observations during the period of the great Antarctic "ozone hole" late in 1987.

Because of the complexity of the circulation patterns in the total ozone on a wide range of space-time scales, the results of this inter-comparison are best presented using an animated videotape to show three side-by-side versions of the development and decay of many different ozone features over the Earth (available from the authors).

Remote Sensing Of Total Ozone

Total column ozone is measured in Dobson Units (DU), equivalent to the height of the ozone if it were all brought to standard temperature and pressure, and scaled so that a 3 mm column of pure ozone is 300 DU. Remote sensing measurements of total ozone can be accurate to a few percent by measuring the differential absorption and scattering in the strong ozone spectral bands. Ozone measurements are available annualLy from a network of approximately 60 active ground-based DOBSON instruments that observe the differential ultraviolet spectrum of sunlight. The 1987 DOBSON observations used here are designated "DOBS(88)" (CAES, 1988).

The TOMS satellite-based instrument uses differential absorption of reflected solar ultraviolet light to estimate total ozone (Fleig et al., 1982). The TOMS data used in this study are version 5 of the globally gridded database with approximately 100 km resolution, designated "TOMS(5)", and have been converted to a convenient video format for browsing and comparison to other datasets (Chesters and Krueger, 1989).

The TOVS physical retrieval algorithm uses the difference between NOAA's pre-processed clear-column infrared 11 µm and 9.7 µm brightness temperatures to determine total ozone. The TOVS physical algorithm adopted by NOAA (Neuendorffer, manuscript in progress) is similar to prior algorithms used to retrieve total ozone from thermal infrared data (Prabhakara et al., 1976; Muller and Cayle, 1983; Shukorov and Shukorova, 1986; Lienesch, 1988). The non-ozone channels are used to determine a "foreground" temperature for the ozone and a "background" temperature for the surface. Then the brightness temperature from the semi-transparent 9.7 µm ozone band channel is used in a linear algorithm to determine the ozone opacity. Finally, a small set of ground-based DOBSON stations in 1984 are used to determine a monotonic relationship between TOVS-determined ozone opacity and the total column ozone. The relationship is otherwise uncoupled from other ozone measurements or time-dependent corrections. Daily synoptic maps are generated from the thousands of individual TOVS physical retrievals from the day- and night-time overpasses from two different satellites, using standard operational polar stereographic mapping routines at approximately 330 km resolution. Regions of missing TOVS ozone data values due to lost orbits are not marked, but are replaced by average values for the latitude band.

Results

Figure 1 presents an approximate black-and-white rendering of TOMS, TOVS and DOBSON ozone imagery and statistics for 4 October 1987. Some features of interest are: (1) low ozone values over the South Pole, as the antarctic "ozone hole" wanes, surrounded by high values of ozone concentrated above the great storms in the "roaring forties"; (2) a lack of ozone observations by TOMS and DOBSON in the arctic polar night, which TOVS can supply; (3) an arctic "mini-hole" over the North Atlantic, observed at greater intensity by TOMS than TOVS, but with no corresponding report from the DOBSON station in Iceland; (4) a large ozone maximum over Siberia, with some difference in location that could be attributed to the difference in TOMS and TOVS local observing times; and (5) general agreement in TOMS and TOVS ozone features over mid-latitudes and the equatorial zones (except for TOVS erroneous ozone "micro-holes" due to cirrus cloud tops above some thunderstorms). TOVS has lower resolution and less contrast than TOMS due to the satellite data pre-processing designed to create nadir-view, cloud-corrected radiances over a synoptic field-of-view. The bottom right-hand panel in Fig. 1 presents the individual DOBSON site observations of total ozone. The DOBSON network is heavily concentrated in North America, Europe, India, Japan, and Australia. The DOBSON-observed ozone values have been also embedded as small squares within the satellite-observed TOMS and TOVS images of the Earth along the right-hand side of Fig. 1. The embedded DOBSON values are unnoticeable when they agree with the satellite estimates, such as for the Antarctic site south of the tip of South America. However, the embedded DOBSON values become noticeable when they do not agree with the satellite estimates, such as for the Antarctic site south of the tip of South Africa. On this day, DOBSON "ground truth" agrees with TOVS but not with TOMS, in the midst of a large area along the coast of Antarctica where TOVS indicates much lower ozone estimates than TOMS. Reststrahlen makes the deserts appear abnormally cold near this wavelength, and has been mapped in some detail over the globe (Prabhakara and Dalu, 1976). Unfortunately, Reststrahlen also causes ozone over-estimates in the TOVS retrieval algorithm, because the lowered emissivity decreases the brightness more at 9.7 µm that at 11 µm, mimicking the absorption effect of the ozone band at 9.7 µm. The effect of Reststrahlen could be minimized if the TOVS infrared sounder carried an 9 µm channel to provide the background emissivity between 8 and 11 µm. Finally, the bottom right-hand scatterplot in Fig 1 presents TOMS and TOVS versus DOBSON statistics at the ozone sites. The mean and root-mean-square differences between TOMS-DOBSON, TOVS-DOBSON, and TOVS-TOMS observations are summarized for the day as 3 error bars in the lower right-hand corner of this scatterplot, with typical values of 20 to 30 DU over the globe. The root-mean-square differences between TOMS-DOBSON, TOVS-DOBSON, and TOVS-TOMS observations over the entire globe average 20 to 30 DU during the last 4 months of 1987, with little seasonal variation (not shown). The corresponding mean differences are shown in Fig. 2, which indicates that TOVS ozone retrievals follow TOMS with a 5 to 10 DU bias during the largest ozone excursions of the decade. There is one bad day on December 29, when none of the four overpasses by TOVS over North America were successful, and the analysis scheme for the TOVS data incorrectly interpolated large amounts of ozone. This error in the TOVS grid likewise affects the TOVS-DOBSON statistics on the same day. Otherwise, the TOVS ozone estimates at the DOBSON stations average 5 to 10 DU less than the "ground truth", remarkably good agreement. During the same period, TOMS has a larger bias, 15 to 20 DU below the DOBSON "ground truth". The larger TOMS bias is traceable to a drift in the TOMS calibration (Krueger, private communication), to be corrected when version 6 of the TOMS database is released.

The TOMS and TOVS algorithms have both been designed and tuned to perform well at mid-latitudes. Indeed, TOMS and TOVS agree well at mid-latitudes, but each DOBSON ground station has its own individual pattern of disagreement with the two satellites. For example, Fig. 3 presents the ozone measurements by the NASA-operated instrument at Wallops Island, VA, USA. In this case, the DOBSON observations are all larger than the two satellites by 10 to 40 DU, a large disagreement between instruments thought to have ±5 DU accuracy and to be useful for detecting long-term trends to ±1% per year. Other DOBSON sites have other error patterns, such as biases lasting a few weeks to months, occasional excursions, no sensitivity to satellite-observed excursions, etc.. The discrepancies between the TOMS, TOVS, and DOBSON time-series at most mid-latitude sites raises questions about using the DOBSON network as "the standard" for measuring total ozone. Rather, one concludes that several different satellite and ground-based systems must be employed to discover and eliminate erroneous values from each, in order to arrive at a reliable consensus estimate for ozone amounts and trends. Each method for measuring ozone should be independently calibrated with respect to an accurate standard, and not with respect to each other.

Figures 4 and 5 present time-series plots of TOMS, TOVS, and DOBSON observations of total ozone before and after the onset of polar night near the North and South poles, respectively. The TOMS, TOVS, and DOBSON measurements inside the Arctic Circle presented in Fig. 4 indicate poor agreement among all three sensors before polar night, with TOMS appearing to significantly under estimate ozone on the edge of polar night, a strong impression also provided by the animated series of daily ozone images. TOVS provides ozone estimates in polar winter, but their quality is untested. The TOMS, TOVS, and DOBSON measurements at the South Pole presented in Fig. 5 indicate good agreement among all three sensors during the "ozone hole" just after the end after polar night. TOVS provides ozone estimates during the initial formation of the "ozone hole" early in polar winter night, but their quality is untested. However, during the breakdown of the polar vortex and dissolution of the "ozone hole" at the beginning of December, TOVS estimates of total ozone are in error, perhaps due to the poor sensitivity of the infrared channels to the changes that occur in the upper stratosphere at this time.

At mid-latitudes, ozone concentrations are definitely correlated with low pressure systems (Schoeberl and Krueger, 1983), and move with the prevailing westerlies rather than with the seasonal reverses in the stratospheric winds. On the synoptic scale, ozone concentrations have been noted near individual storms (Sechrist et al., 1986), possibly concentrated near the tropopause by folding and/or subsidence. Chesters et al. (1990) have shown that local ozone maxima near baroclinic waves actually occur eastward of the tropopause folds associated with mid-latitude troughs and jet streaks. In the tropics, weak ozone concentrations have been linked to upper-level troughs near hurricanes (Rodgers et al., 1986). However, the mechanisms linking total ozone to the weather are still poorly understood. Some of the TOVS ozone errors are due to the background emissivity fluctuations created by cirrus cloudtops. Likewise, polar cirrus may be causing some of the high latitude errors noted in TOMS ozone estimates (A.J. Krueger, private communication).

Conclusions

Semi-operational measurements of total ozone over the Earth from TOMS, TOVS, and DOBSON instruments were compared for the last 4 months of 1987. Each instrument has characteristic strengths and weaknesses, and each of the three systems makes serious errors under some conditions at some times.

The new infrared-based TOVS ozone estimates are generally in good agreement with the venerable ultraviolet-based TOMS measurements. However: TOVS resolution is lower, Reststrahlen causes over estimates above deserts, cirrus clouds cause under estimates near the equator, and temperature-profile and ozone-profile abnormalities cause significant errors during large scale stratospheric disturbances such as the breakdown of the Antarctic ozone hole. TOVS errors in estimating ozone over the deserts would be minimized by the addition of an 9 µm channel to the system.

The TOMS ozone estimates appear generally reliable. However, TOMS estimates at the edge of polar night, especially in the northern hemisphere, are much lower than the corresponding TOVS and DOBSON observations, and are also significantly affected by polar stratospheric clouds. TOMS detects arctic "mini-holes", but appears to over estimate their "depth".

Each site in the DOBSON network has a highly individual pattern of differences with respect to the two satellites. It is difficult to generalize, but satellite biases with respect to mid-latitude stations are often significant (approximately ±30 DU, when ±5 DU is expected), and systematic errors in the "ground truth" are indicated. Some DOBSON sites are completely inconsistent with both satellites and with nearby DOBSON sites.

The TOMS, TOVS, and DOBSON datasets are periodically reprocessed to improve data processing techniques. CAE's DOBSON network database sporadically receives additional and corrected values of the "ground truth" observations from previous years. By the end of 1990, NASA's NIMBUS-7 data processing team will issue "version 6" of the TOMS gridded database, with improved calibration and extended time-coverage. In the latter half of 1990, NOAA's Climate Analysis Center (CAC) will implement the TOVS physical algorithm and provide ozone observations operationally. In 1991, CAC may refine the TOVS algorithm after the ozone data characteristics are studied during the 1990 "ozone hole" episode.

Later in the 1990's, several other satellite systems (METEOR, GOES, UARS and EOS) will be launched carrying ozone sensors to monitor the Earth. Eventually, a consensus among the many observations of total ozone could become a daily operational meteorological and climatological parameter available on demand over the nation's digital communications networks.

References

CAES, 1988. Ozone data for the world. In: Catalog of Ozone Stations and Ozone Data for 1985-1987, Canadian Atmospheric Environment Service, Downsview, Ontario.

Chesters, D., Larko, D.E., and Uccellini, L.W., 1990. Satellite observations of ozone near tropopause folds during the 1982 Atmospheric Variability Experiment. Fifth Conference on Satellite Meteorology and Oceanography, London, U.K., Amer. Meteor. Soc., pp. 443-448.

Chesters, D. and Krueger, A.J., 1989. A video atlas of TOMS ozone data, 1978-88. Bull. Amer. Meteor. Soc., 70: 1564-1569.

Fleig, A.J., Klenk, K.F., Bhartia, P.K. and Gordon, D., 1982. User's guide for the Total Ozone Mapping Spectrometer (TOMS) instrument first year ozone-T dataset. NASA RP-1096, 50 pp.

Lienesch, J.H., 1988. Evaluation of an improved set of predictors for derivation of total ozone from TOVS measurements. J. Atmos. Oceanic Technology, 5: 625.

Muller, S. and Cayle, F.R., 1983. Total ozone measurements derived from TOVS radiances. Planet. Space Sci., 31: 779.

Prabhakara, C., Rogers, E.B., Conrath, B.J., Hanel, R.A. and Kunde, V.G., 1976. The NIMBUS-4 infrared spectroscopy experiment: observations of the lower stratospheric thermal structure and total ozone. Journal of Geophysical Research., 81: 6391.

Prabhakara, C., and Dalu, G., 1976. Remote sensing of the surface emissivity at 9 µm over the globe. Journal of Geophysical Research., 81: 3719-3724.

Rodgers, E., Stout, J., Sterenka, J. and Chang, S., 1990. Tropical cyclone-upper atmospheric interaction as inferred from satellite total ozone measurements. J. Appl. Meteor., 29: 934-954.

Schoeberl, M.R. and Krueger, A.J., 1983. Medium scale disturbances in total ozone during southern hemisphere summer. Bull. Amer. Meteor. Soc., 12: 1358-1365.

Sechrist, F.S., Petersen, R.A., Krueger, A.J., Uccellini, L.W. and Brill, K.F., 1986. Ozone, jet streaks and severe weather. Second Conference on Satellite Meteorology/Remote Sensing and Applications, Williamsburg, VA, Amer. Meteor. Soc., pp. 388-392.

Schoeberl, M.R. and Krueger, A.J., 1983. Medium scale disturbances in total ozone during southern hemisphere summer. Bull. Amer. Meteor. Soc., 12: 1358-1365.

Shukorov, A.Kh. and Shukorova, L.M., 1986. Some results of a determination of the total ozone content in the atmosphere from the 9.6 micron band IR absorption band of O3. Atmos. Oceanic Phys., 22: 180.

U.S. Standard Atmosphere, 1976. NOAA-S/T 76-1562, 227 pp.

WMO, 1985. Atmospheric Ozone. WMO Report No. 16, 3 volumes.

Figure Captions

Fig. 1. Example of a one-day comparison of total column ozone imagery, for 4 October 1987. This image is a frame from a color videotape animation available from the authors. The top row presents TOMS ozone observations: (1) the left-hand rectangle is presented in Cartesian coordinates; and (2) the top right-hand circle is TOMS presented in polar coordinates for the southern hemisphere. The middle row presents the two corresponding TOVS ozone images, using the same intensity scale as TOMS indicated just below the left-hand side of the TOVS rectangle. The bottom row presents three comparisons: (1) the left-hand rectangle shows the location and ozone values from the DOBSON ground stations for the same day; (2) the center rectangle presents the TOVS-TOMS difference over the rectangular maps, using the color scale indicated just above it; and (3) the scatterplot at the right presents the TOMS and TOVS ozone values observed at the DOBSON sites for the day, with the little embedded error bars indicating the mean and RMS differences for TOMS-DOBSON (right-hand bar), TOVS-DOBSON (middle bar) and TOVS-TOMS (right-hand bar) observations.

Fig. 2. Time-series averaged over the Earth of the daily ozone mean differences between TOMS, TOVS and DOBSON measurements, taken in pairs.

Fig. 3. Time-series of daily ozone observations by TOMS, TOVS and DOBSON instruments over the mid-latitude station at Wallops Island, Virginia.

Fig. 4. Time-series of daily ozone observations by TOMS, TOVS and DOBSON instruments over the arctic station at Longyear, Norway (also known as Spitzbergen Island).

Fig. 5. Time-series of daily ozone observations by TOMS, TOVS and DOBSON instruments over the Amundson-Scott site at the South Pole.