S P A R C Stratospheric Processes And their Role in Climate A project of the World Climate Research Programme Home Initiatives Organisation Publications Meetings Acronyms and Abbreviations Useful Links

1.1. Introduction

In this chapter we discuss the sources of data used to study long-term trends in atmospheric ozone, including the measurement methods used, their characteristics and their error sources, in order to understand the extent to which they should be able to throw light on the actual long-term variation. This includes the characteristics of the instruments and those of the data analysis methods used to retrieve the ozone profile from the original signals obtained by the instruments. A discussion of this kind was included in the report of the 1988 Ozone Trends Panel (WMO 1990), but much has been learned since then, and more data sources have become available, specifically the UARS instruments MLS and HALOE, and the lidars, which now have a significantly longer data record.

The magnitude of the reduction in global total ozone, as reported in the WMO/UNEP 1994 Scientific Assessment of Ozone Depletion, appears to be in reasonable agreement among several available measurement techniques, for example surface-based Dobson and satellite-based SBUV and TOMS. The situation is not as good for measurements of trends in the vertical profile of ozone, especially in the lower stratosphere. The 1994 assessment found considerable differences between the trends derived from ozonesondes and from the SAGE II instrument below about 20 km altitude. Considerably better agreement was obtained above 20 km when comparing the SBUV, SAGE and Umkehr techniques for trends.

For the 1998 Assessments, these discrepancies need to be examined in much more detail than was done in 1994. The analysis presented here is intended to determine the drift uncertainties of the calibration of the available instruments, i.e. whether they are adequate in principle to determine ozone trends to the required accuracy, and to identify problems that need attention for future measurements. The discussion will also provide information as to the suitability of data sources in validating the measurements of other instruments. To this end, we describe how the data are produced for each instrument/technique, including the generation of the raw data and the processing used to obtain the final numbers for analysis. The actual performance of instruments can only be determined by intercomparison. These aspects will be discussed in Chapter 2.

Ozone profiling instruments can be generally classified into remote and in-situ devices. Remote instruments such as the satellite sensors (HALOE, MLS, SBUV and SAGE), and ground-based instruments (Dobson-Umkehr, lidar and microwave spectrometers) obtain ozone profile information at a distance by utilising the spectral characteristics of ozone absorption, mainly in the ultraviolet but also in the visible, near infrared and microwave regions of the spectrum. In-situ devices such as balloon-borne ozonesondes and airborne UV absorption (Dasibi-type) instruments sample ambient air drawn into the instrument which is analysed for ozone by chemical or spectral analyses. Both remote and in-situ techniques can be deployed from balloons and aircraft while satellite and ground-based measurements of ozone are obviously limited to remote techniques.

For the analysis of ozone trends, instruments and techniques must have sufficiently long-term records of data with adequate quality and coverage. Ground-based Dobson instruments, observing ozone profiles in the Umkehr mode, have the longest heritage of ozone measurements with substantial records at some stations from 1957. The network of instruments is routinely inter-calibrated and produces ozone profiles in the stratosphere between approximately 20 and 40 km. Vertical resolution is relatively coarse, and spatial coverage is limited by the small number of ground stations. Lidar instruments have fewer ground stations, but in principle, better vertical resolution and precision. They will be important monitoring instruments in the future, and their length of data record is now sufficiently long that they are considered here. Ozonesondes cover a lower altitude range than other techniques, and are important for monitoring the troposphere and lower stratosphere to about 25 km. Useful measurements are available from a small number of stations since about 1970.

Satellite instruments provide a more complete geographical coverage, but data records are shorter. SBUV (on Nimbus 7) and SBUV/2 on NOAA 11 provide daytime data for the stratosphere since 1979, and SAGE I and II provide solar occultation data (twice per orbit), also for the stratosphere, from 1979. More recently two new instruments, HALOE (infrared solar occultation)and MLS (microwave thermal limb emission) have flown on the Upper Atmosphere Research Satellite, launched in 1991. Its six-year record is not yet sufficient to identify trends satisfactorily, but they already provide useful data sources for validation, and will contribute to future trend assessments.

No single existing ozone profiling instrument or technique is capable of measurements at all altitudes, with adequate global and temporal coverage. Thus a combination of instruments and techniques is necessary. There is a range of requirements for data from an instrument or set of instruments to measure ozone trends. These include spatial and temporal coverage and resolution, long-term stability, long-term continuity, length of record and minimal sensitivity to other time varying quantities such as volcanic aerosol.

Horizontal spatial coverage is best obtained by satellite based sensors which measure continuously, such as MLS and SBUV, although SBUV can only make daytime measurements. Occultation instruments such as SAGE and HALOE obtain measurements only at the terminator in a pattern which depends on the spacecraft orbit. Ground-based instruments are limited by the total number of stations that can be provided. The most widely distributed ground-based instrument is the Dobson, for which there are about 90 stations world-wide, although only 19 of those which make Umkehr measurements place their data in the WOUDC.

Vertical spatial coverage varies somewhat. Most of the instruments discussed make measurements in the range from about 25-50 km, but in the critical lower stratosphere from about 15-20 km, only sondes, SAGE and HALOE make useful measurements.

High vertical resolution (better than 1 or 2 km) is obtained by sondes and the occultation instruments SAGE and HALOE and over part of the height range by Lidars. MLS, SBUV and Umkehr all have a rather poorer resolution, around 5±10 km, giving rise to some questions of comparison and interpretation.

All of the radiometric data sources used have some form of internal radiometric calibration. This is particularly simple and reliable in the case of the occultation instruments. The sondes are usually, but not always, referenced to a nearby Dobson total ozone measurement. The precision, accuracy and long-term stability of all of the data sources is discussed in detail in this chapter. Most instruments, namely SAGE, HALOE, SBUV, Lidar and Umkehr are affected to some extent by volcanic aerosol, which has long-term and irregular time variation, and which is eliminated to varying degrees of success. Solar cycle effects may be present in the UV instruments, Umkehr and SBUV, and long-term temperature trends may affect any of the data sources in various ways.

All data sources are subject to problems of long-term continuity of instrumentation. The satellite instruments have a finite lifetime, even if they are reasonably stable during their life, great care is needed when comparing with successor instruments. The same is true of ground-based instruments, but on a shorter time-scale where the temptation to improve an instrument may induce long-term variations in its record. Even the one-off sondes may be subject to long-term changes in manufacturing techniques and suppliers.

The length of the data record is of primary importance in determining trends. The UARS instruments, MLS and HALOE, have only about 5 or 6 years, whilst the SAGE I/II and SBUV record is 18 years, and the Dobson Umkehr record goes back to 1957 at some stations. The question of the influence of data record, sampling and natural variability on the accuracy with which trends can be measured is not strictly part of the error analysis of an instrument, and will be treated separately in Chapter 3.

The error analysis of the data sources follows the same general approach as the 1988 Ozone Trends Panel (WMO 1990). The retrieved profile is taken to be a smoothed version of the true profile, with smoothing functions given by the averaging kernels (Rodgers, 1990). The width of the averaging kernels describes the vertical resolution of the measurement, and the errors in the smoothed profile estimate are analysed in terms of random and systematic errors.

For each instrument type, sources of error are identified and quantified as far as possible, the results being stated in terms of errors in the retrieved ozone profile. For estimating the accuracy with which long-term trends in ozone may be measured, it is particularly important to identify long-term drifts in the systematic errors which could be interpreted as a long-term trend in ozone. Constant systematic errors are of less importance for trends, and with a long record, random errors will generally be reduced to negligible proportions. However the random and constant systematic errors are still important when considering instruments used for validation of others. The identification of trend error sources is in its infancy, and it can involve many subtle effects that may not normally be considered for single-profile error analysis. Therefore the analysis in this chapter should be regarded as preliminary, and be treated with caution.