Uncertainties in upper stratospheric ozone trends from 1979 to 1996

The time series of differences in coincident measurements of ozone by Strat ospheric Aerosol and Gas Experiment (SAGE) and by Solar Backscattered Ultra violet (SBUV), SBUV/2, Umkehr and Microwave Limb Sounder (MLS) are analyzed , and the slopes in the differences are calculated. SAGE ozone measurements are also compared against those by HALOE. The purpose of these comparisons is to look for statistically significant nonzero slopes which could indica te long-term calibration problems in one or more of the measurement systems . It is found that the slopes are remarkably similar between the Northern a nd Southern Hemisphere midlatitudes, and, apart from a few exceptions, the slopes are also similar in the tropics. Slopes of MLS-SAGE differences and HALOE-SAGE trends from approximately 1992 to 1996 have values of approximat ely -0.5 +/- 0.4% yr(-1) (95% confidence limits) in Umkehr layers 7-9 (whic h are centered at similar to 37, 42, and 47 km altitude). Umkehr-SAGE slope s for 1979-1996, however, are almost all positive and in the range -0.1-0.3 % yr-(1) for Umkehr layers 4-8, while SBUV-SAGE slopes for 1979-1989 are es sentially zero in layers 4-7 and 0.3-0.4% yr(-1) in layers 8 and 9. Averagi ng all these results with SBUV-SAGE II slopes from 1985 to 1989, the other sensors minus SAGE slopes are most likely between 0.2 and -0.2% yr(-1) from similar to 20 to 40 km altitude. The results indicate slightly negative sl opes in Umkehr layers 5-7 and positive slopes in the other three layers. Th ere thus appears to be no overall drift in the SAGE ozone measurements from 1979 to 1996, but SAGE sunrise/sunset trend differences >40 km altitude, c ombined with the more accurate SBUV-SAGE slopes for 1979-1989, suggest a mo st likely slope range of 0.4 to -0.4% yr(-1) between 40 and 50 km altitude. SBUV/2 measurements from 1989 to 1994 have an upward trend with respect to SAGE measurements of similar to 0.7% yr(-1) with some altitudinal structur e: this slope exceeds the estimated 95% uncertainties on the SBUV/2 trends.