Trends in Upper-stratospheric Ozone

AGU Spring Meeting
Boston
27 May 1998

Mike Newchurch
Atmospheric Science Department
University of Alabama in Huntsville
mike@atmos.uah.edu

Research sponsored by
IOC/SPARC

Research supported by
JMA, NASA, NOAA, WMO, UKDOE, UNEP

Co-Authors

Detailed estimates of individual measurement-system uncertainties that affect ozone-trend calculations provide the following conclusions:

(1) SAGE II version 5.96 observations may be used to calculate trends within 0.3 %/yr at 20 km and within 0.1 %/yr at 40 km at the 95% confidence level.

(2) Analogous estimates for SAGE I with the Wang et al. [1996] altitude correction are 0.6 %/yr at 20 km and 0.1 %/yr at 40 km.

(3) Trend-uncertainty estimates for Dobson Umkehr observations are 0.4 %/yr (2s ) between 20-40 km.

(4) NIMBUS-7 SBUV version 6.0 observations provide 2s trend constraints between 0.3-0.5 %/yr in the middle and upper stratosphere.

(5) HALOE potential drift uncertainties are similar to SAGE

(6) MLS uncertainties are intermediate SAGE and Dobson Umkehr.

(7) Estimates of the lidar trend uncertainties are significantly better than all other techniques.

Additional analyses of individual sensor systems indicate the following conclusions:

(1) SAGE-II sunrise/sunset ratios differ by ~10% for unexplained reasons above 45 km and below 22 km.

(2) Some residual aerosol interference remains in the ozone retrievals for 2 years after Mt. Pinatubo.

(3) The aerosol effect in the Dobson Umkehr observations is accounted for within 1% in ozone excluding the periods for one year after El Chichon and Mt. Pinatubo.

(4) Concatenation of NIMBUS-7 and NOAA-11&12 SBUV/2 series [SBUV(/2)] with a -15 to +10% empirical overlap adjustment produces a continuous ozone series from 1978-1994. However, problems with the NOAA-11&12 orbital drift and sensor calibration diminish its accuracy. Volcanic aerosols within 1 year of the El Chichon and Mt. Pinatubo eruptions affect the SBUV(/2) ozone retrievals.

Detailed regression analyses of coincident pairs of SAGE and correlative sensors between 1979-1996 indicate the following conclusions:

(1) Dobson Umkehr observations constrain drifts in the SAGE I/II ozone series to within 0.2 %/yr (2s ) between 20-40 km. Fig. 1.

(2) The NIMBUS-7 SBUV observations constrain SAGE I/II drifts to within 0.2 %/yr between 30-50 km. Fig. 2.

(3) Both Dobson Umkehr and SBUV measurements indicate that the SAGE-I altitude correction is adequate.

(4) Over a shorter period (1991-1996), HALOE and MLS observations constrain SAGE-II drifts between 0.3-0.8 %/yr varying with altitude between 20-50 km. Fig. 3.

(5) Over various periods between the 17-year Dobson comparisons and the 5-year UARS comparisons, lidar observations constrain SAGE-II drifts to 0.4-0.7 %/yr between 20-50 km. Fig. 3.

(6) Correlative measurements confirm that the SAGE-II sunrise/sunset ratio possesses an unexplained ~10% anomaly above 45 km.

Blind intercomparison of the statistical models from 10 independent groups analyzing 3 test datasets spanning the range of potential ozone time series indicate:

(1) Agreement between models to within 0.03 %/yr, at the 95% confidence level (smaller than the average individual model trend uncertainty) in the most benign case (complete continuous time series). Fig. 4.

(2) In the case of a discontinuous time series with irregular seasonal coverage all models agreed to within 0.2 %/yr for annual mean trends and to within 0.4 %/yr for worst-case seasonal trends with the model-to-model variance less than or equal to the average model trend uncertainty. Fig. 5.

(3) A major part of this variance could be attributed to the details of how a particular model handles missing data.

(4) Variation with periods equal to or less than or equal to the QBO exert little influence on calculated ozone trends

(5) Because the period between El Chichon and Mt. Pinatubo (~9 years) is close to the 11-year solar cycle, decadal variations in ozone are difficult to correctly partition between those two influences. Fortunately their combined influence is relatively minor.

An example comparison of a time series of ozone observed independently by SAGE I/II, Dobson Umkehr, and SBUV(/2) at 40 km 40^o North show very similar features.

(1) This region dominated by gas-phase reactions is an altitude where changes in ozone were originally predicted to occur. Fig. 6abc.

(2) Independent standard statistical models applied each to SAGE I/II, SBUV(/2), and Dobson Umkehr produced consistent trend results. Fig. 7.

(3) The largest negative trend of -0.6±0.2 (2s ) %/yr occurred at 40 km in mid latitudes diminishing to -0.1±0.1 (2s ) %/yr at 25 km and again becoming larger (more negative) below 25 km.

(4) These trends are largest in the extra tropics, show a factor-of-two seasonal variation with the most negative trends in the winter. Fig. 8,9,10.

(5) At 40 km, the trends exhibit substantial north/south hemispheric symmetry. Fig. 11.

(6) SAGE I/II and Dobson Umkehr trends show good agreement

(7) SBUV/SBUV(2) combined record shows generally less negative trends. We place less confidence in the SBUV/SBUV(2) result due to potential problems with the present version (6.1.2) of the NOAA-11 SBUV(2) data.

These analyses provided an opportunity to make a first attempt at combining the uncertainties from both the sensor systems and from the statistical models.

(1) Combining the trends and uncertainties estimated from all available measurement systems and analogous statistical models in a variance-weighted, root-sum-of-squares sense for northern mid-latitude measurements results in a statistically significant (at the 2s level) negative trend at all altitudes between 20-45 km. Fig. 12.

(2) The combined trend has a local extreme value of -0.74± .0.20 (2s ) %/yr at 40 km altitude and a local minimum value of -0.20± .0.18 (2s ) %/yr at 30 km altitude. Fig. 13.

(3) These variance-weighted rss trend estimates differ somewhat from, but are consistent with, the simple-average results cited above.