Comparison of the daily ozone vertical profiles from satellite (SBUV2-SMOBA) and ground-based (Umkehr) observations

The global (2.5 deg x 2.5 deg) data base of the ozone vertical profile has been constructed by CPC NCEP/NOAA under SMOBA project utilizing daily SBUV/2 observations on board of the NOAA-14 satellite. Using the SMOBA data we interpolate a time series of ozone vertical profile to a location of the ground-based station Belsk (51N, 21E), Poland. Then, the ozone vertical profiles from Umkehr observations by the Dobson spectrophotometer are compared with the interpolated SMOBA time series for 2002. The biases and standard errors of the differences (SMOBA minus Umkehr ozone in percent of Umkehr ozone) are as follows: -4.2 +/- 4.5, 6.5 +/- 7.1, and 3.6 +/- 8.5 for combined Umkehr layers 1-4 (1000-31.2 hPa), 5-7 (31.2-3.91 hPa), and 8-10 (3.91-0.0 hPa), respectively. The best agreement is found for layer 7 (7.81-3.91 hPa) and 8 (3.91-1.95 hPa), i.e., -1.5 +/- 7.1 and 1.9 +/- 9.5, respectively. To examine an accuracy of the ozone profile from the Umkehr observations carried out under non-ideal conditions (zenith cloudy sky) we calculate the SMOBA-Umkehr differences separately for such Umkehr observations. It is found that the Umkehr method provides still valid ozone content in layers 7, 8, and 9. To identify sources of the SMOBA-Umkehr differences for various levels we make a stepwise linear regression based on the following explaining variables: aerosol optical depth at 320 nm (inferred from the concurrent measurements by the Brewer spectrophotometer) for the period of Umkehr observations, temperature and absolute vorticity at 6 (morning Umkehr) and 18 GMT (afternoon Umkehr) for various levels in the atmosphere (500, 250, 100, 50, 30, and 10 hPa). Meteorological data are taken from the NCEP/DOE Reanalysis-2 data. The regression explains about 30-40% of the variance of the differences. The regression coefficients exhibit altitudinal dependence; for example, an increase of aerosol optical depth yields negative (positive) SMOBA-Umkehr difference in lower (upper) Umkehr layers. A possibility of using much more complicated statistical models to explain the differences will be discussed.