Performance of the ground-based total ozone network assessed using satellite data

[1] Dobson and Brewer spectrophotometer and filter ozonometer data available from the World Ozone and Ultraviolet Data Centre ( WOUDC) were compared with satellite total ozone measurements from TOMS ( onboard Nimbus 7, Meteor 3, and Earth Probe satellites), OMI ( AURA satellite) and GOME ( ERS- 2 satellite) instruments. Five characteristics of the difference with satellite data were calculated for each site and instrument type: the mean difference, the standard deviation of daily differences, the standard deviation of monthly differences, the amplitude of the seasonal component of the difference, and the range of annual values. All these characteristics were calculated for five 5- year- long bins and for each site separately for direct sun ( DS) and zenith sky ( ZS) ozone measurements. The main percentiles were estimated for the five characteristics of the difference and then used to establish criteria for suspect'' or outlier'' sites for each characteristic. About 61% of Dobson, 46% of Brewer, and 28% of filter stations located between 60 degrees S and 60 degrees N have no suspect'' or outlier'' characteristics. In nearly 90% of all cases, Dobson and Brewer sites demonstrated 5- year mean differences with satellites to be within +/- 3% ( for DS observations). The seasonal median difference between all Brewer DS measurements at 25 degrees- 60 degrees N and GOME and OMI overpasses remained within +/- 0.5% over a period of more than 10 years. The satellite instrument performance was also analyzed to determine typical measurement uncertainties. It is demonstrated that systematic differences between the analyzed satellite instruments are typically within +/- 2% and very rarely are they outside the +/- 3% envelope. As the satellite instrument measurements appear to be better than +/- 3%, ground- based instruments with precision values worse than +/- 3% are not particularly useful for the analyses of long- term changes and comparison with numerical simulations.