1.6 Validation results
As defined by CEOS, the validation is the process of assessing by independent means the quality of the data products derived from the system outputs. Validation studies must involve a sufficiently large number of correlative measurements of sufficient quality, covering a large area and a wide range of conditions. Good candidates for the validation of GOMOS products are ground-based, balloon-borne (upper troposphere and stratosphere regions), and other satellite measurements (up to the lower thermosphere) such as MIPAS, OSIRIS, ACE, ... Validation studies based on the comparison of individual profiles and on statistical data analysis as well as algorithm verification studies are expected to provide feedback to the calibration, recommendation for the algorithm improvement, and support for the monitoring of geophysical data quality during the mission lifetime.
Various validation activities have been performed since Summer 2002. For O3 products for instance, vertical profiles of local density have been systematically compared with NDSC lidar measurements (methodology described in [Meijer et al., 2004]). The average relative difference (GOMOS-lidar) is between –2% and 1% at altitudes between 18 km and 22 km ( Figure 1.13 ). It is between 0 and –2% between 22 km and 32 km, between 0 and +1% between 32 km and 41 km, and between 0 and –5% between 41 km and 45 km. Best results are obtained at mid-latitudes. In these regions, the average relative difference (GOMOS-lidar) is between –1% and 4% for altitudes between 18 km and 40 km (). For the polar regions, a negative average difference (GOMOS-lidar) is calculated at all altitudes: about –5% for the relative value around 20 km; between 0 and –5% in the altitude range 29 km-36 km; between –5% and –10% in the altitude range 37 km-41 km. For the tropical cases, positive values up to 18% are calculated around 18 km, but this bias decreases to nearly 0 around 23 km. The comparison with balloon measurements at low latitudes also shows a positive average difference in the lower stratosphere.
Figure 1.13: Left figure: Vertical profiles of the average and of the dispersion of O3 local density from GOMOS products (red line), and measured by ground-based lidar instruments (blue line) between 0 and 50km. Right figure: Vertical profiles of the average relative difference between O3 local density from GOMOS products and measured by ground-based lidar instruments (green lines: mean profile of the distribution with variability envelop; black line: median profile of the distribution). Coincidence criteria between GOMOS profiles and lidar profiles are a distance lower than 500 km and a time difference lower than 20 h. Only GOMOS measurements from occultations in pure dark limb have been used.
Figure 1.14: Same as Figure 1.13 for mid-latitude measurements.
Figure 1.15: Same as Figure 1.13 for polar measurements.
Figure 1.16: Same as Figure 1.13 for tropical measurements.
Results of recent studies of validation of all species products from the current operational processor (IPF 5.00) were presented in the Atmospheric Chemistry Validation Experiment workshop held in December 2006. For O3 profiles measured in dark limb, from the comparison with measurements from lidar, sondes, balloon-borne instruments, MIPAS and OSIRIS satellite instruments, it is estimated a bias lower than 5% (upper troposphere, stratosphere and mesosphere). The precision is estimated to be lower than 10%. NO2 profiles have been found free of bias. Their precision is estimated to be about 20%. The precision of NO3 profiles is estimated to be about 30%. No systematic comparison has been performed for aerosol products, but individual profiles have shown a good agreement with balloon measurements. Results from the validation of HRTP have shown that best results are obtained between 23 km and 30 km; however, even in this altitude range, many profiles contain spurious values. Statistical analyses on GOMOS-GOMOS coincidences highlighted an overestimation of the additional error turbulence for O3 local density in the range 20 km -40 km.
It has been stated that the limb flagging is too conservative. It is recommended to revise this limb flagging to a criterion on the solar zenith angle. It has also been stated that the species flagging in the products needs to be improved.