Comparison of GOME tropospheric NO2 columns with NO2 profiles deduced from ground-based in situ measurements
Boersma, K. F.
Kaiser, J. W.
Weiss, A. K.
Eskes, H. J.
Nitrogen dioxide (NO<sub>2</sub>) vertical tropospheric column densities (VTCs)
retrieved from the Global Ozone Monitoring Experiment (GOME) are compared to
coincident ground-based tropospheric NO<sub>2</sub> columns. The ground-based
columns are deduced from in situ measurements at different altitudes in the
Alps for 1997 to June 2003, yielding a unique long-term comparison of GOME
NO<sub>2</sub> VTC data retrieved by a collaboration of KNMI (Royal Netherlands
Meteorological Institute) and BIRA/IASB (Belgian Institute for Space Aeronomy)
with independently derived tropospheric NO<sub>2</sub> profiles. A first comparison
relates the GOME retrieved tropospheric columns to the tropospheric columns
obtained by integrating the ground-based NO<sub>2</sub> measurements. For a second
comparison, the tropospheric profiles constructed from the ground-based
measurements are first multiplied with the averaging kernel (AK) of the GOME
retrieval. The second approach makes the comparison independent from the a
priori NO<sub>2</sub> profile used in the GOME retrieval. This allows splitting the
total difference between the column data sets into two contributions: one that
is due to differences between the a priori and the ground-based NO<sub>2</sub> profile
shapes, and another that can be attributed to uncertainties in both the remaining
retrieval parameters (such as, e.g., surface albedo or aerosol concentration)
and the ground-based in situ NO<sub>2</sub> profiles. For anticyclonic clear sky conditions
the comparison indicates a good agreement between the columns (n=157, R=0.70/0.74 for
the first/second comparison approach, respectively). The mean relative difference
(with respect to the ground-based columns) is −7% with a standard deviation of
40% and GOME on average slightly underestimating the ground-based columns.
Both data sets show a similar seasonal behaviour with a distinct maximum of spring
NO<sub>2</sub> VTCs. Further analysis indicates small GOME columns being systematically
smaller than the ground-based ones. The influence of different shapes in the a
priori and the ground-based NO<sub>2</sub> profile is analysed by considering AK
information. It is moderate and indicates similar shapes of the profiles for
clear sky conditions. Only for large GOME columns, differences between the
profile shapes explain the larger part of the relative difference. In
contrast, the other error sources give rise to the larger relative
differences found towards smaller columns. Further, for the clear sky cases,
errors from different sources are found to compensate each other partially.
The comparison for cloudy cases indicates a poorer agreement between the
columns (n=60, R=0.61). The mean relative difference between the columns is
60% with a standard deviation of 118% and GOME on average
overestimating the ground-based columns. The clear improvement after
inclusion of AK information (n=60, R=0.87) suggests larger errors in the a
priori NO<sub>2</sub> profiles under cloudy conditions and demonstrates the
importance of using accurate profile information for (partially) clouded scenes.