given location, they are computed as a function of pressure using standard vertical profiles (Fig. 12). This
simulation is referenced Case 3.
When ground elevation reaches 1000m (e.g. around 15° S), the optical thicknesses are decreased by 0.006 and
0.008 for molecules and aerosols respectively. Variations on gazeous transmissions are weak since, in this
case, only oxygen transmission varies : ozone lies at about 20 km above ground, while Oort (1983) vapor
water climatology accounts for elevation effect. Globally, accounting for elevation leads to a maximum
difference of 0.05 on NDVI (Fig. 13), which is not negligeable (Maisongrande et al„ this issue)
Figure 12: Molecule (solid line) and aerosol(dash line) Figure 13: NDVI difference along the transect
optical thicknesses along the transect when accounting between Case 1 and Case 3.
for relief (Case 3).
43 Temporal monitoring of reflectances and NDVI
One may wonder if atmospheric corrections improve the temporal monitoring of reflectances and NDVI. As
an example. Figure 14 shows NDVI evolution for a pixel located in the Pampa's meadow (Argentina) and for
nearly four years after NOAA 11 launch. The absolute level of NDVI is modified and increased by
atmospheric corrections, however, atmospheric corrections do not reduce the noise on the temporal signal.
Although part of this noise is due to inaccurate characterization of the atmosphere, the previous sensitivity
study show that additional effects are required to explain its amplitude. Major candidates for short term
variability are i) residual clouds (e.g. Loudjani et al„ 1994); ii) misregistration and c hanging resolution with
scan angle; iii) surface directional effects (enhanced when atmospheric corrections are applied). In addition,
for long term studies, orbital and calibration drifts also hamper the quantitative use of AVHRR data.
NOAA/AVHRR-11 TOA and surtace - Argentina
Figure 14: Temporal evolution of NDVI. Solid line and dash line refer to TOA and surface NDVI respectively.
5. CONCLUSION
Correction of atmospheric effects has been applied to AVHRR time series at global scale except for
stratospheric aerosols produced by Pinatubo eruption in June 1991. These corrections lead to large increases
of NDVI, when compared to TOA calibrated NDVI. Sensitivity of the correction to uncertainties on
atmospheric characteristics has been assessed, as well as the effect of ground elevation. Most important
uncertaimties are due to aerosols effects (Fig 10) rather than to water vapor absorption effects (Fig. 11),
elevation effects are of a secondery importance. Atmospheric corrections do not lower the noise on the
temporal signal, but reducing this noise requires to account for a variety of perturbations. However, these
