anisotropic influences can dramatically alter the spectral contrast between Red and
NIR reflectances and result in angular biases (Cihlar et al. 1997; Schaaf et al.
2002). BRDF effects become more pronounced when atmosphere corrections are
applied to satellite VI data sets (van Leeuwen et al. 1999).
Fensholt et al. (2006) studied the dependence of NDVI on solar and viewing
geometries with MODIS and the Meteosat Second Generation (MSG) Spinning
Enhanced Visible and Infrared Imager (SEVIRI) sensor, and found higher Red
reflectances relative to NIR under backscatter conditions resulting in a decrease of
NDVI, while in the forward scatter direction, Red reflectances were more strongly
reduced due to shadowing relative to the scattered NIR, resulting in higher NDVI.
The EVI responds in an opposite manner, and has a positive bias in the more
sunlit, backscatter canopy view orientation due to the much stronger NIR signal
(Fig. 1.5).
1.3.1 Compositing Approaches
Coarse resolution sensors with wide swath and near-daily imaging are important in
obtaining sufficient acquisitions of cloud-free data necessary to improve temporal
and spatial monitoring of surface vegetation dynamics. However, their wide
swaths of over 2,000 km result in pronounced BRDF effects associated with
sensor-surface-sun observation geometries. The sequential VI imagery is thus
composited over set time intervals to reduce cloud and cloud shadow contamination as well as improve the viewing geometry quality of selected pixels in the
final product.
Standard compositing methods used in coarse resolution satellite data are based
on the maximum value composite (MVC) concept developed for the Advanced
Very High Resolution Radiometer (AVHRR) NDVI time series data (Holben
1986). The MVC method selects the highest NDVI value over a compositing cycle
to best represent the greenness status of an area for that period. In the early
AVHRR era, this was applied to non-atmosphere and non-BRDF corrected
satellite data and accurately presumed that the highest NDVI would occur on the
day with least cloud and aerosol contamination and smallest atmosphere optical
path length (i.e., the most nadir viewing geometry).
However, recent advancements in atmospheric correction have rendered the
MVC approach less useful as surface anisotropy influences are more prominently
revealed in the data. The MVC approach confuses higher VI values associated
with lower residual cloud/aerosol contamination from high VI values caused by
off-nadir viewing angles, i.e., the highest NDVI value within a compositing period
does not necessarily correspond to near-nadir sensor viewing angles or to the least
contaminated measurement (van Leeuwen et al. 1999). Lastly, higher VIs may also
result from over-correction of atmosphere contamination, resulting in a negative
bias in Red reflectances and positive bias in NDVI.
10
A. Huete et al.
NIR reflectances and result in angular biases (Cihlar et al. 1997; Schaaf et al.
2002). BRDF effects become more pronounced when atmosphere corrections are
applied to satellite VI data sets (van Leeuwen et al. 1999).
Fensholt et al. (2006) studied the dependence of NDVI on solar and viewing
geometries with MODIS and the Meteosat Second Generation (MSG) Spinning
Enhanced Visible and Infrared Imager (SEVIRI) sensor, and found higher Red
reflectances relative to NIR under backscatter conditions resulting in a decrease of
NDVI, while in the forward scatter direction, Red reflectances were more strongly
reduced due to shadowing relative to the scattered NIR, resulting in higher NDVI.
The EVI responds in an opposite manner, and has a positive bias in the more
sunlit, backscatter canopy view orientation due to the much stronger NIR signal
(Fig. 1.5).
1.3.1 Compositing Approaches
Coarse resolution sensors with wide swath and near-daily imaging are important in
obtaining sufficient acquisitions of cloud-free data necessary to improve temporal
and spatial monitoring of surface vegetation dynamics. However, their wide
swaths of over 2,000 km result in pronounced BRDF effects associated with
sensor-surface-sun observation geometries. The sequential VI imagery is thus
composited over set time intervals to reduce cloud and cloud shadow contamination as well as improve the viewing geometry quality of selected pixels in the
final product.
Standard compositing methods used in coarse resolution satellite data are based
on the maximum value composite (MVC) concept developed for the Advanced
Very High Resolution Radiometer (AVHRR) NDVI time series data (Holben
1986). The MVC method selects the highest NDVI value over a compositing cycle
to best represent the greenness status of an area for that period. In the early
AVHRR era, this was applied to non-atmosphere and non-BRDF corrected
satellite data and accurately presumed that the highest NDVI would occur on the
day with least cloud and aerosol contamination and smallest atmosphere optical
path length (i.e., the most nadir viewing geometry).
However, recent advancements in atmospheric correction have rendered the
MVC approach less useful as surface anisotropy influences are more prominently
revealed in the data. The MVC approach confuses higher VI values associated
with lower residual cloud/aerosol contamination from high VI values caused by
off-nadir viewing angles, i.e., the highest NDVI value within a compositing period
does not necessarily correspond to near-nadir sensor viewing angles or to the least
contaminated measurement (van Leeuwen et al. 1999). Lastly, higher VIs may also
result from over-correction of atmosphere contamination, resulting in a negative
bias in Red reflectances and positive bias in NDVI.
10
A. Huete et al.