18. SOME RESEARCH AND APPLICATIONS IN THE CSIRO
(AUSTRALIA) EARTH OBSERVATION CENTRE ON SCENE
BRIGHTNESS DUE TO BRDF
165
interact with Maximum Value NDVI compositing and atmospheric
correction. Basically, they concluded the best choice of method is to correct
for “BRDF” first, then composit and finally atmospheric correct. The reason
for this is that the BRDF effect increases the NDVI, which leads to problems
with the compositing. Essentially, the compositing (selecting pixel with
maximum NDVI over a period) was introduced to select against cloud and
edge pixels since these tend to have reduced NDVI. The resulting selected
pixel from the compositing period supplies the Maximum Value NDVI.
However, since atmospheric correction alone increases the BRDF effect, it
makes the compositing result worse and results in pixels from the edges or
with greater Sun/look variations become the selected pixels.
It seems that any attempt to provide consistent and standardized AVHRR
data–especially for environmental monitoring–must account for BRDF and
should not be atmospherically corrected without account of BRDF.
There is little doubt of the value or need for scene brightness correction
in areas as diverse as AVHRR NDVI and video data mosaicking, In each
case, however, we still have a fundamental issue of whether there is a
consistent and simple typology of BRDF? In particular:
Is there a consistent typology of BRDF, which relates to land cover
structure?
Is it representable by simple BRDF (e.g., Kernel) functions that can
provide consistency and standards?
Can remote sensing consistently monitor changes in the coefficients of
the function and/or changes in functional form?
Do the changes recorded key in to significant structural changes in the
surface cover?
The last point raises the land cover structure question.
6.
LAND COVER AND STRUCTURE
Land cover structure concerns vertical and horizontal spatial variation of
components of the vegetation—or the ‘gappiness’ of canopies. It is a key
element in the ecology of a landscape and a key element in fluxes of water,
heat and carbon. Both in models of the ecology and the remote sensing,
biomass is not enough! Land cover derived from spectral data is dominated
by cover. However, it is the spatial distribution, gappiness and variance of
canopies that is needed for monitoring major system changes.
To study scene brightness and structure in forests, CSIRO has also spent
effort on obtaining airborne and field data in the past. For example, low and
(AUSTRALIA) EARTH OBSERVATION CENTRE ON SCENE
BRIGHTNESS DUE TO BRDF
165
interact with Maximum Value NDVI compositing and atmospheric
correction. Basically, they concluded the best choice of method is to correct
for “BRDF” first, then composit and finally atmospheric correct. The reason
for this is that the BRDF effect increases the NDVI, which leads to problems
with the compositing. Essentially, the compositing (selecting pixel with
maximum NDVI over a period) was introduced to select against cloud and
edge pixels since these tend to have reduced NDVI. The resulting selected
pixel from the compositing period supplies the Maximum Value NDVI.
However, since atmospheric correction alone increases the BRDF effect, it
makes the compositing result worse and results in pixels from the edges or
with greater Sun/look variations become the selected pixels.
It seems that any attempt to provide consistent and standardized AVHRR
data–especially for environmental monitoring–must account for BRDF and
should not be atmospherically corrected without account of BRDF.
There is little doubt of the value or need for scene brightness correction
in areas as diverse as AVHRR NDVI and video data mosaicking, In each
case, however, we still have a fundamental issue of whether there is a
consistent and simple typology of BRDF? In particular:
Is there a consistent typology of BRDF, which relates to land cover
structure?
Is it representable by simple BRDF (e.g., Kernel) functions that can
provide consistency and standards?
Can remote sensing consistently monitor changes in the coefficients of
the function and/or changes in functional form?
Do the changes recorded key in to significant structural changes in the
surface cover?
The last point raises the land cover structure question.
6.
LAND COVER AND STRUCTURE
Land cover structure concerns vertical and horizontal spatial variation of
components of the vegetation—or the ‘gappiness’ of canopies. It is a key
element in the ecology of a landscape and a key element in fluxes of water,
heat and carbon. Both in models of the ecology and the remote sensing,
biomass is not enough! Land cover derived from spectral data is dominated
by cover. However, it is the spatial distribution, gappiness and variance of
canopies that is needed for monitoring major system changes.
To study scene brightness and structure in forests, CSIRO has also spent
effort on obtaining airborne and field data in the past. For example, low and