Sie sind auf Seite 1von 11

A PROCESSING SYSTEM FOR GENERATING STANDARD

NADIR BRDF ADJUSTED SURFACE REFLECTANCE (NBAR)


PRODUCTS FOR LANDSAT
Fuqin Lia, Lan-Wei Wanga and David L. B. Juppb
a
National Earth Observational Group, Geoscience Australia, GPO Box 378,
ACT, 2601, Australia Tel: 61 2 62495867; fax 61 2 62499910
Fuqin.Li@ga.gov.au
Lan-Wei.Wang@ga.gov.au
b
CSIRO, Marine and Atmospheric Research, GPO Box 3023, ACT 2601,
Australia, Tel: 61 2 62465867
David.Jupp@csiro.au
Abstract
Normalising for atmospheric, land surface bidirectional reflectance distribution
function (BRDF) and terrain illumination effects are essential in satellite data
processing. It is important both for a single scene when the combination of
land cover, sun, view angles and terrain slope angles create anisotropy and
for multiple scenes in which the sun angle changes.
Geoscience Australia (GA) is establishing a procedure to conduct physically
based atmospheric BRDF and terrain illumination correction for moderate
spatial resolution satellite imagery (10-100 m) such as Landsat using a
coupled atmospheric and BRDF model. In particular, the method is not
dependent on the image data, does not need extensive field data, can be
applied equally to different environments and used with different sensors in a
consistent way. Furthermore, the corrected surface reflectance derived using
this method can be used to calibrate and cross-calibrate satellite sensors.
More importantly, the normalized reflectance can be used for time series
analysis to trace climate change and land cover variation using multiple
sensors (including satellite, airborne and ground based).
In this paper, we will describe the algorithm being progressed at GA.
Preliminary results from the algorithm will be compared with ground based
reflectance measurements for selected validation sites. The paper will also
discuss how the environmental input data for the model, such as aerosol,
water vapour and BRDF parameters are selected and applied.

Introduction
Landsat imagery has been used extensively for terrestrial environment
assessment and natural resources inventory for decades (Goward and
Williams, 1997). However, the radiance received by satellite sensors does not
directly characterize the underlying reflectance of surface objects since, even
for the same surface type, the received radiance at the satellite is modified by
the variations of atmospheric properties, the solar zenith, azimuth and sensor
view angles (Li et al., 2010). The BRDF and atmospheric effects are further
complicated over mountainous or hilly areas where topography affects both
irradiance and BRDF (Dymond et al., 2001). Consequently, to obtain surface
reflectance from remote sensing observations it is necessary to process the
data to reduce or remove these effects. The retrieved surface reflectance can

1
then be used to measure land surface change through a time series or
characterise land cover in a way that does not need to be tailored to each
image. Allowance for atmospheric, BRDF and terrain illumination effects is
also essential in field validation and it allows various sensors and satellite
systems to be directly compared for specific surface covers.

Atmospheric, BRDF and terrain illumination correction has been the focus of
work by many scientists during recent years (Vermote et al., 1997, Schaaf et
al., 2002, Danaher et al, 2001, Furby and Campbell, 2001, Shepherd and
Dymond, 2003, Richter et al., 2006). However, most methods have corrected
atmospheric, BRDF and terrain illumination separately. Some have used
empirical models or parameters, particularly for terrain correction. In this
paper, a physically based model which combines atmospheric and BRDF for
both flat (horizontal) and rugged surfaces is introduced, extending the work
reported by Li et al. (2010).

Atmospheric and BRDF correction for a horizontal (flat) surface

According to Hu et al (1999), Vermote et al (1997) and Li et al. (2010),


radiance received by satellite sensors for non-uniform surface can be
expressed as:
⎛ tV t S ρ S (θ S , θV , δϕ ) + tV t d (θ S ) ρ + t S t d (θV ) ρ ' + ⎞
1 ⎜ ⎟
LTOA = L0 + E 0' cos(θ S )⎜ [t + t (θ )][t S + t d (θ S )]S ( ρ ) 2 ⎟ (1)
π ⎜ t d (θ S )t d (θV ) ρ + V d V ⎟
⎜ 1− Sρ ⎟
⎝ ⎠
Where LTOA is sensor radiance, L0 is path radiance and E0' is solar
exoatmospheric irradiances. ρs (θS, θV, δφ) is the surface reflectance (the BRF
or bi-directional reflectance factor which is π times the BRDF), θS is solar
zenith angle and θV is view zenith angle, δφ is relative azimuth between the
sun and view directions. tS and tV are direct transmittance in the solar and
view directions. td(θS) and td(θV) are diffuse transmittance in the solar and view
directions; ρ , ρ ' and ρ are surface hemispherical-directional, directional-
hemispherical and hemispherical-hemispherical reflectance (or bi-
hemispherical) factors, respectively, and S is atmospheric albedo which is the
bi-hemispherical ratio between upwelling radiation from the surface and the
downwelling radiation
For a horizontal surface, if the semi-empirical BRDF kernel models are used,
equation (1) can be solved as a quadratic equation (Li et al., 2010):
2
(1 − a ) S (1 − Sρ m ) ρ + [a + ρ m (1 − a) S ]ρ − ρ m = 0 (2)
where ρm is the atmospherically corrected Lambertian reflectance and a is
defined as:
a = 1 / α wk (1,α1,α 2)[ fV f S B(θ S ,θV , δϕ ,α1,α 2) + fV (1 − f S )α bk (1,α1,α 2,θV )
+ f S (1 − fV )α bk (1,α1,α 2,θ S ) + (1 − f S )(1 − fV )α wk (1,α1,α 2)]

2
Where fV=tV/(tV+td(θV))= tV/TV and fS=tS/(tS+td(θS))= tV/TS
BRDF shape functions B (θS, θV, φ, α1, α2) and αbk and αwk are derived from
BRDF kernel models and black sky and white sky albedo respectively (Lucht,
et al., 2000). They can be expressed as:
Fvol F
ρ S (θ S , θV , δϕ ) = Fiso (1 + K vol + geo K geo ) = Fiso B(θ S , θV , δϕ , α1, α 2)
Fiso Fiso

α bk (1, α1, α 2, θ ) = ρ / Fiso = 1 + α1( −0.007574 − 0.070987θ 2 + 0.307588θ 3 )


+ α 2( −1.284909 − 0.166314θ 2 + 0.041840θ 3 )

and α wk (1, α1, α 2) = ρ / Fiso = 1 + 0.189184α1 − 1.377622α 2


where Fiso is the isotropic contribution. Fvol and Fgeo are the weights for
volume-scattering and geometric-optical contributions and Kvol and Kgeo are
volume-scattering and geometrical-optical scatting kernel functions. Kvol and
Kgeo are defined functions of θS, θV and δφ. α1=Fvol /Fiso and α2=Fgeo /Fiso.

Atmospheric and BRDF correction for an inclined surface


Let Eh denote the total irradiance on a horizontal surface, Ehdir the direct
component of irradiance on a horizontal surface, and E hdif the diffuse
component of irradiance on a horizontal surface. Because Ehdir = t S E0' cos(θ S ) ,
Ehdif = td (θ S ) E0' cos(θ S ) and E h = Ehdir + E hdif = TS E0' cos(θ S ) , fS can also be
expressed as:
f S = t S / TS = E hdir / E h and 1 − f S = td (θ S ) / TS = E hdif / E h
In this notation, eq. (1) can be rewritten:
⎛ E hdir [ fV ρ S (θ S , θV , δϕ ) + (1 − fV ) ρ '] + ⎞
⎜ ⎟
T
= L0 + V ⎜ ⎟
2
(3)
π ⎜ E hdif [ fV ρ + (1 − fV ) ρ ] + E h ρ S ρ ⎟
LTOA
⎜ 1 − S ρ ⎟⎠

In a similar way to the horizontal surface, equation (3) on an inclined surface
can be expressed:
⎛ E dir [ fV ρ S (it , et , δϕ t ) + (1 − fV ) ρ '(it )] + ⎞
T ⎜ ⎟
LTOA = L0 + V ⎜ dif S(ρ ) ⎟ 2 (4)
π ⎜ E [ fV ρ ( et ) + (1 − fV ) ρ ] + E ⎟
⎜ − ρ ⎟
⎝ 1 S ⎠
Where E, Edir and Edif are total irradiance, direct component of irradiance and
diffuse component of irradiance on a sloping surface, respectively; it and et
are incident and exiting zenith angles between the sun and view directions
and surface normal and δφt is the relative azimuth angle between incident and
exiting directions in the slope geometry.
E, Edir and Edif can be estimated (Shepherd and Dymond, 2003):

3
max(cos(it ),0)
E dir = E hdir
cos(θ S )

E dif = E hdif Vd + E adj and E adj = E hVt ρ adj


Where ρadj is average reflectance of adjacent objects (the terrain reflectance)
and it, et, Vd and Vt can be calculated as:
cos(it ) = cos(θ S ) cos(θ t ) + sin(θ S ) sin(θ t ) cos(ϕ S − ϕ t )
cos(et ) = cos(θV ) cos(θ t ) + sin(θV ) sin(θ t ) cos(ϕV − ϕ t )
Vd = [1 + cos(θ t)] / 2
Vt = [1 − cos(θ t )] / 2 = 1 − Vd
Where θt is the slope angle, φt is aspect angle of the slope and φS and φV are
solar azimuth and satellite view azimuth angle respectively.
Define Rdir = Edir/Eh, Rdif = Edif/Eh and R = E/Eh, then
Equation (4) can be expressed as:
⎛ R dir [ fV ρ S (it , et , δϕ t ) + (1 − fV ) ρ '(it )] + ⎞
ρm ⎜ ⎟
= ⎜ dif S (ρ ) ⎟2 (5)
1 − Sρ m ⎜ R [ fV ρ ( et ) + (1 − fV ) ρ ] + R ⎟
⎜ 1 − S ρ ⎟⎠

In a similar way to the horizontal case, equation (5) can be solved using a
quadratic equation:
2
( R − at ) S (1 − Sρ m ) ρ + [at + ρ m (1 − at ) S ]ρ − ρ m = 0 (6)
with
a t = 1 / α wk (1,α1,α 2){R dir [ fV B(it , et , δϕ t ,α1,α 2) + (1 − f v )α bk (1,α1,α 2, it )]
+ R dif [ fV α bk (1,α1,α 2, et ) + (1 − f v )α wk (1,α1,α 2)]}

Algorithm implementation
Figure 1 outlines the overall work flow for the atmospheric and BRDF
correction algorithm. Landsat images were divided into four rectangles and
MODTRAN (The MODerate resolution atmospheric TRANsmission) run for
the 9 sets of coordinates. For any pixel in the image, MODTRAN outputs are
then interpolated from the 9 coordinates using bilinear interpolation in each
rectangle to get pixel values. Following this, the Lambertian reflectance ( ρ m )
is computed. It is possible for this image to be an intermediate product but in
general, if the BRDF shape function is known then equation (2) can be used
to obtain a surface reflectance normalised to some specified standard sun
and view geometry. If both DEM data and BRDF shape function are known,
the BRDF and terrain illumination corrected surface reflectance can also be
computed through equation (6).

4
Figure 1. Diagram for atmospheric, BRDF and terrain illumination correction
algorithm

Model input data implementation


Atmospheric profiles and water vapour have been difficult inputs to provide for
many years because there are only a few meteorological stations which
launch balloons. The MODTRAN group, at the time they developed Low
Resolution Transmission (LOWTRAN) also established a set of standard
atmospheres based on existing data. These are still commonly used as
defaults in radiative transfer modelling. Accurate water vapour can
theoretically be obtained from the World Meteorological Organization (WMO)
radiosonde network. This would appear to be particularly convenient in
Australia since the Landsat overpass time is near to 2400 hours GMT and it is
a standard time at which to launch balloons as required by WMO. However,
there are still relatively few stations which have radiosonde data and they do
not cover the whole continent. In this algorithm, for the convenience of the
automatic processing system and the data availability (limitation), the default
atmospheric profile from MODTRAN and total precipitable water content
product from NOAA NCEP are used. The NOAA NCEP (National Centers for
Environmental Prediction) reanalysis surface level precipitable water content
product has four times a day individual observation (Kalnay et al., 1996). The
data are available from 1948 to the present and are updated daily. They are
provided on a 2.5x2.5 degree grid with global coverage. Note that the quality
of the data and the way it affects the accuracy of corrections used in the
algorithm needs further investigation in the future.
In order to retrieve surface reflectance accurately, aerosol information and in
particular, quantity is needed; optical depth is a critical parameter. One way to
get data-independent aerosol optical depth is from the global ground
sunphotometer network (http://aeronet.gsfc.nasa.gov). However, there are
only a few sunphotometer stations in Australia and this network has not yet
established sufficient information for operational use. This situation is likely to

5
change in the future. One alternative is to use satellite-derived aerosol
products, such as from the Multiangle Imaging SpectroRadiometer (MISR) or
Advanced Along-Track Scanning Radiometer (AATSR).
Satellite products such as those from MISR and AATSR only provide
infrequent aerosol data and cannot provide the data on a daily basis. This
means that these sensors may not be able to provide data for every satellite
overpass or be relied on even for less frequent coverage. Thus the aerosol
data used in the algorithm is based on a decision tree according to the data
quality and availability. Basically, our first choice is aeronet data. If no aeronet
data are available, then satellite derived data are selected. For details see
Fig. 2.

Figure 2. Decision tree for aerosol data selection

BRDF parameters are obtained from the MODIS BRDF model product
(Schaaf et al., 2002) which is the result of obtaining atmospherically corrected
MODIS data over 16 days. During this time, it is possible to sample directional
reflectance for a range of view angles at each location with view zenith angles
in the range from -60° to +60°. A combination of the Ross Thick volume kernel
and the Li-sparse Reciprocal (Schaaf et al., 2002) geometric kernel models is
used to approximate these data.
In this study, it has been assumed that the BRDF shape needed is one
defined for large regions (regional scale). Finer scale variation in BRDF
functions obtained from MODIS, MISR and POLarization and Directionality of
the Earth's Reflectances (POLDER) have been assumed to be caused by the
specific geometry of the local terrain and cover rather than an average
statistical regional effect. In this way, local variations due to MODIS sampling
and sub-pixel cloud effect are also reduced. To illustrate the methods in this
paper, the regional effect was estimated by averaging the coefficients of the

6
MODIS BRDF product for the appropriate date over the entire scene.
However, for routine implementation it is assumed that stable and
characteristic BRDF shape functions will be defined by spatial region or land
cover at an appropriate scale for the sensor involved. When available, this
can be easily implemented with little increase in algorithm complexity or
reduction in speed
Approximate Digital Elevation Model (DEM) data are very important for terrain
illumination correction. Geoscience Australia and CSIRO released the 1
second SRTM (the Shuttle Radar Topographic Mission) derived Digital
Surface Model (DSM) and Digital Elevation Model (DEM) of Australia in
February, 2010 (see Tickle and Gallant, 2010). For the topography correction,
the DSM data were used for terrain illumination correction after registration
and resampling to the Landsat scale and allowing for noise and other effects
in the DSM that can disturb the terrain correction. The appropriate scale for
the correction is at hillslope scale rather than at the micro-relief scale and this
was taken into account. For places where such DSM data are not available it
is possible to achieve a similar result based only on the global 3 second
SRTM data (as used in Google Earth).

Validation of combined atmospheric and BRDF correction


Two Australian sites were used to test the algorithm. The first one is in the
Gwydir area of New South Wales (centre location 29.58S, 149.62E). The
Gwydir area is covered by forests, crops, grass and bare soil. The second site
is at Lake Frome, which is a dry salt lake in northern South Australia (centre
location 30.78S, 139.68E). The Gwydir campaign was conducted from the
23rd to the 25th of June 2008. A Landsat 7 image was acquired on the 22nd of
June one day prior to the start of field work. The Lake Frome experiment was
undertaken between February 9th and 14th, 2009. A Landsat 7 overpass
occurred on February 12th, 2009. Just prior to the field experiment, a Landsat
5 image was acquired on February 4th. The sampling and processing of the
field data were described in detail in Li et al. (2010).
Figure 3 compares the processed field ASD measurements resampled to
Landsat bands with the retrieved Landsat 5 & 7 reflectance data. Both ASD
and Landsat reflectance were normalised to the same view and solar angles
using the (regional) BRDF model selected for image processing. The figure
shows that the results are generally good with RMSD for both sites being less
than 0.028 (reflectance units) and R values both above 0.950. The data for all
sites and all bands are combined in these data sets and include positioning,
data collection and other unavoidable sources of variation. For the Gwydir
area, the reflectance values retrieved from Landsat are slightly lower than
those from the ground measurements (Figure 3a). This is partly due to the
fact that the image was acquired during the Australian winter period since a
lower solar angle creates a lower signal. In this situation some areas can
generate lower signals due to specific shadow effects etc. A BRDF shape
function is generally not able to account for these effects which are often not
reciprocal. For the Lake Frome data, there was a better agreement since the
experiment was conducted during the summer (Figure 3b). Both Landsat 5

7
and 7 result in good agreement with ground measurements except for band 4
where Landsat reflectance values were generally higher than the ASD values.

0.7 0.7
band1 R=0.956 band1 R=0.985
0.6 0.6
band2 RMSD=0.02 band2 RMSD=0.02
band3 band3
0.5 0.5

Landsat 5&7 BRF


band4 band4
Landsat 7 BRF

0.4 band5 0.4 band5


band7 band7
0.3 0.3

0.2 0.2

0.1 0.1

a b
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

ASD BRF ASD BRF

Figure 3. Comparison between processed ASD diffuse corrected and normalised


BRF and BRDF and atmospheric corrected BRF from Landsat 5 & 7, (a) Gwydir field
measurement and (b) Lake Frome field measurements. Dashed line represents
perfect agreement

BRDF validation using image overlap areas


An image overlap area is one covered by two adjacent paths imaged within a
relatively short time period. For Landsat, the shortest time that can be
theoretically achieved is 7 days but in practice it will usually be longer. If no
rainfall occurs between the two satellite overpasses and the vegetation does
not change rapidly, the reflectance of the area can be assumed the same.
However, because the area is located in different parts of the two images for
the adjacent paths, that is, the western edge of the image for one path and
the eastern edge of the image for the other path, the retrieved reflectance
may be different if no BRDF correction is applied. Therefore, an overlap area
provides a good opportunity to validate BRDF effects. It also has the
additional practical benefit of evaluating whether physically based methods
can be used to create effective mosaics without empirical adjustment.
For this image-based validation, paths 95 and 96 and row 75 of November,
2006 were selected. Path 95 was obtained on November 13, 2006 and Path
96 was obtained on November 20, 2006. Two small uniform targets were
selected in the image overlap area. For target 1, the centre latitude and
longitude are 20.902o S and 144.761o E. The size is around 16 x 19 km and it
is located on the upper left for Path 95, but on the upper right for Path 96. The
central geographical location for target 2 is 22.238o S and 144.471o E. It is
around 12 x 14 km and located on the lower left for Path 95 and on the lower
right for Path 96. Input data, such as water vapour and aerosol were selected
using nearby stations. Figure 4 shows the results of using different processing
methods for target 1 and 2 respectively. Both reflectance values are the
average of the whole areas so that the comparison will minimise error arising
from pixel mismatch. The results clearly show that using the BRDF algorithm
described in the previous sections the reflectance from both paths 95 and 96

8
are very close. However, if a conventional method (which assumes a
Lambertian surface) without BRDF correction is used, the reflectance from
path 95 is 0.010~0.025 higher than path 96. This results in visible brightening
at the left side of image and darkening at the right side of the image when the
two uncorrected paths are mosaicked together.

0.5 0.5
B9 5 B9 5
B9 6 B9 6
0.4 L9 5 0.4 L9 5
L9 6 L9 6

Reflectance factor
Reflectance factor

0.3 0.3

0.2 0.2

0.1
0.1

a b
0
0
B1 B2 B3 B4 B5 B6
B1 B2 B3 B4 B5 B6
Landsat bands Landsat bands

Figure 4. Reflectance factor products comparison computed from different


processing methods for image overlap areas (a) target 1, (b) target 2. Symbols B95,
B96 are reflectance factor with BRDF correction (normalised to 45o solar angle) for
paths 95 and 96, respectively, whereas L95, L96 is reflectance factor without BRDF
correction for paths 95 and 96

Terrain illumination correction


Most Australian regions are relatively flat. However, on the east coast of
Australia, topography effect creates significant variation in the data. After
careful consideration, path 91, row 85 of November 8, 2008 was used to test
the algorithm. The image central geographical location is 36.017oS and
147.85oE. The image is located in the south of NSW and includes the Snowy
Mountain area of NSW which has the most significant topographic effects in
Australia. The aerosol data are from the local aeronet station and total water
vapour data are from NOAA NECP reanalysis data sets. BRDF parameters
are from the MODIS BRDF product. The image was processed using two
different algorithms: with terrain illumination correction (eq. (6)) and without
terrain correction (Li et al., 2010 and eq. (2)). Fig 5 shows the results.
Comparing figures 5a and 5b, we can clearly see the terrain effect at hillslope
scale has been removed in Fig 5b, when compared with BRDF and
atmospheric correction only (Fig 5a).

Conclusions
An algorithm which uses MODIS BRDF shape functions and a relatively fast
method based on the MODTRAN radiative transfer model can provide
consistent and validated surface reflectance products which are relatively free
of major regional sun, view angle and terrain illumination effects. The coupled
BRDF model can be applied for both flat and rugged surfaces, as well as

9
without empirical adjustments. This investigation shows how, using the
coupled BRDF and atmospheric model developed by the MODIS science
team, good estimates of the Landsat normalised surface reflectance and the
BRDF effect on Landsat images can be achieved. Such an algorithm can be
applied in a standardized form to large numbers of images from a variety of
satellites, to generate a standard time series that covers the continent to
monitor the Australian environment. Such a time series would also be
consistent with images processed in a similar way using a number of sensors
and be comparable with sensors imaging other parts of the world at different
times of the year.

(a) (b)

Figure 5. False color composite images near Snowy Mountain areas of NSW (bands
4, 3, 2): (a) with BRDF and atmospheric correction only, (b) with BRDF and
atmospheric correction plus terrain illumination correction.

Acknowledgement
The authors acknowledge Geoscience Australia for the provision of Landsat
images and DSM data. MODIS BRDF model parameters were downloaded
from the NASA MODIS website. Sunphotometer data were from the NASA
AERONET site and radiosonde data and precipitable water were obtained
from NOAA. The Lake Frome field experiment crew collected and processed
reflectance data at the Lake Frome site and Guy Byrne (CSIRO Land and
Water) provided expertise and support for the field data collection in Gwydir
field measurement and he and Paul Daniels (CSIRO Land and Water)
provided support for processing and analysis. Drs Magnus Wettle and Leo
Lymburner from Geoscience Australia kindly contributed their time to review
the paper.

References
Danaher, T., Wu, X. and Campbell, N., 2001, Bi-directional reflectance
distribution function approaches to radiometric calibration of Landsat ETM+
imagery. Proceedings of IGARSS 2001, Sydney, Australia, July 2001

10
Dymond, J.R., Shepherd, J.D. and Qi, J., 2001, A simple physical model of
vegetation reflectance for standardising optical satellite imagery. Remote
Sensing of Environment, 77: 230-239
Furby, S.L. and Campbell, N., 2001, Calibration images from different dates of
‘like-value’ digital counts. Remote Sensing of Environment, 77: 186-196
Goward, S.N. and Williams, D. L. ,1997, Landsat and Earth systems science:
Development of terrestrial monitoring. Photogrammetric Engineering and
Remote Sensing, 63: 887–900.
Hu, B., Lucht, W. and Strahler, A.,1999, The interrelationship of atmospheric
correction of reflectances and surface BRDF retrieval: A sensitivity study.
IEEE Transactions on Geoscience and Remote Sensing, 37: 724-738.
Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L.,
Iredell, M., Saha, S., White, G, Woollen, J, Zhu, Y., Chelliah, M, Ebisuzaki,
W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J., Leetmaa,
A., Reynolds, R., Jenne, R. and Joseph, D., 1996, The NCEP/NCAR 40-Year
Reanalysis Project. Bulletin of the American Meteorological Society, 77: 437-
471
Li, F., Jupp, D. L. B., Reddy, S., Lymburner, L., Mueller, N.,Tan, P. and Islam,
A., 2010, An evaluation of the use of atmospheric and BRDF correction to
standardize Landsat data. The IEEE Journal of Selected Topics in Applied
Earth Observations and Remote Sensing, in press
Lucht, W., Schaaf, C., and Strahler, A., 2000, An algorithm for the retrieval of
albedo from space using semiempirical BRDF models. IEEE Transactions on
Geoscience and Remote Sensing, 38: 977-998.
Richter, R., Schlapfer, D. and Muller, A., 2006, An automatic atmospheric
correction algorithm for visible/NIR imagery. International Journal of Remote
Sensing, 27: 2077-2085
Schaaf, C., Gao, F., Strahler, A.H., Lucht, W., Li, X., Tsang, T., Strugnell,
N.C., Zhang, X., Jin, Y., Muller, J.–P, Lewis, P., Barnsley, M., Hobson, P.,
Disney, M., Roberts, G., Dunderdale, M., Doll, C., d’Entremont, R.P., Hu, B.,
Liang, S., Privette, J.L. and Roy, D., 2002, First operational BRDF, albedo
and nadir reflectance products from MODIS. Remote Sensing of Environment,
83: 135-148.
Shepherd, J.D. and Dymond, J.R., 2003. Correcting satellite imagery for the
variance of reflectance and illumination with topography. International Journal
of Remote Sensing, 24: 3503-314
Tickle, P. and Gallant, J., 2010. 1 second SRTM derived digital surface model
(DSM) & digital elevation model (DEM), User Guide, version 1.0.2, available
through Geoscience Australia and CSIRO
Vermote, E, Saleous, N. El, Justice, C.O., Kaufman, Y.J., Privette, J.,
Remer, L., Roger, J. and Tanre, D., 1997, Atmospheric correction of visible to
middle-infrared EOS-MODIS data over land surfaces: Background,
operational algorithm and validation. Journal of geophysical Research, 102,
D14: 17131-17141

11

Das könnte Ihnen auch gefallen