258
ß = 2.26 (У) - - 73
ß - Angstrom turbidity coefficient
(4)
The Angstrom attenuation coefficient (/?) then
was converted to an aerosol optical depth (r’ a ).
r’ a = /0X“ 01 (5)
The Angstrom wavelength exponent (a) which is a
function of the size distribution of the aerosols
was estimated through an iterative procedure
(Holben et al. (1989b) using an initial estimate
for the aerosol optical depth.
Comparisons of the observer visibility data
versus the aerosol optical depth using the Sun
photometer derived data (n=22, May 28 - June 28,
1990) for Bamako, Mali (Holben et al. 1989b) ,
yielded a correlation coefficient (r) of 0.82.
Hence, there was an indication of a similar trend
between surface visibility observations and
spectral aerosol optical depth -
3.3 Surface Anisotropy
The estimation of a surface, solar albedo frcm
satellite derived radiance (Equation 2) typically
assumes an isotropic reflecting surface.
However, there are several illumination and
viewing geometry conditions where a directional
derived reflectance may misrepresent a
hemispherical, reflectance by over 50% (Kimes et
al. 1985). On the other hand, the assumption of
isotropy for visible and near-IR wavelengths is
adequate for many surface covers between the
solar zenith angles 20° to 40° (Kimes and Sellers
1985). For many of the scenes processed, the Sun
zenith angles were within this illumination range
of 20° to 40°. The highest Sun zenith angle is
for the Oct. 25, 1985 AVHRR scene with an angle
of 50.8° for Tambacounda and 51.3 % for Podor.
Kimes and Sellers (1985) indicates an anisotropic
related error of 5% visible and 7% near-IR for
steppe-grassland with a 50° Sun zenith angle.
Steppe-grassland is a common surface cover in the
study site. Based on these results, the
assumption of surface reflectance isotropy was
made for all the scenes processed, even though a
larger hemispheric reflectance error ( 5-10%)
associated with scenes having large Sun zenith
angles is possible.
3.4 AVHRR Narrow Band to Solar Band Reflectance
Conversion
A model by Brest and Goward (1987) for
converting Landsat Multispectral (MSS) derived
reflectance to solar albedo was further developed
and modified for converting the visible and near-
IR band 1 and 2 derived reflectance to a solar
albedo.
yOsolar = Pvis *k v is + Pnir*knir + Psmir*ksmir (6)
psolar - solar albedo
/°spec ~ derived surface reflectance for the
spectral regions (visible
(0.38-0.72 /mu'), near-IR
(0.72-1.30 /zm) and shortwave middle-
IR (1.30-3.0 /zni)
k — relative proportion of surface, solar
irradiance by spectral region (see/? S pec)
For a vegetative canopy there are four major
spectral regions in the solar region with similar
optical properties (Gausman 1985). The four
spectral regions are ultraviolet (0.30-0.38 /zm),
visible (0.38-0.72 /zm), near-IR (0.72-1.30 /zm)
and shortwave middle-Ш (1.30-3.0 /мп) . In the
visible region, plant pigments chlorophyll,
carotenes and xanthophyll absorb much of the
solar radiation. Leaf absorption is the lowest
for green light (0.45 /zm) with an absorptance
approximately 5% units less than for blue or red
light. The near-IR radiation for plants is
characterized by a low absorption and high
reflectance and transmittance that is reported to
be from internal leaf structure having
substantial intercellular refractive
discontinuities (Gausman 1985). The shortwave
middle-IR (1.30-3.0 /zm) is characterized by
liquid water absorption peaking at 1.45 /zm and
1.94 /zm and also is affected by leaf
intercellular refractive discontinuities (Gausman
1985). In the ultraviolet region there is no
leaf transmittance with approximately 9%
reflectance and 91% absorptance.
The percent of the solar radiation sensed by
the AVHRR bands 1 & 2 bandpass for NOAA-7 and
NOAA-9 at the top and the bottom of the
atmosphere is given in Table 1. To estimate the
atmosphere radiative transfer, a mid—latitude
data set from Dave’ (1978) was evaluated that
modeled an atmosphere with gaseous absorption and
a low aerosol loading. The small proportion of
solar radiation that is represented by the AVHRR
(50% and less), indicates a potential problem for
estimating a broad band, solar albedo.
Analysis of high spectral resolution field
data by Toll (1989) indicates AVHRR band 1 and
band 2 reflectance may be used to estimate both a
total visible (0.38-0.72 /zm) and a total near-IR
(0.73-1.30 /zm) reflectance because of the high
intra-region correlation. The combined visible
and near-IR regions represents approximately 85%
of the total solar radiation at the surface with
the remaining 15% in the shortwave middle-IR
(10%) and ultraviolet (5%) regions. However, a
linear transformation is required when estimating
a total visible and near-IR reflectance (Toll
1989). Errors to 20% may be obtained when using
an AVHRR visible (e.g., 0.57-0.69 /zm) or near-IR
(0.71-0.98 /zm) reflectance to directly estimate a
total visible (0.37-0.72 /zm) or near-IR (0.73-
1.30 /zm) reflectance without a linear
transformation. The derived visible reflectance
was used to represent an ultraviolet reflectance
with only a small reduction in accuracy
accounting for less than 5% of the surface solar
radiation for ultraviolet light. Toll (1989)
indicates a difference of less than 1% when using
a visible derived reflectance to represent an
Tab! e 1. Proportion of M3AA AVHRR band pass to surf ace and
exoatmospheric irradiance by spectral region (i.e., visible and
near-IR) and total solar. Dave’ (1978) Model 3 of gaseous
absorption with light aerosol density for mid-latitude
continental at 30° Sun zenith.
Percent band-pass in visible (0.38-0.72 /An) for
for band-1, in near-IR (0.72-1.30) for band 2 and
total
solar (0.33-3.0 /An)
for bands 1&2
combined.
% Sensed Vis or
Near-Ш
% Sensed in
Total Solar
NUAA-# Channel (¿an)
Surface
Bxoatm.
Surface
Exoatm.
-7
1(.571-,686)
32.2
31.6
15.5
13.6
2(.713-,986)
50.5
53.0
50.5
18.5
1&2
34.0
26.1
-0
1(.570-.699)
34.1
33.6
16.4
14.4
2(.714-.983)
53.1
54.3
53.1
19.5
1&2
35.9
27.1
