(1986) notes the resiliency of the land with an
increasing net primary productivity for much of
the Sahelian region following a period in 1985 of
increased rainfall. For this effort, a dry
northerly site near Podor (16c>30’ N 14024’ W) and
a semi-dry/wet southerly site near Tambacounda
(13047’N 13°40’ W) were selected for analyses.
Image pixel test sites were selected 10-20 km
away from the city center for reducing urban
related effects. An illustration of the study
site region is in Figure 1 with mean annual
isohyets included. 3
Figure 1. Study site location with mean
annual isohyets (Compiled :AGPE/FA0).
3. APPROACH AND ANALYSES FOR PARAMETER DERIVATION
3.1 Shortwave Radiometric Calibration
There is no on-board calibrator for the AVHRR
shortwave visible and near-IR bands. Instead a
vicarious calibration technique is typically used
to determine post-launch calibration changes.
Frouin and Gautier (1987) estimates a 15% post
launch decrease in radiometric gain for the
shortwave NOAA—7 AVHRR bands. Holben et al.
(1989a) through analysis of Saharan derived
reflectivities indicates a time varying gain
change to 20% for both NOAA—'7 and NOAA—9
shortwave bands. The Holben et al. (1989a)
radiometric adjustments were used to correct the
AVHRR spectral data.
The NOAA AVHRR calibrations to exoatmospheric
reflectance were converted to radiance for use in
atmospheric corrections for deriving a surface
reflectance - The exoatmospheric derived
calibrations were converted from the NOAA
provided (Thekaekara 1973) solar irradiance data
to that of the more widely recommended data of
Neckel and Labs (1984) following recommendations
by Price (1987) . For the AVHRR visible band
there is a change in reflectance of 5.9% and for
the near-IR band there is a change of 2.3%. Pre
flight versus post-flight calibrations indicates
a time varying change, worsening with launch for
both NOAA-7 and NOAA-9 AVHRR. The coefficient of
variation of the calibration difference is 31.4%
for the visible reflectance, 29.6 for the near-IR
reflectance, and 29.6% for the NDVI.
3.2 Shortwave Atmospheric Correction
The relationship between a Lambertian surface
reflectance (Psur) and "the upward spectral
radiance (L sa t) measured by the satellite
(Chandrasekhar 1960) is given next with a
spectral dependence (A) implied.
Lsat = Lq + (FsurA") * (T*/?sur/ (l - s*Psur)) (1)
Lsat ~ satellite calibrated radiance
(W-m~2 si—1)
Lq - atmosphere to satellite path radiance
(W-m~2 sr~l)
Fsur - surface, solar irradiance (W-m _ 2)
T - surface to satellite total transmittance
_ (direct + diffuse)
s - atmosphere to surface counter-reflectance
Psur - isotropic surface reflectance
The measured satellite radiance for bands 1 and 2
were converted to reflectance using the
relationship given next.
Psur = (Lsat ~ Lo)/(T’*F 0 (l/7r)+s(Lsat~Lc>)) (2)
T’ - Total downwelling times upwelling
atmosphere transmittance (Ahmad
and Fraser 1982)
The Lsat was derived from the NOAA AVHRR
calibrated radiance. The atmospheric correction
was developed using the radiative transfer model
described by Ahmad and Fraser (1982) . The model
incorporates a multiple scatter radiative
transfer procedure after Dave (1972) with the
addition of atmospheric polarization related
effects. The model assumes a spherical aerosol
for solution using Mie theory. In addition, the
model assumes a horizontally homogeneous
atmosphere bounded by a Lambertian reflecting
surface. The atmosphere is assumed cloud free.
The size distribution of the aerosols was
represented by a power law distribution. The
index of refraction is assumed to be m = 1.54-
0.003i. The real number of 1.54 is considered
typical of the sub—Saharan (e. g., Carlson and
Caverly 1977) and the imaginary part at 0.003 is
less known but was selected by inspecting results
by Patterson et al. (1977) and a personal
communication from Holben (1988) .
Gaseous absorption for input to the radiative
transfer model of Ahmad and Fraser (1982) were
estimated using the opt i cad. properties
(McClatchey et al. 1971) specified in the
tropical model (15oN) from Kneizys et al. (1983) .
The model includes estimates for pressure,
temperature, aerosols, water vapor and ozone.
The pressure, temperature and water vapor
vertical distributions were normalized for
conditions at the time of the satellite overpass
for Tambacounda and Podor by using the near
hourly surface meteorological data frem NCDC
(i.e., ambient air temperature, surface
atmospheric pressure, and dew point temperature).
Holben and Eck (1989) indicates that atmospheric