IEEE Geoscience and Remote Sensing Magazine - June 2019 - 33

aerosol vertical profile, AOD can be related to the meteorological quantity called visible range or visibility. In contrast,
gases contain very sharp absorption features (Figure 2),
which can be used for sensor wavelength calibration and
concentration estimation. Estimates of peak absorption, as
a function of wavelength for various species, are shown
in Table 1.
In general, water vapor plays a major role in atmospheric
transmission loss, as is evident from the lower-right VNIR-
SWIR H2O plot in Figure 2. In the "Physical-Based Methods"
section, we will show that the 0.940- and 1.130-nm watervapor absorption features are particularly interesting because
they tend to be used to obtain column water-vapor amounts.
Molecular scattering (i.e., Rayleigh scattering) also plays a significant role in hampering transmission; it varies with a 1/m 4
dependence (i.e., shorter wavelengths scatter more). We can
see this in Figure 2 as a characteristic sloped shape in reduced
transmission, from approximately 0.85 at 1 nm down to approxiamately 0.45 at 0.4 nm based on a rural aerosol model,
visibility of 23 km, and sensor altitude of 50 km.

particularly vegetation. An effective compensation strategy is to retrieve surface liquid water and atmospheric
water-vapor measurements at the same time, as is done in
several atmospheric methods [11], [12]. Depending on the
specifics of the sensor view and scene content, an assumption of fixed water-vapor absorption may or may not be
sufficient for accurate atmospheric compensation.
ATMOSPHERIC RADIATIVE TRANSFER
SENSOR-REACHING RADIANCE
Before discussing physics-based compensation methods,
we describe the various ways in which solar radiation
interacts with the target, background, and atmosphere.
Thus, in this section we develop a forward sensor-reaching radiance model that accounts for the various paths
solar radiation can take on its way to an aerial sensor. We
then discuss some assumptions and simplifications that
make the reflectance solution to such an equation more
tractable for atmospheric compensation.
Early research in the area of first-principles RT modeling
can be attributed to the work of Chandrasekhar [14] in the
late 1950s. Since then, work has progressed, most notably

0.94
1.13

Transmittance

SPATIAL VARIABILITY OF WATER VAPOR
Water-vapor column amount can vary by an order of magnitude, depending on the geographic
location, season, and particular day.
Therefore, accurate atmospheric
1
1
compensation of hyperspectral data
with a first-principles method re0.9
0.9
O3
quires water-vapor retrieval. Because
CO
0.8
0.8
O2
the apparent water-vapor amount
0.7
0.7
can vary within a scene, these methCH4
ods typically are designed to retrieve
0.6
0.6
water vapor on a pixel-by-pixel basis.
0.5
0.5
Figure 3 shows a water-vapor map
O2
retrieved by the atmospheric comO2
0.4
0.4
pensation code FLAASH (see the
0.5
1
1.5
2
2.5
0.5
1
1.5
2
2.5
"Physics-Based Methods" section)
1
1
from an Airborne Visible/Infrared
H2O
Imaging Spectrometer (AVIRIS) im0.9
0.8
age [9] collected over Woodside,
0.8
California, The derived water-vapor
0.6
column amount is 0.72 cm with a
0.7
standard deviation of 0.07 cm, or
0.4
0.6
lower than 10% of the mean [10].
CO2
0.5
Several factors may contribute
0.2
to the apparent within-scene water
0.4
0
variation besides an actual variation in concentration, including
0.5
1
1.5
2
2.5
0.5
1
1.5
2
2.5
Wavelength (µm)
Wavelength (µm)
topography (i.e., variation in elevation) and changes in viewing angle
across the image, both of which af- FIGURE 2. The atmospheric transmission attenuation in the VNIR-SWIR region due to
fect the length of the upward light extinction from gas molecules, aerosols, and water as derived from Moderate Resolution
path through the atmosphere. An- Atmospheric Transmission [6]-[8]. The simulation parameters included a rural aerosol model,
other potential source of variation visibility of 23 km, sensor altitude of 50 km, and nadir looking with a spectral resolution of
is interference from liquid water 0.003 µm. The concentrations were 0.4 atm-cm for O 3, 0.15 ppmv for CO, 200 ppmv for CO 2,
on or within the surface material, 1.8 ppmv for CH 4, and 3,500 atm-cm for H2O. ppmv: parts per million by volume.
JUNE 2019

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

IEEE Geoscience and Remote Sensing Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2019