Atmospheric and

Radiometric Corrections
for Remote Sensing Data

Rüdiger Gens
 Outline

y radiometric correction
Atmospheric and radiometric corrections

x striping
x (partially) missing lines
x illumination and view angle effects
x sensor calibration
x terrain effects
y atmospheric correction

2
 Striping

y due to non-identical detector response

Atmospheric and radiometric corrections

x detector characteristics
x changes with time / rise of temperature
x failure
y various methods (sometimes used in
combination)
x look up tables (radiometric response measurements
at different brightness levels)
x onboard calibration
x histogram matching (gain and offset) – line pattern

3
 Striping – Landsat TM
Atmospheric and radiometric corrections

Processed by A. Prakash, GI, UAF Processed by A. Prakash, GI, UAF

Striping De-striped

4
 (Partially) missing lines

y errors in
Atmospheric and radiometric corrections

x sampling or scanning equipment

x transmission or recording of image data
x reproduction of the media containing the data

y two methods
x interpolation using data from adjacent scan lines
x interpolation data at the same scan line from
different spectral bands

5
 Partially missing lines - Example
Atmospheric and radiometric corrections

Source: CCRS Remote Sensing Tutorial

6
 Sun angle correction

y position of the sun

Atmospheric and radiometric corrections

relative to the earth

changes depending on
time of the day and
the day of the year
y in the northern
hemisphere the solar
elevation angle is
smaller in winter than
in summer Adapted from Lillesand and Kiefer

7
 Sun angle correction

y an absolute correction involves dividing the

Atmospheric and radiometric corrections

DN-value in the image data by the sine of the

solar elevation angle
y size of the angle is given in the header of the
image data

DN
DNcorr =
sin α

8
 Sun Illumination

y position of sun
Atmospheric and radiometric corrections

x sun elevation (sun-angle)

x sun - earth distance
y correction elevation
x normalization
x division of each pixel
value by the sine of solar
elevation angle for
particular time and
location per spectral
band
Adapted from Lillesand and Kiefer
y correction distance
x sun irradiance decreases
with square of distance
x normalization
9
 Sensor calibration

y necessary to generate absolute data on

Atmospheric and radiometric corrections

physical properties
x reflectance
x temperature
x emissivity
x backscatter
y values provided by data provider / agency

10
 Terrain effects

y cause differential solar illumination

Atmospheric and radiometric corrections

x some slopes receive more sunlight than others

y magnitude of reflected radiance reaching the

sensor
x topographic slope and aspect
x bidirectional reflectance distribution function (BRDF)

11
 Terrain correction

y Minnaert correction
Atmospheric and radiometric corrections

x first order correction for terrain illumination effects

k −1
L n = L ⋅ cos( e) ⋅ cos(i)k

Ln – normalized radiance
L – measured radiance
e – slope (derived from DEM)
i – incidence angle of solar radiation
k – Minnaert constant (estimated for each image)

12
 Terrain correction

y shaded relief model (SRM)

Atmospheric and radiometric corrections

x requires digital elevation model

x generated with constant albedo (brightness
dependent solely on topographic effects)
x ratio of image and SRM yields spectral radiance of
ground cover (noisy)
x alternative

DNcorr = m ⋅ (DN − SRMDN ) + a

13
 Why do atmospheric correction?

y physical relation of radiance to surface

Atmospheric and radiometric corrections

property
x atmospheric component needs to be removed
y multispectral data for visual analysis
x scattering increases inversely with wavelength
y image ratios
x leads to biased estimate
y time difference between image acquisition and
ground truth measurements

14
 Atmosphere and radiation

y relationship between radiance received at the sensor

Atmospheric and radiometric corrections

(above atmosphere) and radiance leaving the ground

L s = H ⋅ ρ ⋅ T + Lp
Ls – at sensor radiance
H – total downwelling radiance
ρ – reflectance of target
T – atmospheric transmittance
Lp – atmospheric path radiance (wavelength dependent)

15
 Atmospheric correction methods

y image-based methods
Atmospheric and radiometric corrections

x histogram minimum method

x regression method
y radiative transfer models
y empirical line method

16
 Histogram minimum method
Ls = H ⋅ ρ ⋅ T + Lp
Atmospheric and radiometric corrections

y histograms of pixel values in all bands

y pixel values of low reflectance areas near zero
x exposures of dark colored rocks
x deep shadows
x clear water
y lowest pixel values in visible and near-infrared
are approximation to atmospheric path
radiance
y minimum values subtracted from image
17
 Regression method
Ls = H ⋅ ρ ⋅ T + Lp
Atmospheric and radiometric corrections

y applicable to dark pixel areas

y near-infrared pixel values are plotted against
values in other bands
y least square line fit using standard regression
methods
y resulting offset is approximation for
atmospheric path radiance
y offset subtracted from image

18
 Radiative transfer models

y limited by the need to supply data about

Atmospheric and radiometric corrections

atmospheric conditions at time of acquisition

y mostly used with "standard atmospheres"
y available numerical models
x LOWTRAN 7
x MODTRAN 4
x ATREM
x ATCOR
x 6S (Second Simulation of the Satellite Signal in the
Solar Spectrum)

19
 ATCOR

y originally developed DLR (Germany)

Atmospheric and radiometric corrections

y different versions
x satellite imagery – ATCOR 2 (flat terrain), ATCOR 3
(rugged terrain)
x airborne imagery – ATCOR 4 (flat and rugged)
y various versions commercially available
x standalone version in IDL
x ERDAS Imagine
x PCI Geomatics

20
 Surface reflectance R
Empirical line method

y selection of one dark

Atmospheric and radiometric corrections

and one bright target

R = s (L-a)
y ground reflectance
bright measurement
target
x field radiometer
y sensor radiance
dark computed from image
target
y slope s = cos (α) and
α
intercept a of line joining
sensor radiance L
atmospheric two targets
radiance a

21
 Haze

y due to Rayleigh scattering

Atmospheric and radiometric corrections

x particles size responsible for effect smaller than the

radiation's wavelength (e.g. oxygen and nitrogen)
y haze has an additive effect resulting in higher
DN values
y decreases the general contrast of the image
y effect is wavelength dependent
x more pronounced in shorter wavelengths and
negligible in the NIR

22
 Haze – Example Indonesia
Atmospheric and radiometric corrections

Source: C. Pohl, ITC Source: C. Pohl, ITC

Hazy Corrected

23
 Atmospheric and radiometric corrections

Questions

24