Atmospheric and Radiometric Corrections for Remote Sensing Data
Rüdiger Gens
2
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Outline
radiometric correctionstriping(partially) missing lines illumination and view angle effectssensor calibrationterrain effects
atmospheric correction
3
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Striping
due to non-identical detector responsedetector characteristicschanges with time / rise of temperaturefailure
various methods (sometimes used in combination)
look up tables (radiometric response measurements at different brightness levels)onboard calibrationhistogram matching (gain and offset) – line pattern
4
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Striping – Landsat TM
De-stripedStripingProcessed by A. Prakash, GI, UAF Processed by A. Prakash, GI, UAF
5
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
(Partially) missing lines
errors in sampling or scanning equipmenttransmission or recording of image datareproduction of the media containing the data
two methodsinterpolation using data from adjacent scan linesinterpolation data at the same scan line from different spectral bands
6
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Partially missing lines - Example
Source: CCRS Remote Sensing Tutorial
7
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Sun angle correction
position of the sun relative to the earth changes depending on time of the day and the day of the yearin the northern hemisphere the solar elevation angle is smaller in winter than in summer Adapted from Lillesand and Kiefer
8
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Sun angle correction
an absolute correction involves dividing the DN-value in the image data by the sine of the solar elevation anglesize of the angle is given in the header of the image data
α=
sinDNDNcorr
9
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Sun Illumination
position of sunsun elevation (sun-angle)sun - earth distance
correction elevationnormalizationdivision of each pixel value by the sine of solar elevation angle for particular time and location per spectral band
correction distancesun irradiance decreases with square of distancenormalization
Adapted from Lillesand and Kiefer
10
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Sensor calibration
necessary to generate absolute data on physical properties
reflectancetemperatureemissivitybackscatter
values provided by data provider / agency
11
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Terrain effects
cause differential solar illuminationsome slopes receive more sunlight than others
magnitude of reflected radiance reaching the sensor
topographic slope and aspectbidirectional reflectance distribution function (BRDF)
12
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Terrain correction
Minnaert correctionfirst order correction for terrain illumination effects
Ln – normalized radianceL – measured radiancee – slope (derived from DEM)i – incidence angle of solar radiationk – Minnaert constant (estimated for each image)
kkn )icos()ecos(LL ⋅⋅= −1
13
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Terrain correction
shaded relief model (SRM)requires digital elevation modelgenerated with constant albedo (brightness dependent solely on topographic effects)ratio of image and SRM yields spectral radiance of ground cover (noisy)alternative
( ) aSRMDNmDN DNcorr +−⋅=
14
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Why do atmospheric correction?
physical relation of radiance to surface property
atmospheric component needs to be removedmultispectral data for visual analysis
scattering increases inversely with wavelength image ratios
leads to biased estimatetime difference between image acquisition and ground truth measurements
15
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Atmosphere and radiation
Ls – at sensor radianceH – total downwelling radianceρ – reflectance of targetT – atmospheric transmittanceLp – atmospheric path radiance (wavelength dependent)
ps LTHL +⋅ρ⋅=
relationship between radiance received at the sensor (above atmosphere) and radiance leaving the ground
16
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Atmospheric correction methods
image-based methodshistogram minimum methodregression method
radiative transfer modelsempirical line method
17
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Histogram minimum method
histograms of pixel values in all bandspixel values of low reflectance areas near zero
exposures of dark colored rocksdeep shadowsclear water
lowest pixel values in visible and near-infrared are approximation to atmospheric path radianceminimum values subtracted from image
ps LTHL +⋅⋅= ρ
18
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Regression method
applicable to dark pixel areasnear-infrared pixel values are plotted against values in other bandsleast square line fit using standard regression methodsresulting offset is approximation for atmospheric path radianceoffset subtracted from image
ps LTHL +⋅⋅= ρ
19
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Radiative transfer models
limited by the need to supply data about atmospheric conditions at time of acquisitionmostly used with "standard atmospheres"available numerical models
LOWTRAN 7MODTRAN 4ATREMATCOR6S (Second Simulation of the Satellite Signal in the Solar Spectrum)
20
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
ATCOR
originally developed DLR (Germany)different versions
satellite imagery – ATCOR 2 (flat terrain), ATCOR 3 (rugged terrain)airborne imagery – ATCOR 4 (flat and rugged)
various versions commercially availablestandalone version in IDLERDAS ImaginePCI Geomatics
21
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Empirical line method
selection of one dark and one bright targetground reflectance measurement
field radiometer
sensor radiance computed from imageslope s = cos (α) and intercept a of line joining two targetssensor radiance L
Sur
face
refle
ctan
ce R
R = s (L-a)
atmosphericradiance a
brighttarget
darktarget
α
22
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Haze
due to Rayleigh scatteringparticles size responsible for effect smaller than the radiation's wavelength (e.g. oxygen and nitrogen)
haze has an additive effect resulting in higher DN valuesdecreases the general contrast of the imageeffect is wavelength dependent
more pronounced in shorter wavelengths and negligible in the NIR
23
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
CorrectedHazy
Haze – Example Indonesia
Source: C. Pohl, ITC Source: C. Pohl, ITC
24
Atm
osph
eric
and
radi
omet
ric c
orre
ctio
ns
Questions