Lectures Remote Sensing ATMOSPHERIC CORRECTION dr.ir. Jan Clevers Centre of Geo-Information Environmental Sciences Wageningen UR
Atmospheric Correction of Optical RS Data Background When needed? Model understanding Model implementation Validation
Background: Atmospheric Composition RS no clouds of importance are: absorption: mainly due to water vapour scattering by: air molecules (N 2, O 2 ) λ >> D Rayleigh scattering: λ -4 aerosols (dust) Mie scattering: independent of λ λ D rain droplets neglectable with RS λ << D
Atmospheric Correction Methods Step 0: When needed yes/no? monotemporal multitemporal classification parameter estimation monitoring / change detection Step 1: Model understanding simplification: abstraction of reality assumptions: limitations, possibilities model choice: variables, equations planetary reflectance L&K sections 1.4 and 7.2 Richards other models Step 2: Model implementation how do we get good estimates of variables and parameters? Step 3: Results: validation
Step 0 Correction: No at classification: little sensitivity for linear transformations topology remains intact little need for atmospheric correction TM DN DP 1 77 2 28 3 22 4 13 5 5 7 1 DN reflectance (%) 170 80 TM band 4 soil line 13 0 0 22 TM band 3 30 90 reflectance (%) DN
Correction: Yes In case of parameter estimation monitoring atm. correction important Processes monitoring (dynamics)
Step 1: Absence of an atmosphere Planetary reflectance: the reflectance measured at the top of the atmosphere (without atmospheric correction). The planetary reflectance (r p ) equals the measured radiant exitance (W.m -2 ) at the sensor divided by the irradiance (W.m -2 ) at the top of the atmosphere, originating from the sun. sun sensor L s E NO-ATM E DIF E DIR object atmosphere
Planetary Reflectance So: r p = π L s / E NO-ATM L s can be calculated through internal sensor calibration parameters directly from the measured digital number (DN): L s = A 0 + A 1 DN E NO-ATM = E EXO e cosθ λ E NO-ATM = incoming radiation at the top of the atmosphere in Wm -2 E EXO = exo-atmospheric irradiance of the sun in Wm -2 µm -1 (for average earth-sun distance) e = eccentricity correction factor of the earth s orbit θ = solar zenith angle λ = width of the wavelength band in µm
planetary reflectance TM Reflectance Tunesia - April 0.6 0.5 0.4 0.3 0.2 TM band 1 TM band 7 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 ground reflectance
planetary reflectance 0.6 TM Reflectance Tunesia - December 0.5 0.4 0.3 0.2 TM band 1 TM band 7 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 ground reflectance
Presence of an Atmosphere Lillesand & Kiefer, section 1.4 L tot = ρ E/π T + L p (1.9) L tot = L s ρ = r E = E TOT T = T TOT,ϕ So: L s = r T TOT,ϕ E TOT /π + L a This is the starting point for equation (5) in the syllabus
Variety of atmospheric correction techniques Simple approaches Radiative Transfer modelling Histogram matching (assuming the same surface reflectance histograms) Darkest pixel subtraction (assuming negligible surface reflectance) DN cor = DN meas - DN DP Empirical line correction (using invariant bright and dark objects) ρ = a + b DN
Atmospheric Model According to RICHARDS irradiation from sun path radiance component 1 radiance to sensor path radiance component 2 EDIF sky irradiance component 1 EDIR θ φ sky irradiance component 2 atmosphere pixel (ground resolution element) neighbouring pixel
Equations Richards L s = r T TOT,ϕ E/π + L a E = E DIR + E DIF L s = r T TOT,ϕ (E DIR + E DIF )/π + L a E DIR = E NO-ATM T TOT,θ E NO-ATM = E 0 cosθ λ So: E DIR = E 0 T TOT,θ cosθ λ
Step 2: Model Implementation (Richards) Calculation of direct irradiation (E DIR ) Calculation of diffuse irradiation (E DIF ) Calculation of atmospheric transmittances (T TOT ) (for Rayleigh, Mie and water vapour) Calculation of path radiance (L a ) End result: r = π. (L s - L a ) / [T TOT,ϕ. (E DIR + E DIF )]
Model Extensions (Richards) Extension 1: Multiple reflections between atmosphere and ground Extension 2: Diffuse upward transmittance through atmosphere End result: r = π. (L s - L a ) / [T 1. T 2. E NO-ATM + r a. π. (L s - L a ) ] This is the equation that will be used in the exercise.
Model Verhoef (used for validation) E s E - E + E o
Model Verhoef -2- τ ss E NO-ATM = E 0 s cos θ s τ oo πl tot Ozone layer ρ so τ τ τ sd do τ ss oo Troposhere ρ dd E sun E sky E upw πl obj E DIR E DIF r dd Earth r sd r do rso
Step 3: Results: Validation Reflectance (%) TM band 3, 16 June 1986 100 80 60 40 Verhoef Richards 20 0 50 100 150 200 250 Digital number
Difference Richards - Verhoef Richards Verhoef atmosphere 1 layer 2 layers O 3 absorption no yes H 2 O absorption yes no adjacency effects no yes Mie transmittance from visibility darkest pixel method
Some examples of other models MODTRAN: MODerate resolution TRANSsmittance http://www2.bc.edu/~sullivab/soft/modtran4.html 6S: Second Simulation of Satellite Signal in the Solar Spectrum ACORN: Atmospheric CORrection Now (uses MODTRAN 4 radiative transfer modeling) http://www.aigllc.com/acorn/intro.asp ATCOR: ATmospheric CORrection, topographic correction and haze removal of hyperspectral satellite images (add-on module to ERDAS IMAGINE ) http://www.geosystems.de/atcor/index.html