Atmospheric Correction Algorithm - PowerPoint Presentation

Atmospheric Correction Algorithm for the GOCI Jae Hyun Ahn * Joo-Hyung Ryu * Young Jae Park* Yu-Hwan Ahn * Im Sang Oh** Korea Ocean Research & Development Institute Seoul National University

I n d e x _ Introduction _ Atmospheric CorrectionAtmospheric Algorithms of the GOCI> Standard NASA Algorithm > SGCA > SSMM Process of Atmospheric Correction _Standard NASA AlgorithmSGCASSMMResult & Validation _ResultValidationConclusion _ Ocean Color

1. Introduction _ Atmospheric Correction M(λ) *L TOA (λ)*Rrs(λ) Chl SSCDOM … Radiometric Calibration Atmospheric Correction L2 algorithms L TOA (555 nm ) Rrs (555 nm ) AtmosphericCorrection *L : radiance * Rrs : remote sensing reflectance

1. Introduction _ Atmospheric Correction Clear water / thin aerosol case * Lr: Radiance of molecular scattering La : Radiance of aerosol scattring*Lw : Radiance of Ocean Case 1 water : LW is 1~7% of LTOA

1. Introduction _ Atmospheric Correction Issue : GOCI has longer optical path than the polar orbit satellite (MODIS : 0˚ < Satellite zenith angle < 40˚) 26˚ < Satellite zenith angle < 55˚ Observation area Earth GOCI equator

Introduction _ 3 atmospheric Algorithms of the GOCIStandard NASA algorithmA classical standard atmospheric correction algorithm Developed by M.Wang & H.R.GordonAerosol selection, turbid-water iterative method, diffuse transmittance models are updated by J.H.AhnSSMM (Spectral Shape Matching Method)Developed by Y.H.Ahn & P.ShanmugamUsing reference siteAerosol models updated by J.H.Ahn SGCA (Sun-Glint Correction Algorithm) Developed by HYGEOSRemoving sun-glint & atmospheric signalPolynomial fitting algorithm (ocean color & atmospheric model )

2. Process of Atmospheric Correction _ Geometric Corrected TOA Radiance Image L TOA(λ) Raw Image Reflectance of TOA Image ρ(λ)=ρ‘ (λ) + ρR ( λ) Reflectance of Ocean + Aerosol Image ρ ‘ ( λ ) = T d (λ )ρW(λ) + ρA(λ) + ρRA(λ) Reflectance of Ocean Image ρ W(λ) Level 2 ProductChl, SS, CDOM, Kd490, … Radiometric Calibration & Geometric Correction Downward Solar Irradiance Normalization  Longitude, Latitude, Time, SZA, VZA, AZA Remove Rayleigh & Sun-glint Reflectance & Mask  Radiative Transfer Equation, Cox&Munk Model Remove Aerosol Reflectance Radiative Transfer Equation, Aerosol Model Underwater Algorithm Reflectance of Ocean Image Rrs ( λ ) Atmospheric Correction Standard NASA Algorithm SSMM SGCA

2. Process of Atmospheric Correction _ Step 1. Downward Solar Irradiance Normalization Downward Solar Irradiance Normalization L TOA ( λ ) cos ( θ S ) * θ S : solar zenith angleF0( λ) : Extraterrestrial spectral irradianceρTOA (λ)

0 1 2 3 4 5 7 6 9 8 12 13 14 15 11 10 2. Process of Atmospheric Correction _ Slot Correction of Solar Irradiance Normalization cos ( θ S ) Step 1. Downward Solar Irradiance Normalization

2. Process of Atmospheric Correction _ Step 2. Remove Rayleigh Signal ρ TOA (443 nm ) ρ R (443 nm ) ρ ‘ (443 nm)

2. Process of Atmospheric Correction _ Remove direct & sun-glinted Rayleigh reflectance Computed by radiative transfer equation Integrate with GOCI bands’ spectral response Using pre-computed LUT Wind speed : 0~16 m/s Step 3. Remove Rayleigh & Sun-glint Reflectance Scattering off a rough sea surface Molecular scattering

M 2. Process of Atmospheric Correction _ Step 3. Land & Cloud Masking Using threshold of Band8 (865nm) Masking 5x5 around the above threshold pixel M M M M M M M M M M M M M M M M M M M M M

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal ρ ‘ (555 nm ) ρ A (555 nm )+ ρ RA (555 nm )ρW (555nm )

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal Standard NASA algorithm Basic Assumption : ρW(NIR) = 0 (GOCI’s NIR Band : 745nm, 865nm) Atmospheric Correction Select 2 Aerosol Type Multiple Scattering to Single Scattering for all Aerosol Types Get Two Aerosol Models (model1/model2) ε model1 ( B7, B8 ) < ε ave(B7, B8 ) < εmodel2(B7, B8) Look-up Table from RTE (6S) Calculate Multiple Scattering of Specific Aerosol type Get ε (λ, B8) for all band Calculate Single Scattering of 2 Specific Aerosol type Calculate Single Scattering Reflectance for all Band  ρ as model ( λ ) 2 Aerosol Models sza / vza / aza ρ as model1 ( λ ) ρ as model2 ( λ ) Get ρ a ( λ ) + ρ ra ( λ ) and t( λ ) of 2 models Interpolate ρ a ( λ ) + ρ ra ( λ )and t(λ)of 2 models Calculate Rayleigh Scattering

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal Standard NASA algorithm Aerosol model selection (Modified) Select 2 Aerosol Type Multiple Scattering to Single Scattering for all Aerosol Types Get Two Aerosol Models (model1/model2) ε model1 ( B7, B8 ) < ε ave (B7, B8) < ε model2(B7, B8) Average all aerosol models’ ε(B7, B8)Select 4 aerosol models Average 4 aerosol models’ ε(B7, B8) Select 2 aerosol models Get weight of 2 aerosol models

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal Aerosol models Maritime (RH 50%, RH 80%, RH 99%) Urban (RH 50%, RH 80%, RH 99%) Continental (RH 50%, RH 80% RH 99%) Band 8 signal(aerosol signal) Aerosol model selection result Aerosol removed signal (pure ocean signal : ρ w (443)) East sea East sea East sea East sea

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Reflectance SSMM (Spectral Shape Matching Method) Assumption : ρW(NIR) = 0 (GOCI’s NIR Band : 745nm, 865nm) Assumption : ρaerosol_model_1(λ ) + ρ aerosol_model_2(λ) = 0 Use reference site’s spectrum shape Atmospheric Correction LUT Reflectance of Specific Aerosol type 2 Aerosol Models sza / vza / aza ρ a ( λ ) + ρra( λ)and t(λ) Calculate Rayleigh Scattering Reference site Get Aerosol reflectance Get Two Aerosol Models & mixing ratio from LUT

ρ TOA (NIR)= ρ r (NIR) + ρa(NIR) + ρra(NIR) + t(NIR) ρf(NIR) + t(NIR) ρw(NIR) ρr (λ ) calculated by RTEρa ( λ ) + ρ ra ( λ ) calculated by LUTt(NIR)  calculated by LUT + RTEρf(NIR) calculated by Cox&Munk’s Eqρ w (λ) chl , ss Atmospheric Correction Underwater Algorithm CHL, TSM  ρw (NIR) Ocean Color Model ρ w ( λ ), chl  corrected ρ w ( λ ) BRDF 2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Reflectance Iterative Method of NASA Standard Algorithm & SSMM Turbid water : ρ W (NIR) ≠0

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal Iterative Method of NASA Standard Algorithm & SSMM Rrs(NIR) = f/Q*bb(NIR)/(a(NIR)+bb(NIR)) Bb(NIR) = b b w (NIR)+bbchl(NIR) + bbnc(NIR) a(NIR) = aw(NIR)+ achl(NIR) + anc(NIR ρW (865nm) ρ W (865 nm )

2. Process of Atmospheric Correction _ Step 4. Remove Aerosol Signal ρ‘ (λ) Td(λ) ρ WMOD(λ) + ρA(λ)+ρRA(λ)+ error(λ)ρWMOD parameters (λ, chl, Bb S ) ρ Aerosol MOD parameters (C 0 , C1 , C2) Min-error (λ) Final value( chl, C0, C1, C2) ρW(λ) SGCA (Sun-glint Correction Algorithm) Basic Assumption : ρ W MOD ( λ ) is valid Polynomial fitting : ρ W MOD ( λ ) & ρ Aerosol MOD ( λ ) ρ W MOD ( λ ) : Using Biogenic optical model (by A.Morel ) ρ Aerosol MOD ( λ ) : C 0 + C 1 λ -2 + C 2 λ -4

B1 2. Process of Atmospheric Correction _ Step 5. Apply Diffuse Transmittance Extract Rayleigh diffuse transmittance Generic Rayleigh diffuse transmittance model τ r(λ) : use H.R.Gordon’s model B3 B4 B8 Td r cos ( Ф ) Model’s Td r RTE’s Td r

2. Process of Atmospheric Correction _ Step 5. Apply Diffuse Transmittance Extract Rayleigh diffuse transmittance A simple Rayleigh diffuse transmittance model C 6 C 5 C4 C 3 C 2 C 1 C 0 412nm 2.446662E+00 -8.426278E+00 1.091486E+01 -5.986775E+003.424127E-011.212632E+00 3.582148E-01443nm 2.439042E-016.214171E-02-2.343571E+00 4.741604E+00 -4.368938E+002.218751E+00 3.401276E-01 490nm -3.409564E+00 1.368336E+01 -2.270315E+01 2.024385E+01 -1.059768E+01 3.364536E+00 3.456215E-01 555nm -6.190158E+00 2.375412E+01 -3.712744E+01 3.049661E+01 -1.420755E+01 3.801402E+00 4.276636E-01 660nm -6.027454E+00 2.276901E+01 -3.481947E+01 2.770477E+01 -1.228477E+01 3.025252E+00 6.094426E-01 680nm -5.722233E+00 2.158916E+01 -3.295611E+01 2.615090E+01 -1.154451E+01 2.820577E+00 6.416646E-01 745nm -4.680227E+00 1.760824E+01 -2.677182E+01 2.111729E+01 -9.234431E+00 2.219140E+00 7.273351E-01 865nm -3.040593E+00 1.140555E+01 -1.727012E+01 1.354123E+01 -5.866066E+00 1.386646E+00 8.353374E-01

2. Process of Atmospheric Correction _ Step 5. Apply Diffuse Transmittance Get aerosol diffuse transmittance from AOT Aerosol model, single scattering reflectance, single scattering albedo, phase function  Get aerosol optical thickness A simple aerosol diffuse transmittance model (Hajime Fukushima, 1998) Using Aerosol+Rayleigh LUT (Future work) A generic data driven method

GOCI with NASA standard 2011/03/17 03:16 (UTC) 3. Result & Validation _ ResultComparison images of GOCI & MODIS (NASA Standard Algorithm) MODIS with NASA standard 2011/03/17 05:05 (UTC)

3. Result & Validation _ Result Comparison spectrums of GOCI & MODIS (with NASA Standard Algorithm) B1 : 412nm B2 : 443nm B3 : 490nm (MODIS : 488nm)B4 : 555nm (MODIS : 551nm)B5 : 660nm (MODIS : 667nm)B6 : 680nm (MODIS : 678nm) GOCI MODIS GOCI MODIS

SSMM Rrs (412 nm ) SSMM Rrs (443 nm) SSMM Rrs (490 nm ) SSMM Rrs (555 nm )MODIS Rrs(412 nm) MODIS Rrs(443nm) MODIS Rrs(490nm) MODIS Rrs(555nm )GOCI : SSMM 2010/09/17 04:16 (UTC) MODIS : NASA Standard Algorithm 2010/09/17 04:45 (UTC) 3. Result & Validation _ Result Comparison images of SSMM & MODIS (NASA Standard Algorithm)

SSMM nLw (555nm): 2010. 08. 20 04:16 (UTC) SGCA nLw (555nm): 2010. 08. 20 04:16 (UTC) MODIS nLw (555nm): 2010. 08. 20 04:25 (UTC) Comparison nLw spectrums of SSMM & SGCA & MODIS (NASA Standard Algorithm) 3. Result & Validation _ Validation SSMM SGCA NASA Standard (MODIS)

4. Conclusion _ NASA Standard Algorithm for the GOCI Basic schema is all implemented. Need to improve the ocean color model Add more good arrangement aerosol models Need to consider the new aerosol model for the GOCI observation area Change to the look up table based diffuse transmittance estimation Aerosol model selection and weight method update SSMM Looks reasonable but needs more tuning Better result high turbidity water and blue absorption aerosol case Also consider about horizontal aerosol type changes Collect more reference site SGCA Relatively good matching at the high optical thickness case Improvement for turbid water Needs more local tuning

THANK YOU