Atmospheric Correction Algorithms for Remote Sensing of Open and Coastal Waters

Zia Ahmad

Ocean Biology Processing Group (OBPG) NASA- Goddard Space Flight Center

GEO-CAPE Workshop, August 18-20, 2008

Acknowledgements

Charles R. McClain and Ocean Biology Processing Group (OBGP), NASA-GSFC

Overview

• Background

• Atmospheric Correction (General)

• Overview of the Operational Method

• Recent Enhancements

• Summary and Conclusions

Ocean-Atmosphere Model

Dust Layer

SunSatellite

Scat. & Abs. in the Ocean

Upwelling Diff. Radiance

Downwelling Diff. Radiance

Ozone Layer

Ig IWL

Goal: LWL(λ)

IWC

Ltot (λ) = Latm (λ) + t1Lg (λ) + t2Lwc (λ) + t3 LWL (λ)

LWL carries valuable information about organic matter, phytoplankton, particulate matters, and other constituents of the upper ocean

Remote Sensing Reflectance Rrs

Rrs(λ)=LWL (λ)/Ed (0+) (λ)

Effect of PhytoplanktonEffect of CDOM

Water

absorption

scattering

Abs. coeff. for 765nm band is ~400 times higher than the abs. coeff for 412 nm band. Scattering coeff. for 765 nm band is ~16 times lower than scatt. coeff at 412 nm

Chlorophyll

Chlo. specific abs. coeff (a*) for 443 nm band is ~3.2 times higher than the chlo. specific abs. coeff for 555 nm band.

Scattering and Absorption Coefficients of Water and Chlorophyll

Ltot (top-of-atmosphere) and ILw

• Contribution from water-leaving radiance (t3LWL) to the TOA radiance (Ltot) @ 412 nm is ~ 12% for open ocean and ~5% for C. Bay

Atmospheric Correction

Ltot (λ) = [Latm (λ) +t1 Lg (λ)+ t2 LWC (λ)] + t3 LWL (λ)

• Latm(λ) = f (scattering by air molecules and aerosols in the atmosphere, and absorption by aerosols and trace gases like O3, H2O and NO2. Also, Fresnel reflection and sea state characterized by wind speed)• Lg(λ) = f (Fresnel reflection, and sea state - characterized by wind

speed and direction over the ocean)

Goal:

Determine [Latm (λ) +t1 Lg (λ)+ t2 LWC (λ)] as accurately as possible

Top-of-Atmosphere Radiance:

• LWC(λ) = f (sea state - characterized by wind speed)

Aerosol Models for Atmospheric Correction

• Howard Gordon’s (HG) Aerosol Models are Based on Shettle and Fenn’s Models for Tropospheric and Oceanic Aerosols

-Twelve (12) aerosol models are used in operational processing

-OceanicO99

-Maritime (1% oceanic and 99% tropospheric)M99, M90, M70, M50

-Coastal (0.5% oceanic and 99.5% tropospheric)C99, C90, C70, C50

-TroposphericT99, T90, T50

Some Properties of Operational Aerosol Models

c50

c90

Phase Function

- Effective radius varies from 0.14 to 4.74 μm

Size Distribution

c50

c90

- Single Scattering Albedo (SSA) varies from 0.98 (T50) to 1.0 (O99)

Atmospheric Correction Methodology

• Gordon and Wang’s algorithm uses measurements in NIR bands to select aerosol model

ε765, 865 = ρ765/ρ865ρWL (λ)=0 ρWL (λ)=0

- Select two models that bracket the observed ε765, 865 - In operational processing, ratio of single-scattering-reflectance values are used to compute ε765, 865

ελ, 865 = ρλ/ρ865ε765, 865 = ρ765/ρ865

Atmospheric Correction Methodology (cont.)

• NIR Correction

-For higher concentration of chlorophyll (chlo > 0.7 mg/m3 ) the assumption that water-leaving radiance in the NIR bands is zero is no longer valid -The correction is based on a bio-optical model that relates the Rrs in the NIR as:

Rrs (λ) = Rrs(λ0)*[atot(λ0)/atot(λ)]*[bb(λ)/bb(λ0)]η

atot(λ)=aw(λ)+aph(λ)+adg(λ)

bb(λ)=mλ+c

Here, λ0=670-nm, and λ=765- and 865- nm

• The objective of vicarious calibration is to normalize the observed TOA radiances to RT computed radiances.

• The method uses in situ data from MOBY site to calibrate the visible bands, and data from South Pacific Gayer and South Indian Ocean sites to calibrate NIR band.

• LwL is assumed to be zero for NIR bands (765 and 865 bands),

Vicarious Calibration

Vicarious Calibration (cont.)

Locations of in situ Data Time series of Gain Coefficients

- The gain coefficient for IR channel (765 nm) is determined from match-up data collected over the South Pacific Gayer (SPG) and the South Indian Ocean (SIO) sites

- The gain coefficients for all VIS Channels are determined from match-up data over MOBY site

Results for Deep Water (d >1000 m)

Vicarious Calibration (Validation)

Locations of in situ Data

- Results for deep water show very good agreement between in situ data and satellite retrievals.

•Use of SWIR Bands in Retrieving ILw Over Bay Area

• Wang and Shi’s Algorithm for Coastal Areas- Uses SWIR Bands to select aerosol models

Chlo. using NIR BandsRGB Image Chlo. using SWIR Bands

Recent Enhancements

A Comparison of NIR and SWIR Based Retrieval over the Bay Area

An Example of Negative nLw(412 nm) over the Eastern Coast of US

April 7, 1998

• Possible Reasons

- Operational aerosol size dist. are not representative of Bay area aerosols

- Perhaps Bay area aerosols are more absorbing than what is assumed in the operational processing

Chesapeake Bay (AERONET Stations)

SERC

Wallops Island

COVE

SERC: 38o, 53/ N, 76o, 30/ W

COVE: 36o, 54/ N, 75o, 42.5/ W

Wallops Island: 37o, 56.5/ N, 75o, 28.5/ W

Physical and Optical Properties of Aerosols over Bay Area

Monthly Mean Modal Radius Monthly Mean Modal Std. Dev.

Monthly Mean SSA

Effective Radius

Spectral Dep. Of SSA Single Scattering Albedo

SERC Cove Wallops Island

Aerosol Opt. Thickness (MODIS vs. AERONET)

AOT Based on New Models

AOT Based on Operational Models

Ewa KwiatkowskaCOVE:: Red Wallops Island:: Blue SERC: Black

Summary and Conclusions

• Gordon and Wang (GW) algorithm for atmospheric correction works reasonably well over open ocean (Case 1 waters). The retrieved values of water-leaving radiances and chlorophyll amount compare favorably with in situ data.

• Absorbing aerosols are problems. Also, over coastal areas, GW algorithm sometimes gives negative water-leaving radiance in 412 nm band. Work is progress to address this problem.

• With new aerosol models, the retrieved AOT in the four bands of the SeaWiFS sensor are in fairly good agreement with the AERONET data over the Chesapeake Area.