MODIS Ocean Science TeamContributors to the Ocean Presentations

Mark Abbott Oregon State University Barney Balch Bigelow Otis Brown University of Miami Dennis.K.Clark NOAA Janet Campbell University of New Hampshire Ken Carder University of South Florida Wayne Esaias NASA Robert Evans University of Miami Howard Gordon University of Miami Frank Hoge NASA Eddie Kearns University of Miami Ricardo Letelier Oregon State University Peter Minnett University of Miami Ken Voss University of Miami

L’Aquila Summer Course - August 26, 2002

Just the sunglint slides…

Where to get data and more information

Information locations:

MODIS Oceans home page

– http://modis-ocean.gsfc.nasa.govMODIS Oceans QA Browse 36km Imagery (MQABI)

– http://jeager.gsfc.nasa.gov/browsetool/

Select Terra collection 4Useful links to documentation and related web pages

– http://modis-ocean.gsfc.nasa.gov/doclinks.html

Data Ordering locations:

NASA GES DAAC WHOM (NASA - Goddard DAAC)

http://daac.gsfc.nasa.gov/

– Select Ocean color ->MODIS->oceanEOS DATA GATEWAY EDG

- http://modis.gsfc.nasa.gov/data/ordering.html

http://modis-ocean.gsfc.nasa.gov/qa

MQABI Quality Assurance Web Page

… provides access to the available MODIS Ocean data collections, the ability to browse selected fields and links to information concerning the status of the Ocean products.

MODIS Compared to AVHRR, SeaWiFS

• 12 bit digitization vs 10 bit --> improved precision

• Lower noise detectors --> subtle features better resolved

• Global 1 km data stream vs 4km --> larger data sets

• Additional spectral channels --> improved and additional product algorithms, better quality determination

• Equatorial crossing time --> significant impact on atmospheric correction

• Shared calibration sources: – optical:- MOBY buoy, – infrared: MAERI interferometer, buoys

Radiance sources and sinks affecting visible and IR wavebands

Time-series of M-AERI measurements on Time-series of M-AERI measurements on Explorer of Explorer of the Seasthe Seas

The Explorer of the Seas is a Royal Caribbean Cruise Liner, operating a bi-weekly schedule out of Miami. It is outfitted as an oceanographic and atmospheric research vessel, very suitable for satellite validation. For more details see http://www.rsmas.miami.edu/rccl/

M-AERI

Atmospheric CorrectionEffects of the Atmosphere

• We want to measure the "color" of the ocean, but the satellite actually measures “ocean + atmosphere”. The atmosphere is 90% of the signal in the ‘blue’ segment of the spectrum, and it must be accurately modeled and removed.

• Some of the atmospheric effects that are included in visible “atmospheric correction” for retrieval of ocean water leaving radiance or reflectance include:

– Gaseous absorption (ozone, water vapor, oxygen).

– Molecular scattering (air molecules), also referred to as Rayleigh scattering. Reason for blue skies and red sunsets.

– Aerosol scattering and absorption (haze, dust, pollution). Whitens or yellows the sky.

• Adapted from http://seawifs.gsfc.nasa.gov/SEAWIFS/TEACHERS/CORRECTIONS/ Bryan Franz, SeaWiFS Project

Ocean-Atmosphere SpectrumWater-Leaving Radiance Retrieval Challenge

MODIS -

At satellite radiance

Rayleigh removed radiance

Water leaving radiance: Lw

Atmospheric Correction Equation

t : total reflectance measured at satellite

w : water-leaving reflectance.

r : contribution due to molecular (Rayleigh) scattering, which can be accurately modeled. MODIS requires accurate measurement of change in its mirror reflectivity with angle of incidence

a + ra : contribution due to aerosol and Rayleigh-aerosol scattering, estimated in NIR from measured radiances and extrapolated to visible using aerosol models.

wc : contribution due to whitecaps, estimated from statistical relationship with wind speed.

g : Sunglint reflectance from sea surface; SeaWiFS avoids by tilting sensor; MODIS does not tilt so sunglint must be removed.

t = r + (a + ra) + twc + tg + t w

Atmospheric Correction Differences: MODIS vs SeaWiFS

• Glint : Spectral diffuse glint term, modified Cox-Munk distribution with spectral weighting, must be removed for MODIS (non-tilting), SeaWiFS minimizes glint by tilting

• Rayleigh : Polarization varies with satellite and solar zenith angles, MODIS mirror angle of incidence (AOI) affects reflectivity, SeaWiFS has constant AOI mirror

• Multiple (10) detectors per spectral band : Affects Rayleigh, especially near nadir, SeaWiFS has 1 detector/band

• Equatorial crossing time : SeaWiFS noon, MODIS Terra 1030, MODIS Aqua 1330. Affects sun glint, bidirectional reflectance.

Atmospheric correction enhancements

for MODIS

• Instrument effects : detector and mirror side banding (banded appearance)

• Polarization (effect on MODIS optical train)

• Sunglint• Sun-satellite-observation point viewing

geometry, overpass time, BRDF (bidirectional reflectance)

Sunglint Correction

Sun glint influences large portions of the image. Several approaches to correcting glint problem were investigated.

a) Assumed sunglint was direct, i.e. no scattering component. Result: Lw’s (water-leaving radiances) decreased as increased sunglint was removed.

b) Removed diffuse Rayleigh scattering component of the sun glint. Result: Lw retrievals showed a spectral behavior. Lw’s were correct in green region but under-corrected in blue.

c) Added a diffuse aerosol component to the sun glint.

Result: improved Lw retrievals in regions of sun glint contamination with reasonable spectral behavior.

Uncorrected La 865nmYellow and red region is glint contaminated (Lg > 5*La).> 70% of swath affected.

Corrected La 865nm Sun glint removed

Glint Corrected La 865nm

Note wind-wave-current interaction. Glint suppressed in Gulf Stream region

Future Sunglint Enhancement

• Present sunglint correction relies on assimilation-modeled wind fields and Cox-Munk surface reflectance model (no wind direction)

• Does not reflect local, kilometer-scale winds or wind-current interaction

• Alternative : ‘measure’ surface reflectance by utilizing 4 and 11 micron infrared bands

Note aerosol fronts off U.S. East Coast and mid-oceanGoal : remove sunglint but retain aerosol signal

865-nm Aerosol Op Depth: May 8, 2000

412nm Water-Leaving Radiance: Cox-Munk glint correctedNote residual glint along Gulf Stream Core

869-nm Reflectance, Total minus RayleighMODIS IR regression fit F(4/11) where 4

includes reflected sunlight, 11 does not

Alternate approach: replace Cox-Munk with MODIS IR

IR sun glintcorrection

Cox-Munksun glintcorrection,slope afterpixel 800 isresidual sunglint

Cross-scan glint behaviour

Cox-Munk Glint Field

MODIS IR Channel Glint Field (‘Measured’)

Note IR glint field not aliasedby presence of aerosol front

Glint Reflectance: Modeled vs Empirical

Use of IR bands enables detection of sunglint missed with Cox-Munk approach

Cox-Munk minus IR-Empirical reflectance

MODIS, a nadir pointing sensor, is significantly affected by sunglint

Successful removal of sun-glint extends useful coverage ~20% of total scan or 30% of non-saturated scan

Glint modeled using Cox-Munk and diffuse Rayleigh+Aerosol spectral propagation provides excellent correction if a proper wind field is available and ocean velocity is negligible

MODIS IR bands are being investigated as an alternative providing a ‘measured’ surface reflectance to better accommodate wind-wave interaction

Initial results are encouraging

Conclusions

Modis Chl A3, May 8, 2000 K. Carder Semi-analytic

University of Miami

Rosenstiel School of Marine

and Atmospheric Science

February 2002

Strategies for sun-glint identification and correction in water-leaving radiances measured from MODIS

Robert H. Evans

Edward Kearns

Katherine Kilpatrick

Daily images of Chlorophyll_a2 (SeaWiFS equivalent)

Dec 4, 2000 (solstice) Apr 8, 2001 (equinox) June 10, 2001 (solstice)

MODIS Polarization and cross-scan correction

with reduced sunglint mask

Sun Glint Reflectance Equation

glint reflectance g = tg * t

tg = glintsc() * zglint * zbst() * t_star()

• glintsc() = 2.4-2.5 (spectral scale factor)

• zglint = sun glitter coefficient using Cox and Munk (1954a,b; 1956) ; assumes vector wind, modified wind speed input to adjust shape

• zbst = two-way (Sun --> surface --> satellite) Rayleigh and ozone diffuse transmittance;

• t_star = one-way (surface --> satellite) aerosol diffuse transmittance using chosen models;