Transcript ppt

Atmospheric Correction for
Ocean Color Remote Sensing
Geo 6011
Eric Kouba
Oct 29, 2012
Ocean Color Overview
Measured data
Top of Atmosphere Radiance
Need to do
Atmospheric Correction
Desired signal
Water-Leaving Radiance
or Remote Sensing Reflectance
****************************************************************************
Use other equations and methods...
Proxy parameter
e.g. Chlorophyll-A concentration
Biological parameter
Phytoplankton primary productivity
Desired goal
Information about health of the ocean
Sunlight to Surface to Sensor
1. Solar spectrum at top of atmosphere
2. Atmospheric absorption, scattering, etc
3. Clouds
Thin clouds allow some visibility
4. Reflection from top layer of ocean
Case 1 clear waters (tens of meters)
Case 2 turbid waters (less penetration)
5. Atmospheric absorption, scattering, etc
6. Radiance measured by satellite sensor
Only 10% to 20% of signal comes from ocean waters
Ocean Color Sensors and Satellites
CZCS sensor on Nimbus-7 satellite
SeaWiFS sensor on OrbView-2 satellite
MODIS sensor on Terra satellite
MERIS sensor on ENVISAT satellite
MODIS sensor on Aqua satellite
ETM+ sensor on LANDSAT satellite
Sensors and Satellites - Past
www.ioccg.org
Sensors and Satellites - Current
www.ioccg.org
Sensors and Satellites - Planned
www.ioccg.org
Major types of correction methods
Know which one your software uses
Dark object subtraction
Invariant object subtraction
Histogram matching
Cosine estimation of atmospheric transmittance
Contrast reduction
Path extraction
Spectral shape matching method
Others
->
CAAS
SeaWiFS
seadas.gsfc.nasa.gov
SeaDAS Atmospheric Correction
Wavelength dependent equation
ρT(λi) = ρR + ρA + ρC + ρSG + ρWC + transmittancesurf-sensor * ρW
ρT
ρRA
ρAS
ρCPL
ρSG
ρWC
ρW
Top of atmosphere reflectance
Rayleigh scattering
Aerosol scattering
Coupled Rayleigh-Aerosol effects
Sun glint
Whitecaps (wind on sea surface)
Water-leaving reflectance
Sensor
Not difficult
Difficult
Assume zero
Assume zero
Less than 7 m/s
Desired
Basically a dark object subtraction, with additions
Assumes NIR (765 and 865 nm) should be zero
12 aerosol models -> 25000 simulation runs -> Lookup tables
Works well for deep, clear, and low growth water
Fails in shallow, turbid, and high growth water
Options when standard correction fails
Flag and ignore regions that are difficult to process
If available, use SWIR instead of NIR for dark subtraction
but
MODIS has SWIR signal to noise problem
Simultaneous spectroradiometer measurements in field of view
but
Not practical for daily operations
Take spectroradiometer measurements nearby
but
Atmospheric parameters vary in time and space
Use other algorithm with standard atmospheres
but
Atmospheric parameters vary in time and space
Use algorithm to get aerosol correction from within image data
e.g. Shanmugam (2012)
aeronet.gsfc.nasa.gov
CAAS - Basic Equation
LT(λi) = LR + LA + LC + transsun-surf * LSG + LWC + transsurf-sensor * LW
LT
LR
LA
LC
LSG
LWC
LW
Top of atmosphere reflectance
Rayleigh scattering
Aerosol scattering
Coupled Rayleigh-Aerosol effects
Sun glint
Whitecaps (wind on sea surface)
Water-leaving reflectance
Sensor
Not difficult
Estimated
Estimated
Estimated
Ignored for now
Desired
Use Rayleigh-corrected Radiance to derive Aerosol correction
Spectral shape matching method
See Shanmugam (2012) pg 205-207 for math discussion
Conclusions
NIR dark subtraction fails in shallow, turbid, and high growth water
Like counting worldwide trees, but failing in the jungle
Aerosol modeling remains difficult
Atmospheric correction
is important...
Choose wisely
Questions?