Estimation of Aerosol Properties from CHRISPROBA Data

1
Estimation of Aerosol Properties from CHRIS-PROBA
Data

• Jeff Settle
• Environmental Systems Science Centre
• University of Reading

2
The Importance of Aerosols

• Atmospheric Correction of Images
• Aerosols and Climate
• Aerosols and Air Quality

3
The Need for Directional Measurements

• Reflection properties of the surface depend on
  position of the sun, and the geometry of sensing.
• Multi-temporal data can be properly evaluated
  only if they are normalised for these directional
  effects.
• Albedo is determined accurately only by
  integrating incoming and outgoing flux over all
  directions.
• Information on the structure of vegetation
  canopies may be retrievable by inversion of
  directional reflectance
• Data driven atmospheric correction is possible

4
The Need for Atmospheric Correction
Ground and TOA Reflectance Values in Green Light
Ground and TOA NDVI values
The main source of error in atmospheric
correction is uncertain knowledge of aerosol
loading
5
Aerosols and Climate

• Aerosols have direct and indirect effects on
  atmospheric radiation
• Direct
• They scatter and absorb radiation
• Indirect
• They act as cloud condensation nuclei, and affect
  the microphysical structure of the clouds formed
• Interaction between aerosols and clouds a major
  source of uncertainty

6
Sulphate Aerosols and Radiative Forcing of the
Climate
7
Global Aerosol Data

• Aerosols are highly variable in space and time
  concentrations vary by a factor 1000.
• Global climatologies are model based, or
  extrapolations from a small number of
  observations.
• Aerosol models exist, limited validation.
• Observational network (Aeronet) highly skewed
• tropospeheric aerosol loading is very poorly
  measured (NASA 1993, Modeling the Earth System
  in the Mission to Planet Earth

8
Aeronet Network
9
Aeronet Sites, Quality Assured
10
Correction of ATSR2 Images

• ATSR2 characteristics
• 1 km pixel size
• 2 view angles (0-20 and 50-55)
• 4 spectral channels (555, 655, 870, 1600 nm)
• Correction approach based on premise that surface
  reflectance is of the form (shape function) x
  (spectral function)

11
Methodology

• The essential method is inversion of a radiative
  transfer model for the TOA radiance field.
• The inversion is constrained by requiring the
  surface reflectance field to follow a certain
  generic pattern. A simpler version has been used
  successfully on ATSR2 data (2 view directions, 4
  wavelength channels). It is robust to the
  aerosol optical depth.
• The method is described in North et al (1999)
• (IEEE Trans. Geoscience and Remote Sensing, 37(1)
  pp 526-537)

12
ATSR-2 Atmospheric Correction (With thanks to
Peter North, ITE)
Green Channel Correction
Before Correction
After Correction
13
ATSR-2 Atmospheric Correction (With thanks to
Peter North, ITE)
NDVI Correction
Before Correction
After Correction
14
ATSR-2 Atmospheric Correction BOREAS SSA, 25-9-95
(With thanks to Peter North, ITE)
Top of atmosphere
Corrected image
False colour composite r1630nm (nadir), g870nm
(nadir), b555nm (along-track)
15
Validation of AATSR atmospheric correction (with
thanks to Peter North, ITE)
Aerosol optical thickness
Validation against sun photometer data
16
Sites to be Used in this Study
17
Wavelength Calibration of CHRIS
CHRIS has no spectral calibration device on board
so we need to find an external method of
spectral calibration. We aim to determine the
spectral displacement dl of the spectral response
curve resulting from launch conditions to within
an accuracy of 0.5 nm.
Method Observe a scene that is spectrally
bland, and preferably dark, through the
atmosphere and use observations of a prominent
atmospheric absorption feature, matching observed
and expected profiles.
The atmospheric absorption feature used is the O2
absorption at 762 nm, The ocean surface is
effectively black over the wavelength range 750 -
780 nm.
18
Wavelength Calibration of CHRIS Data
Within a spectral region encompassing just the O2
absorption, locate the detector j recording the
lowest observed signal and read the signals from
adjacent detectors j-2,j-1 and j1,j2.
This dip is due to the O2 absorption
Observed Detector Signal
Compare the observed signals with those predicted
using Radiative Transfer Theory and the known
CHRIS Spectral Response Curves Ri(l) shifted by a
range of possible dl between 3.5 nm (See the
figure).This is done for a typical range of
atmospheric optical depths t (i.e. visibilities)
- the instrument signal is effectively
independent of moisture and ozone content in this
spectral range. The predicted signals constitute
a Look-Up Table (LUT).
j-2
j1
j2
j-1
j
CCD detector cells about the minimum signal cell
j - aligned in the spectral direction
19
Simulated detector signals for an increasing
spectral shift dl at 2 different atmospheric
visibilities
dl 3.0 nm
Solar zenith is 40 degrees View zenith is 45
degrees
dl - 3.0 nm
Visibility 17 km
dl 0.0 nm
NEdL
dl 0.0 nm
dl - 3.0 nm
Detector Radiance mW/cm2/sr/nm
Increasing dl
dl 3.0 nm
NEdL
Increasing dl -1.5 to 1.5 nm step 0.5 nm
Visibility 26 km
CCD detector index
20
Results
The mean rms retrieval accuracy (over all
wavelength shifts) of the dl was found to be
better than 0.53 nm in the presence of detector
noise. Worst case 1.3 nm (50km visibility). We
found that the method was robust to uncertainties
in the (unknown) surface albedo and atmospheric
optical depth. Averaging the darkest pixels in a
calibration image will reduce the uncertainty.
The method will be extended to include the water
vapour absorption profile at 900-1000
21
(No Transcript)
22
Bands at Full Spectral Resolution
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)