CRCSI: Project 4.1

1
CRC-SI Project 4.1 PhD1 A method for the
remote sensing of aerosols based on MODIS
time-series data within an operational system for
near real-time atmospheric correction
2

• Topics
• Simple explanation of aerosol optical depth (AOD)
  algorithms used with MODIS
• Validation of these algorithms over Western
  Australia
• Stefan Maiers reflectance change process from
  the Bidirectional Reflectance Distribution
  Function (BRDF) algorithm used with MODIS
• A proof of concept for using reflectance change
  to determine AOD

3
Current Methods of AOD determination using MODIS
NASA institutional algorithms

• MOD04
• Uses surface and TOA reflectance values to select
  an appropriate model then an AOD value from a
  look-up table
• The surface reflectance values are determined by
  the relationships

4

• This relationship and this approach works best
  when the reflectance at 2.13 µm is within the
  range
• This reflectance range is generally only found
  over dense dark vegetation

5

• If the reflectance value of band7 is too high
  then no AOD value is calculated
• An AOD value is interpolated for 0.55 µm (band4)
  using the values for the other two wavelengths
• The AOD is determined on a 10km x 10km grid with
  1 value per wavelength per grid

6

• MOD09
• MOD09 uses the same wavelength relationships to
  determine surface reflectances at 0.47 µm and
  0.66 µm and a similar LUT approach to determine
  AOD
• The upper limit at 2.13 µm is increased to 0.4
• AOD values are determined on a 1 km grid with 1
  AOD value per wavelength
• An AOD determination is done for band8 rather
  than band4

7

• Validation
• Ground validation will involve the use of OD data
  from solar photometers and reflectance/radiance
  data from a spectroradiometer
• MOD04 and MOD09 AOD are being evaluated by
  examination of the data, specifically the
  coverage provided over W.A

8
Cimel Solar photometer
MRF-7 rotating shadowband radiometer
9
(No Transcript)
10
ASD Fieldspec Spectroradiometer
11
MOD04 AOD composites displayed as RGB of
Corrected Optical Depth Land for bands 1,4,3
Jan 6,7,11 Feb 3,5,8,10 Oct
13,19,20,23
12
MOD09 AOD composite displayed as RGB of Corrected
Optical Depth Land for bands 1,3,8
Oct 20
Oct 13,19,20,29
13
MOD04 Oct 13,19,20,23
MOD09Oct 20
True Colour Oct 20
14

• Stefan Maiers reflectance change algorithm
• The reflectance change is determined on a pixel
  by pixel basis by taking the value predicted by
  MOD43 BRDF model from the actual pixel
  reflectance
• MOD43 uses values over 16 days of observations
  from a single MODIS instrument
• A BRDF is produced for all seven MODIS
  reflectance bands

15

• This reflectance change is either a result of an
  actual change in the surface reflectance (land
  clearing, fire) or a change in the atmospheric
  contribution to the TOA signal
• The most likely candidate for changing the
  atmospheric contribution is aerosol
• Examining reflectance change in this manner
  should provide information on atmospheric aerosols

16

• Proof of concept for using reflectance change to
  determine AOD
• The following are examples of comparisons of the
  reflectance change as determined by Stefans
  algorithm and the change in AOD of MOD04 from
  mean aerosol values

17
-
MOD04 Comp Jan 6,7,11
MOD04 band3 Jan 17
18
The points where there is both RC and AOD change
are selected

AOD change band3 Jan 17
Ref. C change band3 Jan 17
19
These points are then plotted to examine the
correlation
Points to be compared
Correlation plot
20
This process was repeated using average values
from the MFRSR at Merredin for the areas that had
no MOD04 values
Extra values for comparison
New composite
21
New plot
Old plot
22
MOD04 Comp Oct 13,19,20,23,29
AOD change band3 Nov 2
23
Cloud
Correlation plot
24
MOD04 B3 Nov 2
Ref. C B3 Nov 2
True Colour Nov 2
25
(No Transcript)
26

• In conclusion
• MOD04 and MOD09 do not provide complete coverage
  of WA
• The AOD values show evidence of contamination by
  an over estimation of the path radiance due to a
  large contribution by surface reflectance
• Time-series data (MOD43) allows changes in
  reflectance due to atmospheric aerosol to be
  determined
• This shows a correlation with both ground and
  space derived AOD