MODIS Aqua PSF (Point Spread Function)

1
MODIS Aqua PSF (Point Spread Function)

• Gerhard Meister, OBPG
• NASA Goddard Space Flight Center
• (Futuretech Corp.)
• January 4, 2006
• 
  

2
Background

• Ocean color remote sensing imposes very stringent
  requirements on radiometric accuracy of bands
  relative to each other.
• Stray light effects are very large for bright
  clouds next to dark ocean water.
• Theoretical solution correction with
  point-spread function (PSF).
• Problems PSF not well characterized, light
  source intensity sometimes unknown,
  computationally very expensive.

3
PSF definition
Lm (i0,j0) Sij PSF(i- i0,j- j0)LT(i,j) Lm
Measured radiance LT True radiance PSF Point
Spread Function i line index j column index
PSF kernel 25x25 i i0-12, i0 -11,..., i011,
i012 j j0-12, j0 -11,..., j011, j012
4
Available from SBRS

• LSF (Line Spread Functions) smaller slit, 1
  pixel from either side of slit, dynamic range
  1-10-3, scan and (simulated) track direction
• NFR (Near Field Response) measurements scans of
  1x10 slit (slit in track direction), up to 100
  pixels to either side of peak, dynamic range
  1-10-7
• NFR modeling scatter model, certain shape
  parameters, no absolute scale for PSF

5
Creation of PSF
Assume that PSF of adjacent pixel in scan
direction is given by 0.125/0.75 (theoretical
value from Geolocation ATBD) of center value,
adjacent pixel in track direction is 5 (from
LSF)
Adjust center value and scale parameter from
SBRS model (Harvey-Shack) so that NFR
measure- ments are reproduced
6
NFR measurements and model Band 11
7
NFR measurements and model All bands
8
Paper by Qiu et al. on Terra PSF
9
(No Transcript)
10
Aqua and Terra Band 11 with Lcloud/Ltypical20
11
Aqua all bands with Lcloud/Ltypical20 (PSF
size512x512)
12
Aqua all bands with Lcloud/Ltypical from 10-40
13
Same as before, but 10x10 pixel cloud
14
10x10 pixel cloud, band 11 only, different PSF
sizes
15
100x100 pixel cloud, band 11 only
16
Summary

• Created Aqua PSF based on SBRS model and NFR
  measurements
• For huge clouds, TOA radiances 50 pixels away
  from cloud will be contaminated up to 1 (even
  after correcting with 101x101 PSF)
• For 100x100 clouds, radiances will be up to 0.5
  too high after correction with 25x25 PSF (0.2
  for 51x51)
• For small clouds, radiances can be adequately
  corrected with 25x25 PSF as close as 2 pixels to
  the cloud

17
Outlook

• Still need to check the model in track direction
• Refinement needed for adjacent pixel
• Validate with real data (nLw should look
  better, but more noise is expected)
• If successful, implement into SeaDAS as an option
• Work with MCST on Terra PSF derivation/validation

18
Backup slides