You are on page 1of 18

SLC-Off Gap Filled

Product Generation
Richard Irish
SSAI, NASAs Goddard Space Flight Center
Richard.irish@nasa.gov
Pat Scaramuzza
SAIC, Center for Earth Observations and Science (EROS)

Remote Sensing Basics | June 14-20, 2009

Scan Line Corrector Failure


On May 31, 2003, Landsat 7 experienced an anomaly causing the
Scan Line Corrector (SLC) to stop functioning normally.
The telescope focuses energy onto a pair of motion compensation
mirrors (i.e. SLC) where it is redirected to the focal planes and
accounts for the compound effect of along-track orbital motion and
cross-track scanning which leads to significant overlap and underlap
in ground coverage between successive scans.
While missing data is lost, it was possible to modify existing
algorithms to produce imagery containing roughly 80% of the
expected scene.
The processed post-anomaly data continues to maintain its
expected radiometric and geometric fidelity.
Remote Sensing Basics | June 14-20, 2009

Scan Line Corrector

Remote Sensing Basics | June 14-20, 2009

SLC-off Archive Potential


Landsat 5 has outlived its design life by an astounding 22
years yet its demise will likely precede that of Landsat 7.
Over 660,000 SLC-off scenes have been acquired, archived,
and are available for free. Many uniquely represent the only
recent Landsat observations over foreign soil.
The goal of the algorithm presented here is to tap the value
represented by the SLC-off archive by filling missing pixels
with radiometry that has scientific purpose.
With a LDCM (Landsat - 8) scheduled for a late 2012 launch,
there maybe no choice.

Remote Sensing Basics | June 14-20, 2009

Algorithm Description
Step 1 - Scene Reframing (reframe_l1t.pro)
Although, the standard ETM+ L1T product is geo-located and
Orthonormalized, scenes for common path/rows are not
necessarily aligned. Scene reframing (using the GeoTIFF header)
establishes a common frame and achieves registration through
translation.
Step 2 - Gap Filling (gap_fill.pro)
Common neighorhood pixels (non-zero) are identified for each
primary scene pixel gap and its fill scenes counterparts.
Linear least-squares regression analysis, using 144 data
points yields the predictive equation used for filling.
Remote Sensing Basics | June 14-20, 2009

Adaptive Window Logic

(x,y)

Anchor scene

(x,y)

Fill scene 1

(x,y)

(x,y)

Anchor scene

Fill scene 1
= gap

Remote Sensing Basics | June 14-20, 2009

(x,y)

Common pixels = 4

(x,y)

Common pixels = 12
= valid data

Predicter Relationship Developed

Remote Sensing Basics | June 14-20, 2009

Frame_and_fill
The primary goal of this endeavor was to create freeware so anyone
with a desire could take greater advantage of the ETM+ SLC-off archive.
Unfotunately, reframe_l1t.pro and gap_fill.pro require an IDL license to run.
Fortunately, ITT provides a Virtual Machine a mechanism for creating
standalone applications called runtime distrubutions. The only penalty
is an ITT splash screen at program launch.
Creating a runtime distribution required learning the IDL GUI language,
a treat, and creating a solitary application called frame_and_fill.
Currently versions exist for Windows - 32 and Mac PPC. Plans are to
make additional versions for Mac Int (32, 64 bit), Windows-64, Linux
and Solaris (32 and 64 bit).
http://landsathandbook.gsfc.nasa.gov/handbook/software/gap_filling_software.html
Remote Sensing Basics | June 14-20, 2009

Frame_and_fill Data Staging Rules


Theprimaryscenetobefilledistermedtheanchorscene.It
MUSTresideinafoldernamedanchorforthesoftwaretowork.
Thefillerscenesresideinseparatefoldersatthesamedirectory
levelastheanchorfolder.
Theyaresonamedfill_scene_1,fill_scene_2...toreflecttheir
processingpriority.
Thesefoldernamingconventionsmustbefollowedforthe
softwaretowork.

Remote Sensing Basics | June 14-20, 2009

Frame_and_fill Opening Window

Note: Frame_and_fill will work on all L1T GeoTIFF formatted


Products available from EROS (i.e. Landsat 5, L7 SLC-on).
Also, if youve previously framed your scenes you bypass
the re-frame step.

Remote Sensing Basics | June 14-20, 2009

Reframe_L1T Parameters

Remote Sensing Basics | June 14-20, 2009

Gap Fill Parameters

Remote Sensing Basics | June 14-20, 2009

File Naming Conventions


Frame_and_fill expects the EROS file naming conventions for the input L1T
scenes and won't work if these are altered. A typical file name for band one
will appear as:
L71145039_03920081124_B10.TIF.
After re-framing the new band one file name will be:
L71145039_03920081124_B10_reg.TIF
After gap filling the final band one file name will appear as:
L71145039_03920081124_B10_reg_filled.TIF
Output band files are GeoTIFF formatted. Additionally, processing results text
files are produced.

Remote Sensing Basics | June 14-20, 2009

ETM+ Sub-scene Anchor

Path 190, Row 52 acquired on 1/20/08 in Western Nigeria.


Remote Sensing Basics | June 14-20, 2009

ETM+ Fill Sub-Scene #1

Path 190, Row 52 acquired on 2/22/08 in Western Nigeria.


Remote Sensing Basics | June 14-20, 2009

ETM+ Fill Sub-Scene #2

Path 190, Row 52 acquired on 3/08/08 in Western Nigeria.


Remote Sensing Basics | June 14-20, 2009

ETM+ Filled Sub-Scene

Path 190, Row 52 acquired on 2/22/08 in Western Nigeria.

Remote Sensing Basics | June 14-20, 2009

SLC-off Scene Selection Criteria


If possible, fill scenes should be cloud-free with minimal changes
In scene content (e.g. snow).
Choose fill scenes that are temporally close to the anchor to minimize
changing ground cover. A second choice would be anniversary scenes
where scene content is similar.
Select fill scenes that are within 4 WRS cycles of the anchor. The
ETM+ scan cycle changes slowly over time meaning the relative scan
gap phase will vary across the scene if the acquisition date interval
Is too great. What appears to be a perfect match may not be.
One might consider using an SLC-on scene as the final fill
component. This will ensure complete gap coverage.
Remote Sensing Basics | June 14-20, 2009

You might also like