This site has moved to www.usgs.gov/landsat
Please visit the new site and update any bookmarks you have. 
 
Top Page    Section 1    Section 2    Section 3    Section 4     Section 5    Section 6    Appendix A     Appendix B     Appendix C    Appendix D    Appendix E    References

Appendix A - Known Issues

A.1 Overview

Analyses performed on earlier Landsat data revealed the existence of imperfections or image artifacts caused by the instrument's electronics, dead or dying detectors, as well as downlink errors. As a descendant of the Landsat 4 and Landsat 5 TM sensor, the ETM+ was expected to generate data with similar characteristics. In the past, these anomalous effects were ignored or artificially removed using cosmetic algorithms such as histogram equalization during radiometric processing by the ground system. In keeping with the mission's goals (see Section 1), a more proactive approach was developed for Landsat 7 before launch. One of the goals of the enhanced Landsat ground system is to detect and remove image artifacts prior to radiometric processing. Remnant artifacts, if they exist, can be subsequently removed in a post-processing step using cosmetic algorithms as can be done for TM data.

Based on analyses of TM data, the ETM+ image artifacts were expected to be scan correlated shift, memory effect, modulation transfer function, and coherent noise. Dropped lines and inoperable detectors can also impact ETM+ data via decommutating (data stream synchronization) errors and detector failure, respectively. On-orbit analyses of ETM+ have also identified some impacts related to the operation of the Scan Line Corrector (SLC) (anomalous coherent noise and detector ringing), as well as a coherent noise storm in the data that was acquired as the SLC failed. Due to one or more of these issues, remnant artifacts may exist in a given scene and can appear as banding and striping. A discussion of these effects and characterization methodologies follows in this section. Eliminating their presence from data products is addressed in Appendix B.4. The ETM+ scan line corrector failure in 2003 is discussed in Section 2.2.3.3.

A.2    Scan Correlated Shift (SCS)

SCS is a sudden change in bias that can occur in all ETM+ detectors simultaneously. The bias level switches between two states and not all detectors are in phase; some are 180 degrees out of phase (i.e. when one detector changes from low to high bias, another may change from high to low). All detectors shift between these two bias states over time with the shift frequency varying from days to months. Measurement of SCS levels is affected by another instrument artifact known as Memory Effect (ME).

Characterization of possible SCS is performed at three levels. First, SCS levels are obtained for each detector in each line of the scene. This must be done to all scenes that require removal of SCS. Second, the value for the SCS levels is calculated for each detector. Third, the exact location of the transition from one SCS state to another within a scan line is determined, (assuming that a continuous flow of data from the detectors is available). SCS can be diagnosed by its random nature, by its fixed magnitude states, and by its appearance in otherwise completely dark nighttime data. On-orbit analyses of ETM+ data indicate that SCS does not exist in Landsat 7 data.

A.3    Memory Effect (ME)

TM data from Landsat 4 and Landsat 5 was rife with the artifact known as ME. It is manifested in a noise pattern commonly known as banding. It can be observed as alternating lighter and darker horizontal stripes that are 16 pixels wide in data that have not been geometrically corrected. These stripes are most intense near a significant change in brightness in the horizontal (along scan) direction, such as where a cloud (bright) and water (dark) are imaged together. Because of this, it was formerly termed 'Bright Target Saturation' or 'Bright Target Recovery.' Another artifact known as 'Scan Line Droop' was originally thought to be a separate phenomenon, but has since been shown to be simply another manifestation of ME. Because of its nature, ME impacted data have historically been the cause of significant error in calibration efforts because its effect on the Internal Calibrator (IC) data is scene dependent. It is present in TM Band 1 through Band 4 acquired at the PFP, and nearly absent in the Bands 5-7 acquired at the CFP.

ME is known to be caused by circuitry contained in the pre-amplifiers immediately following the detectors in the TM electronics. It is primarily due to a portion of a feedback circuit that contains a resistor / capacitor combination with a time constant of approximately 10 µm. This directly corresponds to time constants of 1100 minor frames (pixels) that have been derived from night scenes.

ME is characterized by both a magnitude and a time constant, which is a measure of the length of the artifact. If both values are known, ME is correctable (Helder et al., 1997). Most modern processing systems correct ME by default; thus, it usually appears only in Level 0 data or in data processed by older systems. Analysis of ETM+ data indicates that ME has never been detected in reflective band data, but insignificant amounts may exist in the thermal band. Ongoing studies will assess if it becomes a significant problem as the ETM+ sensor ages.

A.4 Coherent Noise (CN)

CN appears as a repeating pattern in satellite imagery and is most visible over dark homogenous regions (see Figure A-1). These patterns may appear in only one band or in several bands and may or may not be phase-locked to the instrument scan time. Although CN can be corrected, this correction often degrades other parts of the image; therefore, correction is performed only for very high magnitude noise sources.

Figure A-1 Example of CN in Band 3 of a Landsat 7 ETM+ Level-1
Figure A-1. Example of Coherent Noise in Band 3 of a Landsat 7 ETM+ Level-1.

Currently, the only CN correction performed for Landsat 5 and Landsat 7 is for Nyquist Noise (see Figure A-2). Nyquist frequency CN is phase-locked to the mirror scan and arises in the analog / digital conversion circuits of an instrument. Nyquist noise is easily removed and usually unseen in Level-1 images.

Figure A-2 Example of Nyquist Frequency CN in Band 1 in a Landsat 5 MSS Image
Figure A-2.  Example of Nyquist Frequency Coherent Noise in Band 1 in a Landsat 5 MSS Image.

The other form of CN is Anomalous Coherent Noise, which appears when a CN source changes in magnitude across a single instrument scan. It has been seen in Band 1 and Band 8 on Landsat 7 ETM+. It only affects a few detectors in each band. The Landsat 7 ETM+ anomalous CN is largest in the IC region and at the edges of a scene. It drops to zero magnitude in the center of a scene.

The Anomalous Coherent Noise sources on Landsat 7 appear at 20 Kilohertz (kHz), which is the same operating frequency as the SLC. As a result, a link between the anomalous noises and the SLC has been theorized. This theory appears to have been confirmed with the failure of the SLC in June 2003. After ETM+ returned to normal operations, it was observed that the 20 kHz noises in Band 1 and Band 8 were decreased in magnitude, and the anomalous nature of CN has been greatly reduced.

Anomalous Coherent Noise is a known problem on Landsat 7. Even though it is uncorrectable, it is not a cause for concern. However, new Anomalous Coherent Noise sources require further analysis.

A.5 Dropped Scan Lines

Dropped scan lines can occur in ETM+ Level 0R data due to decommutating errors in the raw data stream ingested by Landsat Processing System due to transmission problems between the satellite and the receiving ground station. Because Landsat 7 data are downlinked and processed in two different data format streams, transmission-related data loss usually only affects one format—either Bands 1-5 and Band 6 low gain, or Bands 7-8 and Band 6 high gain. Missing scan lines may also be caused by a temporary problem in the ETM+ main scan mirror. An example of dropped scans in Level-0 data is shown in Figure A-3.

Figure A-3 Dropped Scan in Band 7 of Landsat ETM+ Level-0 Data
Figure A-3.  Dropped Scan in Band 7 of Landsat ETM+ Level-0 Data.

A.6 Inoperable Detectors

Also known as Detector Failure, Inoperable Detectors is a term used when transient or permanent anomalies arise in an individual detector. Detector Failure is uncorrectable, although interpolation of data from adjacent detectors can mitigate its effects. All forms of Detector Failure are matters for serious concern and may indicate progressive problems onboard the instrument.

Transient Detector Failure, also known as a “flaky detector”, occurs when a single detector undergoes a sudden and drastic change in bias. Usually, the detector slowly returns to nominal behavior over the course of several scans. The cause is unknown, but is most likely due to energetic particle strikes within the detector circuitry. Transient failure has been observed once in Landsat 7 ETM+ data (see Figure A-4).

Figure A-4 Transient Failure in Landsat 7 ETM+ Level-0 Band 6, Detector 6
Figure A-4. Transient Failure in Landsat 7 ETM+ Level-0 Band 6, Detector 6.

A Flat Detector is a permanent but partial detector failure. Either the detector’s bias values, gain values, or both, are affected, resulting in reduction in dynamic range. The detector return values may only vary within a range of a few DN values, or they may register a single constant value. Also, detector values may exhibit larger than usual random noise. Flat detectors have been seen in Side B engineering data in Landsat 7 ETM+, but should not appear in normal imagery.

Absolute, permanent detector failure is known as a Dead Detector. Dead Detectors do not return any signal, and have a constant DN value of 0. Although previous Landsat instruments have exhibited Dead Detectors near the end of their operating lifetime, there are currently no Dead Detectors on any instruments aboard Landsat 7.

A.7 Banding

Banding or "scan-to-scan striping", are common terms for a visible noise pattern that affects Landsat data. Three artifacts are known to cause Banding.

SCS (see Section A.2) causes banding by forcing each band to randomly choose one of the two or more bias states. This banding has a stable magnitude across each scan. Figure A-5 shows an example of SCS-caused Banding.

Figure A-5 Example of Banding due to Scan Correlated Shift (SCS)
Figure A-5. Example of Banding Due to Scan Correlated Shift.

ME (see Section A.3) causes banding-like patterns that change in magnitude across a scan (see Figure A-6). This banding is caused by a large radiation transition in the scanning direction (from bright target to dark or from dark target to bright), either in the imagery or in the Internal Calibrator (IC) region of the data. For ME to create banding, the target is usually many scan lines in size, and if the target is in the imagery, the banding may visibly change over the bright target.

Figure A-6 Example of Banding due to Memory Effect (ME)
Figure A-6. Example of Banding Due to Memory Effect.

The third artifact known to cause banding is Calibration Error. Also known as True Banding or Sweep Striping, Calibration Error Banding is caused by errors in calibration algorithms that treat forward scans and reverse scans separately, so that one direction of scan has a different bias than the other. Calibration Error Banding is stable across the scan and consistently alternates between high and low bias. Calibration Error Banding is rare and corrected when found.

More than one of these artifacts can affect an instrument, causing banding of a complex character that may be difficult to diagnose with a cursory visual inspection. All forms of banding are known artifacts, all correctable and do not cause a concern.

A.8 Striping

Striping is a line-to-line artifact phenomenon that appears in individual bands of radiometrically corrected data. Its source can be traced to individual detectors that are calibrated incorrectly with respect to one another (true striping, or "calibration striping"). Another form of this problem is "saturation striping" which can occur over bright snow/ice surfaces or clouds (see Figure A-7). Saturation striping is normal and expected in Level-1R products and is corrected during LPGS processing.

Figure A-7 Saturation Striping in Landsat 7 ETM+ Level-1R Band 3 Data
Figure A-7. Saturation Striping in Landsat 7 ETM+ Level-1R Band 3 Data.

Saturation striping may also appear in some Landsat 7 gain-change scenes. In these scenes, part of the scene is converted from one gain setting to another. If high gain data with a small dynamic range is converted to low gain data with a large dynamic range, it may saturate at a value less than 255 DN. The processing system does not recognize it as saturated data and may cause striping in radiometric correction processing. Saturation striping in Level-1R or L1G gain-change scenes is expected and not a cause for concern. If saturation striping is visible in a normal Level-1 product, corrections need to be made to the calibration of the instrument.

A.9 Scan Line Corrector Failure

On May 31, 2003, the SLC aboard Landsat 7, malfunctioned. The failure of the SLC eliminated the instrument’s ability to compensate for the forward motion of the satellite. Subsequent efforts to recover the SLC were not successful, and the problem is permanent since that date. The ETM+ is still capable of acquiring useful image data with the SLC turned off, particularly within the central portion of any given scene. The center of an SLC-off scene is very similar in quality to previous Landsat 7 data. However, the scene's edges contain alternating scan lines of missing data (see Figure A-8). Figure A-8  shows that the scan gaps increase in magnitude toward the scene edges.The precise location of the affected scan lines varies from scene to scene. See Section 2.2.3.3 for additional material on this instrument artifact

Figure A-8 A subset of a Landsat 7 ETM+ scene showing the SLC-off scan gaps. The scan gaps increase in magnitude toward the scene edges
Figure A-8. A Subset of a Landsat 7 ETM+ Scene Showing SLC-off Scan Gaps. 

 


Top Page    Section 1    Section 2    Section 3    Section 4     Section 5    Section 6    Appendix A     Appendix B     Appendix C    Appendix D    Appendix E    References