Download Landsat 8 (L8) Data User Handbook (.pdf 4.39 MB)
Certain artifacts are expected in all satellite-borne sensors, and Landsat is no exception. However, moving to a push-broom sensor configuration with few moving parts, along with thorough testing, has dramatically reduced the number of issues and artifacts that have been observed on L8. Of note, several artifacts that existed on previous Landsat instruments, including cross-track banding, scan correlated shift, and dropped scan lines, are not possible on the OLI and TIRS instruments. Other artifacts such as coherent noise and memory effect exist only at very low levels and are virtually undetectable.
USGS maintains a list of L8 OLI and TIRS calibration notices that describe new or temporary artifacts and how they are addressed and corrected, and provides up-to-date information and details regarding known issues and current artifacts. In general, the known OLI and TIRS instrument image artifacts consist of stray light (TIRS only), striping, oversaturation (OLI only), and Single Event Upsets (SEUs).
The TIRS instrument is well within requirements for noise and stability, as determined using its onboard calibration systems. Comparisons of L8 calibrated data to surface buoy-based predictions, however, indicated a significant overestimation of radiance (TIRS results are too hot) and high variability in these comparison results. Differences between the ground-based results and TIRS results ranged up 5 K in the TIRS Band 10 and up to 10 K in TIRS Band 11. Some of this variation is related to the time of year (i.e., the temperature differences were larger during the summer, when the land surrounding the calibration sites was warmer). Additionally, some TIRS data are affected by significant banding, a low frequency variation in signal across the FOV, particularly over the three focal plane sensor chip assembly boundaries, even when viewing uniform regions. In addition, the banding may vary within a given scene in the along-track direction.
The banding that occurs in some measurements, but not in others, was hypothesized to be caused by stray light entering the optical path from outside the direct FOV. In these cases, the out-of-field light introduced additional incident energy on the detectors that was not uniform across the detector arrays. For example, during on-orbit check out, the image in Figure A‑1 was acquired over Lake Superior. The image shows that the uniformity appeared to change as the TIRS instrument scanned along track from northeast to southwest across the lake. This banding change is indicated in Figure A‑1 by red arrows.
Figure A-1. TIRS Image of Lake Superior Showing Apparent Time-Varying Errors
Several scans from the Earth to the Moon were acquired in order to investigate the potential stray light in the optical system. The analysis of these lunar scans, along with subsequent optical modeling efforts, confirmed that radiance from outside the instrument’s FOV was adding a non-uniform signal across the detectors and caused the observed banding. Additional scans of the Moon were acquired in order to better quantify the amount and location of stray light affecting TIRS imagery. Figure A‑2 shows the results of this lunar scanning. The gray lines indicate the angle between the TIRS boresight and the Moon where there was no ghost visible in any of the TIRS detectors. The blue lines indicate where a lunar ghost was observed by at least one detector within the TIRS focal plane.
Figure A-2. TIRS Special Lunar Scan to Characterize the Stray Light Issue
The larger amount of stray light observed in Band 11 is consistent with previous observations of variation in the accuracy of the imagery based on ground measurements and the larger amount of banding. Therefore, there is little doubt that the errors observed in the current thermal band data are caused by stray light.
At this time, efforts to understand the stray light paths fully are continuing and correction methods are being developed. In the meantime, the Landsat CVT adjusted the TIRS band’s radiometric bias in order to improve (but not fully eliminate) the absolute radiometric error for typical Earth scenes during the growing season. The bias corrections were implemented when the L8 data were reprocessed in February 2014. These corrections minimized the bias in temperatures derived from TIRS instrument data for typical Earth scenes. An uncertainty of ± 1 K (1 sigma) remains in temperatures derived from Band 10 data and an uncertainty of ± 2 K (1 sigma) remains in temperatures derived from Band 11 data for the test data set.
Specifically, this stray light error was estimated to be 0.29 W/m2/sr/mm for Band 10 and 0.51 W/m2/sr/mm for Band 11. Investigations showed the TIRS instrument reported a higher radiance than the water buoy measurements, as shown in the left graph of Figure A‑3. The graph on the right of Figure A‑3 shows the results after accounting for this average radiance error, which is the calibration adjustment that was implemented for reprocessing in February 2014. As described above, this adjustment minimizes the apparent bias, but does not change the variance in the data resulting from stray light. Additionally, colder scenes tend to be overcorrected due to this bias adjustment. The variability in the offset, as shown in Table A‑1, is about twice as large in Band 11 as Band 10, and Band 10 is about twice as large as Landsat 7 ETM+ thermal band uncertainty. Due to the larger calibration uncertainty associated with Band 11, it is recommended that users refrain from relying on Band 11 data in quantitative analysis of the TIRS data, such as the use of split window techniques for atmospheric correction and retrieval of surface temperature values. Users should note the errors from this stray light effect are dependent on the surrounding area temperatures (including clouds), so 1-sigma variances may be misleading.
Figure A-3. Thermal Band Errors (left group) Prior to Calibration Adjustment and (right group) After Calibration Adjustment
Table A-1. TIRS Band Variability
Correction for the TIRS stray light requires knowledge of the temperature of the areas surrounding TIRS scenes. Several methods are being pursued to obtain this information: (1) using GOES or another coincident data set to estimate the out of FOV sigma; and (2) estimating the out view signal based on in-scene statistics. Each of these approaches has pros and cons, and none has been proven effective in every case. However, all of these approaches are in early exploratory stages and improvements are expected.
Once the stray light issue is fully understood and a correction method chosen, all TIRS products will be reprocessed to include a stray light correction.
Striping is a phenomenon that appears as columns of consistently lighter or darker pixels in a single band of radiometrically corrected data. Banding is a similar phenomenon, but occurs across multiple contiguous columns. Both are often caused by incorrect calibration of detectors with respect to one another. These effects in OLI imagery are less than 4 DN in its 12-bit dynamic range, or less than 0.5 percent of the radiance of a typical Earth image, and are generally seen in OLI Bands 1 (CA), 2 (Blue), and 9 (Cirrus). This low level of non-uniformity is typically not visible to most users of L8 data. Figure A‑4 and Figure A‑5 are examples of striping and banding.
Figure A-4. Striping and Banding Observed in Band 1 (CA Band)
Figure A‑4 displays the striping and banding observed in Band 1 (Coastal Aerosol band) in image LC80160042013118LGN00, which is over a very homogenous region of the Greenland ice sheet. Single-pixel wide columns are considered striping; the thicker columns are banding.
Figure A-5. Striping and Banding observed in Band 2 (Blue)
Figure A-5 displays the striping and banding observed in Band 2 (Blue band) in image LC80750882013131LGN01, which is over clear, calm ocean water.
Striping and Banding also affect TIRS imagery, as seen in Figure A‑6.
Figure A-6. Striping and Banding observed in TIRS Band 10
Figure A‑6 displays the striping and banding observed in TIRS Band 10 in image LC80750882013131LGN01, which is over clear, calm ocean water. This low level of non-uniformity is typically not visible to most users of L8 data.
Normal radiometric processing removes the majority of banding and striping in OLI and TIRS imagery. The CPF used during Level-1 processing provide the parameters for corrections. However, minor responsivity changes in individual detectors cannot be accounted for, so some non-uniformity will remain in the data. Some of this striping may be corrected with future calibrations.
Both OLI and TIRS are designed with discrete SCAs, which have some detectors overlapping the adjacent SCAs. This causes a small region between each SCA to be viewed by multiple detectors. The imagery for this region is created by averaging over the detectors viewing each pixel. In some situations, this can cause an artifact that resembles banding, but it is in fixed locations and often transient, depending on the underlying terrain.
SCA overlap artifacts are most visible over high clouds, as the detectors on each SCA view a slightly different area of each cloud. Figure A‑7 shows an example of this.
Figure A-7. SCA Overlap Visible in Band 9 (Cirrus Band)
Figure A-7 displays the SCA overlap artifact observed in Band 9 (cirrus band) in image LC80140362014091LGN00, where high clouds were visible at the boundary between two SCA’s.
SCA overlaps are commonly seen in TIRS imagery because of contributions from the stray light artifact.
Figure A-8. SCA Overlap Visible in TIRS Band 10
Figure A-8 displays the SCA overlap artifact observed in TIRS Band 10 in image LC80372482013137LGN01 at the boundary between two SCA’s. In this image, stray light has shifted the calibration of two adjacent SCAs, making the overlap region visible.
Oversaturation occurs when a detector views an object that is much brighter than the maximum radiance the instrument was designed to handle. This causes the detector to deliver a voltage that is larger than expected by the 12-bit electronics, so the detector's value rolls over the 12-bit limit and records as a very small integer. Therefore, this artifact appears as dark spots in the middle of very bright objects. Figure A‑9 represents an oversaturation example in OLI SWIR Bands 6 and 7.
Figure A-9. Oversaturation Example in OLI SWIR Bands 6 & 7
Oversaturation artifacts are typically rare, but are repeatable artifacts that normally occur in OLI Bands 7 and 6 over large fires or volcanic events. Additionally, oversaturation may occur in other OLI bands over land surface objects that are very bright or exhibit strong specular reflectance. Oversaturation does not cause permanent harm to the instrument, and the detectors recover immediately with no visible memory effect.
An SEU is a “catch-all” term used for any electronic fault that causes brief, instantaneous artifacts in the imagery. Usually, SEUs are caused by charged particles from the Earth's radiation belts striking the detectors or instrument electronics. These particle hits are relatively rare, but are seen more commonly over the poles and over the South Atlantic Anomaly – a region where the Earth's magnetic field is weakest and the radiation belt is at its lowest altitude. Transmission errors can also cause SEUs, but due to the design of L8, these errors are not possible with OLI or TIRS imagery.
SEUs appear in OLI as single-frame bright spots that may affect several detectors in a line. The electronics design is such that SEUs often affect only odd or even detectors, but as seen in the lower portion of Figure A‑10, can affect both odd and even detectors, with a two-pixel gap between the odd and even detectors. Users may never notice an SEU because they are rare, typically only a single pixel, and geometric resampling distorts or erases these small artifacts in L1 products. They are most visible in Level-1R (radiometrically corrected) or L0 imagery. Figure A‑10 is an example of a couple SEU events measured by OLI. As noted above, the SEU clearly manifests as a line of single-frame bright spots.
Figure A-10. Example of SEU Event Measured by OLI – SEU Manifests as a Line of Single-Frame Bright Spots
SEUs occur at random. They cannot be corrected, but their effect on the imagery is minor. They do not cause permanent harm to the instrument detectors.
The L8 IAS geometry algorithms use ten coordinate systems. These coordinate systems are referred to frequently in the remainder of this document and are briefly defined here to provide context for the subsequent discussion. They are presented in the order in which they would be used to transform a detector and sample time into a ground position.
The OLI LOS coordinate system is used to define the band and detector pointing directions relative to the instrument axes. These pointing directions are used to construct LOS vectors for individual detector samples. This coordinate system is defined so that the Z-axis is parallel to the telescope boresight axis and is positive toward the OLI aperture. The origin is where this axis intersects the OLI focal plane. The X-axis is parallel to the along-track direction, with the positive direction toward the leading, odd numbered, SCAs (see Figure A‑11). The Y-axis is in the across-track direction with the positive direction toward SCA01. This definition makes the OLI coordinate system nominally parallel to the spacecraft coordinate system, with the difference being due to residual misalignment between the OLI and the spacecraft body.
Figure A-11. OLI Line-of-Sight (LOS) Coordinate System
The orientations of the TIRS detector LOS directions and of the TIRS Scene Select Mirror (SSM) are both defined within the TIRS instrument coordinate system. TIRS LOS coordinates define the band and detector-pointing directions relative to the instrument axes. These pointing directions are used to construct LOS vectors for individual detector samples. These vectors are reflected off the SSM to direct them out of the TIRS aperture for Earth viewing. The TIRS LOS model is formulated so that the effect of a nominally pointed SSM is included in the definition of the detector lines-of-sight, with departures from nominal SSM pointing causing perturbations to these lines-of-sight. This formulation allows TIRS LOS construction to be very similar to OLI. This is described in detail below, in the TIRS LOS Model Creation algorithm.
The TIRS coordinate system is defined so that the Z-axis is parallel to the TIRS boresight axis and is positive toward the TIRS aperture. The origin is where this axis intersects the TIRS focal plane. The X-axis is parallel to the along-track direction, with the positive direction toward the leading SCA (SCA02 in Figure A‑12). The Y-axis is in the across-track direction with the positive direction toward SCA03. This definition makes the TIRS coordinate system nominally parallel to the spacecraft coordinate system, with the difference being due to residual misalignment between the TIRS and the spacecraft body.
Figure A-12. TIRS Line-of-Sight (LOS) Coordinates
The spacecraft coordinate system is the spacecraft-body-fixed coordinate system used to relate the locations and orientations of the various spacecraft components to one another and to the OLI and TIRS instruments. It is defined with the +Z axis in the Earth-facing direction, the +X axis in the nominal direction of flight, and the +Y axis toward the cold side of the spacecraft (opposite the solar array). This coordinate system is used during Observatory integration and prelaunch testing to determine the prelaunch positions and alignments of the attitude control sensors (star trackers and SIRU) and instrument payloads (OLI and TIRS). The spacecraft coordinate system is nominally the same as the navigation reference system (see below) used for spacecraft attitude determination and control. However, for reasons explained below, these two coordinate systems are treated separately.
The navigation reference frame (a.k.a., the attitude control system reference) is the spacecraft-body-fixed coordinate system used for spacecraft attitude determination and control. The coordinate axes are defined by the spacecraft ACS, which attempts to keep the navigation reference frame aligned with the (yaw-steered) orbital coordinate system (for nominal nadir pointing) so that the OLI and TIRS boresight axes are always pointing toward the center of the Earth. The orientation of this coordinate system relative to the inertial coordinate system is captured in spacecraft attitude data.
Ideally, the navigation reference frame is the same as the spacecraft coordinate system. In practice, the navigation frame is based on the orientation of the absolute attitude sensor (i.e., star tracker) being used for attitude determination. Any errors in the orientation knowledge for this tracker with respect to the spacecraft body frame will lead to differences between the spacecraft and navigation coordinate systems. This becomes important if the absolute attitude sensor is changed, for example by switching from the primary to the redundant star tracker during on-orbit operations. Such an event would effectively redefine the navigation frame to be based on the redundant tracker, with the difference between the spacecraft and navigation frames now resulting from redundant tracker alignment knowledge errors, rather than from primary tracker alignment knowledge errors. This redefinition would require updates to the on-orbit instrument-to-ACS alignment calibrations. Therefore, the spacecraft and navigation reference coordinate systems are different because the spacecraft coordinate system is fixed but the navigation reference can change.
The spacecraft orientation rate data provided by the spacecraft attitude control system’s inertial measurement unit are referenced to the SIRU coordinate system. The SIRU consists of four rotation-sensitive axes. This configuration provides redundancy to protect against the failure of any one axis. The four SIRU axis directions are determined relative to the SIRU coordinate system, the orientation of which is itself measured relative to the spacecraft coordinate system both prelaunch and on-orbit, as part of the ACS calibration procedure. The IAS uses this alignment transformation to convert the SIRU data contained in the L8 spacecraft ancillary data to the navigation reference coordinate system for blending with the ACS quaternions.
The orbital coordinate system is centered at the spacecraft, and its orientation is based on the spacecraft position in inertial space (see Figure A‑13). The origin is the spacecraft’s center of mass, with the Z-axis pointing from the spacecraft’s center of mass to the Earth’s center of mass. The Y-axis is the normalized cross product of the Z-axis and the instantaneous (inertial) velocity vector, and corresponds to the negative of the instantaneous angular momentum vector direction. The X-axis is the cross product of the Y- and Z-axes. The orbital coordinate system is used to convert spacecraft attitude, expressed as Earth-Centered Inertial (ECI) quaternions, to roll-pitch-yaw Euler angles.
Figure A-13. Orbital Coordinate System
The ECI coordinate system of epoch J2000 is space-fixed with its origin at the Earth's center of mass (see Figure A‑14). The Z-axis corresponds to the mean north celestial pole of epoch J2000.0. The X-axis is based on the mean vernal equinox of epoch J2000.0. The Y-axis is the cross product of the Z and X axes. This coordinate system is described in detail in the Explanatory Supplement to the Astronomical Almanac published by the U.S. Naval Observatory. Data in the ECI coordinate system are present in the L8 spacecraft ancillary data form of attitude quaternions that relate the navigation frame to the ECI J2000 coordinate system.
Figure A-14. Earth-Centered Inertial (ECI) Coordinate System
The Earth-Centered Earth Fixed (ECEF) coordinate system is Earth-fixed with its origin at the Earth’s center of mass (see Figure A‑15). It corresponds to the Conventional Terrestrial System defined by the Bureau International de l’Heure (BIH), which is the same as the U.S. Department of Defense World Geodetic System 1984 (WGS84) geocentric reference system. This coordinate system is described in the Supplement to Department of Defense World Geodetic System 1984 Technical Report, Part 1: Methods, Techniques, and Data Used in WGS84 Development, TR 8350.2-A, published by NGA.
Figure A-15. Earth-Centered Earth Fixed (ECEF) Coordinate Systems
The geodetic coordinate system is based on the WGS84 reference frame, with coordinates expressed in latitude, longitude, and height above the reference Earth ellipsoid (see Figure A‑16). No ellipsoid is required by the definition of the ECEF coordinate system, but the geodetic coordinate system depends on the selection of an Earth ellipsoid. Latitude and longitude are defined as the angle between the ellipsoid normal and its projection onto the Equator and the angle between the local meridian and the Greenwich meridian, respectively. The scene center and scene corner coordinates in the Level-0R product metadata are expressed in the geodetic coordinate system.
Figure A-16. Geodetic Coordinate System
Level-1 products are generated with respect to a map projection coordinate system, such as the UTM, which provides mapping from latitude and longitude to a plane coordinate system that is an approximation of a Cartesian coordinate system for a portion of the Earth’s surface. It is used for convenience as a method of providing digital image data in an Earth-referenced grid that is compatible with other ground-referenced data sets. Although the map projection coordinate system is only an approximation of a true local Cartesian coordinate system at the Earth’s surface, the mathematical relationship between the map projection and geodetic coordinate systems is defined precisely and unambiguously.