Landsat Collection 2

Landsat Collection 2 Level-1 Data

Return to Collection 2 Overview

Go to Collection 2 Level-2 Product Information 

 

Several data processing, geometric, and radiometric improvements, along with a new data distribution process, define Collection 2 Level-1 data. This page provides details about the changes users can find in Landsat Collection 2 Level-1 data. See the instrument-specific Data Format Control Books listed at the bottom of this page for a more comprehensive overview of these data characteristics. 

Level-1 Data Processing Improvements

  • Remove Extreme Latitude DEM Constraint
    Landsat Collection 1 Level-1 scenes with a Digital Elevation Model (DEM) value beyond the geographical extent (-57.0 degrees latitude) of the elevation source only process to a Level-1 Systematic Terrain (L1GT) product. Removing this constraint allows for more extreme latitude acquisitions to process to a Level-1 Terrain Precision (L1TP) product. For example, this is most pronounced at the southern tip of South America (path 227/row 99).
     
  • Landsat 7 ETM+ and Landsat 4-5 TM Precision Correction Logic
    Improvements were made to the Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Landsat 4-5 Thematic Mapper (TM) precision correction algorithm that includes a “decision-tree” like process to identify a good collection of correlated Ground Control Points (GCPs), and to constrain the model fit parameters depending on these GCPs within the scene content.  The number of Landsat 7 ETM+ and Landsat 4-5 TM Level-1 scenes achieving the Collection Tier 1 inventory criteria increases due to these improvements.
     
  • Landsat 8 Instrument Alignment
    In December 2014, Landsat 8 experienced a Thermal Infrared Sensor (TIRS) Scene Select Mirror (SSM) anomaly. After the change in operations due to this anomaly, a minor seasonal dependency in sensor alignment between the TIRS and Operational Land Imager (OLI) sensor was observed which is accounted for in Collection 2 processing.
     
  • Landsat 7 ETM+ Thermal Band Detector Alignment
    Landsat 7 ETM+ Thermal Band 6 includes a minor odd-to-even detector misalignment due to pre-launch detector-based focal plane values. After extensive historical on-orbit characterization and calibration efforts of the thermal band detector placement, revised focal plane values have been determined. The revised focal plane values provide improved detector-to-detector alignment of Band 6. This page provides more details about this improvement. 
     
  • Remove GCPs Constraint in Level-1 Processing
    In Collection 1, a constraint of 250 GCPs is used in Landsat 4-5 TM and Landsat 7 ETM+ Level-1 processing. This GCP constraint reduces the number of acquisitions that could process to a Level-1 Terrain Precision (L1TP) product. Together with the European Space Agency's Copernicus Sentinel-2 Global Reference Image (GRI) harmonization, removing this constraint allows more Landsat 4-7 scenes to process to a L1TP product.
     
  • Increase DEM Scene Framing Extent
    Present in some Collection 1 scenes was a duplicated ‘wedge’ of data in the extreme southwest corner of the scene. This primarily occurred in Band 1 as a result of insufficient scene coverage by a DEM tile. An increase in the DEM framing extent ensures scenes prone to this issue are captured by all elevation tiles, and the scene is precisely terrain corrected.
     
  • Landsat 7 ETM+ Instrument to Attitude Control System Alignment Update
    The Landsat 7 ETM+ instrument to Attitude Control System (ACS) alignment was updated between January 2013 - December 2018 and impacted Level-1 Systematic Terrain (L1GT) acquisitions only. The implementation of this update results in an increase accuracy of all Landsat 7 ETM+ L1GT acquisitions. Visit this page for more information. 
     
  • Day Transition Error Fix
    Previously, some Landsat 7 ETM+ and Landsat 4-5 TM Level-1 acquisitions captured during a day transition (primarily over Australia) failed to produce a Level-1 product or produced a misregistered Terrain Precision (L1TP) or Systematic Terrain (L1GT) product. The acquisitions affected by this error can now process successfully to a L1GT or L1TP product. Visit this page for more details and to see an example of the day transition error and fix in Collection 2.
     
  • Landsat 7 ETM+ and Landsat 4-5 TM Shutter Intrusion Indicated in New Sensor Anomaly Metadata Field
    Due to improved Internal Calibrator shutter intrusion detection for Landsat 7 ETM+ and Landsat 4-5 TM, a new metadata field (<SENSOR_ANOMALIES = “SHUTTER_INTRUSION”>) is included in the metadata file if intrusion is detected. Visit this page to see examples of shutter intrusion in Landsat imagery. 
     
  • Adjust Landsat 7 ETM+ Focal Plane for Improved Band Alignment
    The Landsat 7 ETM+ cold-to-warm focal plane alignment is adjusted to account for small misalignments that occurred during changes in operations, such as the Scan Line Corrector failure in 2003. This update produces improved co-alignment between the cold and warm focal planes of the Landsat 7 ETM+ instrument. Visit this page for additional information about focal plane calibration adjustments.

 

Geometric Accuracy Improvements

The Global Land Survey (GLS) database is the primary source for the GCPs library used to georeference Landsat Level-1 data. The most recent GLS 2000 data set is a collection of images acquired by Landsat 7 ETM+.

The absolute geolocation onboard Landsat 8, which was launched in 2013, is more precise than that of the GLS 2000 data set due to Landsat 8's onboard Global Positioning System. GLS 2000 updates have been implemented in phases since 2014 to improve the accuracy of the Landsat GCP library, with the most recent update occurring in May 2016 (Phase 3).  

Visit the Landsat Ground Control Points page for more information on GCPs, and to review details about each GCP Improvement phase.

Ground Control Points Phase 4
Landsat Ground Control Points (GCPs) Phase 4 involves re-baselining the Landsat 8 OLI GCPs to the European Space Agency's Copernicus Sentinel-2 GRI for Collection 2 data. This involved updating over 5.1 million GCPs and included the extraction of 2.5 million new Landsat 8 GCPs globally.

This update improves the per-pixel geodetic accuracy and interoperability of the global Landsat archive spatially, temporally, and with Sentinel-2.

he extent of the five geographic regions used in the Phase 4 Landsat ground control point improvement for Collection 2

Map 1. Geographic regions of Phase 4 GCP improvements. (Click to enlarge)

Map 1 displays the extent of the five geographic regions used in Phase 4 GCP improvement.  More information about Phase 4 GCP, including additional graphics showing the estimated net offsets between Landsat Collection 1 and Collection 2 Ground Control Points for each Landsat path/row, and summary statistics of the global GCP re-adjustment effort to harmonize Landsat 8 OLI-based GCPs with Sentinel-2 Global Reference Image can be found in the January 2020: Phase 4 - Collection 2 section of the Landsat Geometry Ground Control Points page. 

The article titled “Bundle Adjustment Using Space-Based Triangulation Method for Improving the Landsat Global Ground Reference” provides a description of the triangulation based method used in this Phase 4 GCP update.

 

 

Radiometric Accuracy Improvements

  • Landsat 8 TIRS Absolute Radiometric Calibration Update
    Landsat 8 TIRS absolute radiometric calibration has been updated to correct for errors detected by the vicarious calibration teams. The calibration parameters were adjusted to reduce radiometric calibration errors in Collection 1 products after the TIRS stray light correction implementation in early 2017.The calibration update is a linear gain and bias adjustment, though the magnitude of the correction varies over time and by signal level. All Collection 2 data are corrected for this error. The Landsat 8 OLI and TIRS Calibration Notices page provides details about each calibration update applied to improve the Landsat 8 data.
     
  • Landsat 8 TIRS Relative Gains
    The detector-to-detector radiometric calibration of both Landsat 8 TIRS bands was updated to correct for striping observed in Landsat Collection 1 image data. Analysis of on-board calibrator data established that striping has progressively deteriorated over the lifetime of the mission.  The images shown to the right compare the striping that is evident in TIRS Band 10 with Collection 1, with noticeable reduction in the Collection 2 image. (Data acquired December 26, 2019.)
Landsat Collection 1 vs Collection 2 TIRS Striping Comparison

Sample of a TIRS Band 10 image with visible striping when processed into Collection 1 (left), and the reduction of striping in Collection 2 (right). (Click to enlarge)

Both images are scaled to the same radiometric range. The reduction in the level of striping is clear across the relatively uniform water body.

Blackbody data were used to help determine new detector-to-detector relative gain parameters for every calendar quarter.  Blackbody data will continue to be used to update relative gains as new stripes are detected.

Plots were created to present a comparison between Collection 1 and Collection 2 striping factors derived from Band 10 and Band 11 for path 186 row 36 over the instrument’s lifetime (view the plots here).

The striping factor is a quantitative metric for the amount of striping present in an image that is obtained through calculation of spatial-domain statistics. Larger values for this metric tend to positively correlate with more visually apparent striping. The plots clearly suggest significant improvement in striping characteristics of thermal bands in Collection 2.
 

 

 

  • Landsat 8 OLI Absolute Gain Model Change
    The radiometric calibration model for the Landsat 8 OLI is based on an average response of the on-board calibrators (three lamp pairs, two solar diffuser panels and lunar acquisitions). The data from the working lamp pair, used daily, and from the working solar diffuser panel, used weekly, show trends that disagree with the other calibrators, indicating that their characteristics are changing independently of the instrument. The calibration model was updated to no longer include the working solar diffuser and the working lamp pair.

    Only the Coastal Aerosol (Band 1) and Blue (Band 2) are affected by this change. The radiometric calibration for these two bands is actively being updated with the on-orbit model; the radiometric calibration of the other OLI bands has remained stable since launch. This update influences all data in the Coastal Aerosol and Blue bands since launch (February 11, 2013) of up to 0.15 percent.
     

  • Landsat 8 OLI Relative Gain Update
    The Landsat 8 OLI relative gains are updated every quarter to account for variability in per-detector responsivity. The relative gains updates are derived from the solar diffuser collects acquired during the prior quarter.  For Collection 2, all relative gains estimated since the start of Collection 1 processing are updated with relative gains based on the data acquired during the applicable quarter itself. (Collection 1 products for data acquired before the start of Collection 1 processing have already been updated to make use of quarter-specific relative gains.)
    The differences between Collection 1 and Collection 2 images will generally be less than 0.2 percent; however, in the SWIR bands, there will be some individual detectors that change by up to 2.0 percent.  For the future quarters, the relative gains will continue to be updated based on the estimates from solar diffuser data acquired during the prior quarter.

 

  • Landsat 8 OLI Bias Calculation Update
Landsat 8 OLI Collection 1 vs Collection 2 Improved Bias Estimate

These Landsat 8 OLI images of Path 165 Row 198 over the Pacific Ocean (Band 3, acquired October 10, 2019) compare the striping in Collection-1 (left) and the reduction in striping in the Collection 2 image due to the improved bias estimate (right). (Click to enlarge)

The method of calculating the Landsat 8 OLI bias, used for dark subtraction in the calibration process, was modified to make use of Video Reference Pixels (VRPs).  Rather than base the dark-subtraction solely on shutter data (acquired within +/-40 min of the image data), the new bias estimate will consider VRP dark signal information acquired simultaneously with the image data.

The dark signal in the OLI drifts over time and the new bias estimate takes into consideration the short-term drift between shutter collects. The improved bias estimates will reduce along-track noise, improve image uniformity and improve signal-to-noise ratio.

The effect of this update will be primarily noticeable in low radiance scenes, as shown in Landsat 8 example to the right. The reduction in striping is a result of the improved bias estimate.

 

 

 

  • Landsat 5 TM Band 6 Calibration Update
    The Landsat 5 TM Band 6 radiometric calibration was updated to account for a residual bias error of -0.044 W/m2 sr um (-0.33K at 300K) for all data acquired on or after Jan 1, 1997. This error was detected after the previous calibration update (see the May 1, 2010 Calibration Notice) was applied. While the 2010 calibration update was signal dependent, this correction is constant for all signal levels. 

 

Data Access and Downloads

Landsat Collection 2 Level-1 data naming convention and band identifiers follow the same nomenclature as Landsat Collection 1. Level-1 data are delivered as Digital Numbers (DN) in an unsigned 16-bit integer format and can be converted to Top of Atmosphere (TOA) reflectance or radiance using the radiometric scaling factors provided in each scene metadata file.

The following bands are available for download individually or as a bundle from EarthExplorer

  • Spectral Bands in Cloud Optimized GeoTIFF (COG) format
    • Bands 1-11 (Landsat 8 OLI/TIRS)
    • Bands 1-8 (Landsat 7 ETM+)
    • Bands 1-7 (Landsat 4-5 TM)
    • Bands 1-4 (Landsat 4-5 MSS)
    • Bands 4-7 (Landsat 1-3 MSS)
  • Angle Files
    • Solar Azimuth Angle (SAA)
    • Solar Zenith Angle (SZA)
    • Sensor Azimuth Angle (VAA)
    • Sensor Zenith Angle (VZA)
    • Angle Coefficient File
  • Quality Assessment (QA)
    • Pixel Quality Assessment Band (QA_Pixel)
    • Radiometric Saturation and Terrain Occlusion QA Band (QA_RADSAT)
  • Metadata
    • Material Library File (MTL) (Object Description Language-based) format
    • Extensible Markup Language (XML)

Landsat Collection 2 Level-1 data will be available to download from EarthExplorer. The USGS Landsat no-cost open data policy remains intact since its inception in 2008.

New for Collection 2, users may select individual bands for download. For individual scene downloads, two or more band files will be grouped in a .tar file in order to provide a single download. EarthExplorer also maintains a bulk download option, which allows individual band selection, using the Bulk Download Application (BDA). Full Resolution Browse downloads will continue to be made available.

The image below is an example of the Full Resolution Browse and Level-1 band download options that will be available in EarthExplorer.  The Landsat Data Access page will soon be updated with Collection 2 information, for additional data access and download options. 

EarthExplorer Landsat Collection 2 Level-1 Download Options

Documentation

Users are encouraged to reference the COG Data Format Control Book (DFCB) for a detailed description of the data, product, and Spatiotemporal Asset Catalog (STAC) metadata files. Together with the instrument specific Level-1 DFCBs, the COG DFCB provides the user with a comprehensive overview of the new Landsat Collection 2 products.

Landsat Cloud Optimized GeoTIFF (COG) Data Format Control Book

Level-1 Documents