Landsat 8 Data Processing and Archive System (DPAS)

The USGS is also developing the DPAS to manage the mission data transmitted from the LDCM observatory. The DPAS will be operated at EROS along with the rest of its Landsat data archive and it will consist of several subsystems: Storage and Archive (SA), Ingest System (IS), Subsetter System (SS), Image Assessment System (IAS), Level 1 Product Generation System (LPGS), and User Portal (UP). These subsystems will work together to ingest, store, and archive LDCM data and will also generate LDCM data products for distribution.

The GNE will send LDCM mission data to EROS over the internet. The Storage and Archive (SA) subsystem will first receive, store, and archive the data in the file-based format received from the GNE. These data will be archived offline with an additional backup copy archived off site. The SA will also perform all archive and storage functions in support of the other DPAS subsystem functions.

The SA will pass data from the GNE to the Ingest System (IS). The IS will process the OLI and TIRS file-based mission data into an interval-based format for the on-line archive and will create the associated inventory metadata. In the process the IS will decompress the OLI data, analyze the data for impulse noise and saturated pixels, fill dropped data, eliminate duplicate data, correct ancillary data, generate metadata, and collect processing metrics. The data in this format will be called Level-0 Reformatted archive (L0Ra) data. The DPAS will use Hierarchical Data Format 5 (HDF5) for the L0Ra data with the data grouped in multiple HDF5 files(Folk & Choi, 2004), one file per OLI or TIRS spectral band plus an ancillary data file, metadata file, and checksum file. Each raw 12-bit datum from OLI or TIRS will be stored across two eight-bit bytes. The IS will send the Level 0Ra data files to the SA for online storage. The L0Ra data will be the format archived for long-term storage and this archive will be part of the USGS National Satellite Land Remote Sensing Data Archive (NSLRSDA).

The Subsetter System (SS) will retrieve L0Ra data from the SA and subset the OLI and TIRS L0Ra data into Landsat WRS-2 scenes for distribution or for the generation of Level-1 products and their associated metadata. These scene-based data sets will be called Level-0 Reformatted product (L0Rp) data and will be stored in the same HDF5 format as LORa data. No data processing will be performed up to this point with the exception of reformatting and filling in dropped data; L0Ra and L0Rp data files will contain raw OLI and TIRS data. With the use of 2 B for 12-bit data each uncompressed LDCM L0Rp data product will consist of 1384 MB for a full WRS-2 scene. In comparison, LDCM ETM + data products, with 8-bit data stored on single bytes, consist of less than 500 MB per scene. The DPAS will therefore store a daily average of 540 GB of L0Rp data given an average reception of 400 LDCM scenes per day.

L0Rp data files will be sent back to the SA for on-line storage and the Level-1 Product Generation System (LPGS) will retrieve each L0Rp scene to radiometrically and geometrically correct the image data. The radiometric correction will transform raw OLI data to digital counts linearly scaled to top-of-the-atmosphere spectral reflectance and will transform raw TIRS data to digital counts linearly scaled to at-aperture spectral radiance. In both cases the LPGS will scale the raw 12-bit data to 16-bit integers for the Level 1 products.

The geometric correction will use digital elevation models and ground control points to resample the radiometrically corrected data using cubic convolution and create orthorectified images of Earth’s surface registered to the Universal Transverse Mercator (UTM) cartographic projection or, in the case of polar scenes, registered to the Polar Stereographic projection. The OLI pixels will be resampled to a 30 m ground sample distance for each spectral band with the exception of the panchromatic band resampled to a 15 m ground sample distance. The TIRS data will be over-sampled from the 100 m sensor resolution to a 30 m ground sample distance for alignment to the OLI data in the final product. The co-registered and terrain corrected OLI and TIRS data will be merged to create a single integrated Level-1 data product.

The LPGS will also generate a full-resolution browse image and a quality assurance band that identifies filled-in pixels, identifies pixels obscured by terrain, and provides a cloud cover mask generated by an automated cloud cover assessment algorithm. The corrected digital images along with metadata and the quality assurance band will be referred to as Level-1T (L1T) data. The L1T data product will consist of a set of uncompressed GeoTiff files for each of the OLI and TIRS spectral bands, a file containing the scene-based metadata, and a file for the quality assessment band. LPGS will routinely generate a L1T product and a browse image for each of the 400 scenes collected per day.

The Image Assessment System (IAS) will perform OLI and TIRS data characterization, analysis, and trending over the operational life of the mission to monitor LDCM observatory performance and to create the calibration parameters required by the LPGS for Level-1 product generation. Calibration coefficients will initially be derived from pre-launch OLI, TIRS, and integrated observatory testing. The IAS will update coefficients during mission life using on-orbit OLI and TIRS observations of observatory calibration sources, surface calibration sites, and the moon. The IAS will make the calibration coefficients available to accompany requests for L0Rp data as well as to International Cooperators to help them remain current and consistent with products distributed by the LDCM ground system. The IAS will also manage and update the auxiliary data sets used by the LPGS to create L1T products. These auxiliary data sets include a ground control point library and digital elevation models.

The L0Rp and L1T WRS-2 scenes will constitute the standard LDCM science data products. The general public will be able to search, browse, and order L0Rp and L1T scenes from the User Portal (UP) on the internet. The UP will electronically transmit ordered scenes over the internet to LDCM data users at no cost to the users. Access through the UP will be nondiscriminatory and no restrictions will be placed on the use and redistribution of the data.

These subsystems of the DPAS will together orchestrate LDCM data archiving, processing, and distribution to meet key requirements. L1T data product quality requirements include a geolocation uncertainty of less than 12 m circular error, a band-to-band co-registration uncertainty of less than 4.5 m for the OLI spectral bands, a co-registration uncertainty of less than 24 m for the two TIRS spectral bands, and a co-registration uncertainty of less than 30 m between the OLI and TIRS spectral bands where all of these uncertainties are at a 90% confidence level. Ground system performance requirements include the processing of 85% of the LDCM scenes to L1T products within 48 h of data collection by the observatory, the processing of 400 scenes per day to L1T, and distributing 3500 scenes per day by the third year of the mission. Current best estimates indicate that the ground system will be capable of substantially exceeding these performance requirements with a predicted data processing latency of 12 h for 85% of the data, the capacity to process 890 scenes per day, and the ability to distribute 4700 scenes per day.

Excerpted from Remote Sensing of Environment 122, James R. Irons, John L. Dwyer, and Julia A. Barsi , The next Landsat satellite: The Landsat Data Continuity Mission, 11-21, Copyright 2012, doi:10.1016/j.rse.2011.08.026, with permission from Elsevier

Courtesy of the journal Remote Sensing of the Environment