Comparing geometric differences between Landsat Collection 1 to Collection 2 level-1 products

Abstract

In late 2020 the U.S. Geological Survey (USGS) began the distribution of Landsat products associated with their collection 2 reprocessing of the archive. Several changes were implemented within the Landsat Product Generation System (LPGS) and the calibration parameters applied to the Landsat imagery for the collection 2 processing. When comparing between collection 1 and collection 2 products, radiometric and geometric differences will be present. One of the most substantial changes between the two collections was an adjustment to the ground control which made the control more accurate from an absolute and relative perspective. Some of the other changes associated with collection 2 processing were to the Digital Elevation Model (DEM) used in terrain correction, the Thermal Infrared Sensor (TIRS) relative gains, TIRS absolute calibration, Operational Land Imagery (OLI) absolute gain model, OLI relative gain, and OLI bias calculation to name a few. Although these changes also have an effect on the differences between the product associated with these two collections, the change in the ground control, although typically less than one 30-meter multispectral pixel in magnitude, will have the largest effect on the differences between the products. This change in ground control is also a spatially dynamic change that although is low in spatial frequency, is nonlinear and is not a change that can be modeled on a global or even on a local Worldwide Reference System-2 (WRS-2) path and row scale. The effects of these ground control changes will be discussed and demonstrated within this paper along with examples showing their effect on specific datasets. This paper demonstrates some of these geometric differences associated with the ground control through both the registration statistics created during product generation and through an example comparison of a set of collection 1 and collection 2 products.