Publications

Xu, HQ; Ren, MJ; Yang, LJ (2023). Evaluating the consistency of surface brightness, greenness, and wetness observations between Landsat-8 OLI and Landsat-9 OLI2 through underfly images. INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION, 124, 103546.

Abstract
One important mission of the newly launched Landsat-9 is to collaborate with Landsat-8 to reduce the revisit period of Landsat Earth observations to eight days. This requires a high level of consistency between the two satellite data. Previous cross-calibrations between Landsat-8 Operational Land Imager (OLI) and Landsat-9 OLI2 have been performed via band-by-band approaches, and the spectral deviation revealed between the two sensors was within 1%. However, it remains uncertain whether this deviation will persist when multiple bands are combined to perform certain calculations and whether the offset found in each band will be amplified in the combined multiband performance. Therefore, a comprehensive cross-comparison using a multiband combination approach is necessary to ensure agreement between the two sensor data. This study conducted a multiband combination-based cross-comparison using simultaneous underfly data of both sensors along with a Tasseled Cap Transform (TCT) performance. The coefficients of Landsat-8 were utilized to calculate the three components of Landsat-9 TCT. The calculation of the TCT involves six bands of both OLI sensors, which enables a more thorough examination of the consistency of the two sensor data. In addition, calculating Landsat-9 TCT directly using Landsat-8 coefficients allows further investigation of the similarity between the two sensors by determining whether Landsat-8 TCT coefficients are suitable for Landsat-9 TCT. Both top of atmosphere (TOA) reflectance data and surface reflectance (SR) data were employed for this comparison. Apart from whole image pair-based comparisons, land cover category-based comparisons were also performed. The results show that the three TCT components of Landsat-9, calculated using Landsat-8 coefficients, are similar to those of Landsat-8, with an average R2 of 0.983 and RMSEs on the order of 0.009 in most scenarios. The deviations observed in the TCT components between the two sensors are primarily due to the higher radiometric resolution of Landsat-9 (14 bits) compared to Landsat-8 (12 bits), as the deviations occurring in bright and dark areas are larger than those in other areas. Besides, the uncertainty in the green band for vegetative targets and the uncertainty revealed in typical vegetative surfaces also contributed to the higher divergence in the greenness component. The 1% difference detected in band-by-band cross-calibrations increased by 0.2-0.5 percentage points in this comprehensive multiband comparison. Overall, this cross-comparison study generates confidence that the Landsat-8 OLI and Landsat-9 OLI2 data are in strong agreement even when performing multiband combination operations. This demonstrates that the synergistic use of the two sensor data can well maintain the continuity of Landsat Earth observations.

DOI:
10.1016/j.jag.2023.103546

ISSN:
1872-826X