Roy, DP; Li, ZB; Zhang, HKK; Huang, HY (2020). A conterminous United States analysis of the impact of Landsat 5 orbit drift on the temporal consistency of Landsat 5 Thematic Mapper data. REMOTE SENSING OF ENVIRONMENT, 240, 111701.

The Landsat 5 orbit changed more than expected over its three-decade mission life. Previous modelling research indicated that the resulting Landsat 5 overpass time and so acquisition solar zenith differences will have caused non-negligible reflectance and normalized difference vegetation index (NDVI) temporal differences. Rather than use a modelling approach this paper presents the results of a comprehensive examination of the Landsat 5 Thematic Mapper (TM) data. Ten years of summer Landsat 5 TM Analysis Ready Data (ARD) sampled at >19.3 million 30 m pixel locations across the conterminous United States (CONUS) are considered. The acquisition solar zenith, red, near infrared (NIR), shortwave infrared (SWIR) (similar to 2.1 mu m) surface reflectance, and NDVI, are compared between five consecutive periods (1995 versus 2006, 1996 versus 2007, 1997 versus 2008, 1998 versus 2009, and 1999 versus 2010) that span years with very different Landsat 5 overpass times, and are also compared between 1998 and 1999 that have similar overpass times. The different years of Landsat 5 TM data are plotted against each other and reduced major axis (RMA) regressions passing though the origin are developed and compared with the acquisition solar geometry changes. The red, SWIR and NDVI RMA slopes are proportional to the magnitude of solar zenith change between years and are indicative of CONUS-wide differences due to orbit drift. The greatest differences occurred between 1995 and 2006, the years with the greatest solar zenith differences (median CONUS difference 8.43 degrees) with RMA regression slopes quite far from unity (1.14 for the red and SWIR bands, 0.91 for NDVI). In years with smaller solar zenith differences the RMA slopes are closer to unity. As agricultural crops may be managed inconsistently through time, and as drought may have pronounced effects on reflectance, the analysis is also undertaken masking out crop pixels and considering locations that have similar precipitation and evapotranspiration conditions between years. These experiments indicate that, at CONUS scale, the cause of the changing RMA regression slopes between years is not due to agricultural changes or to the influence of above normal drought or wetness. The results confirm previous published modelling based findings but are based on direct examination of CONUS Landsat 5 TM data. They indicate that certain research and applications that require temporally consistent red, SWIR and NDVI data may need to take Landsat 5 orbit drift effects into account.