Performance of Landsat-8 and Sentinel-2 surface reflectance products for river remote sensing retrievals of chlorophyll-a and turbidity

• Published in: Remote sensing of environment Vol. 224; pp. 104 - 118
• Main Authors: Kuhn, Catherine, de Matos Valerio, Aline, Ward, Nick, Loken, Luke, Sawakuchi, Henrique Oliveira, Kampel, Milton, Richey, Jeffrey, Stadler, Philipp, Crawford, John, Striegl, Rob, Vermote, Eric, Pahlevan, Nima, Butman, David
• Format: Journal Article
• Language: English
• Published: Goddard Space Flight Center Elsevier Inc 01.04.2019; Elsevier; Elsevier BV
• Subjects: Agricultural ecosystems; Agriculture; Amazon River; Atmospheric correction; Chlorophyll; Columbia River; Detection; Earth Resources And Remote Sensing; ecosystems; freshwater; industry; Inland waters; Landsat; Landsat satellites; Landsat-8; Mississippi River; Ocean color; Radiometry; Reflectance; Remote Sensing; river water; Rivers; Satellite imagery; Sentinel-2; transportation; Turbid waters; Turbidity; United States Geological Survey; Windows (intervals); Amazon River; Landsat-8; Atmospheric correction; Sentinel-2; Remote Sensing; Ocean color; Amazon River; Columbia River; Mississippi River; Turbid waters; Inland waters; Ocean Color; Atmospheric Correction
• ISSN: 0034-4257; 1879-0704; 1879-0704
• DOI: 10.1016/j.rse.2019.01.023

Rivers and other freshwater systems play a crucial role in ecosystems, industry, transportation and agriculture. Despite the >40 years of inland water observations made possible by optical remote sensing, a standardized reflectance product for inland waters is yet forthcoming. The aim of this work is to compare the standard USGS land surface reflectance product to two Landsat-8 and Sentinel-2 aquatic remote sensing reflectance products over the Amazon, Columbia and Mississippi rivers. Landsat-8 reflectance products from all three routines are then evaluated for their comparative performance in retrieving chlorophyll-a and turbidity in reference to ship-borne, underway in situ validation measurements. The land surface product shows the best agreement (4% Mean Absolute Percent Difference) with field measurements of radiometry collected on the Amazon River and generates 36% higher reflectance values in the visible bands compared to aquatic methods (ACOLITE and SeaDAS) with larger differences between land and aquatic products observed in Sentinel-2 (0.01 sr−1) compared to Landsat-8 (0.001 sr−1). Choice of atmospheric correction routine can bias Landsat-8 retrievals of chlorophyll-a and turbidity by as much as 59% and 35% respectively. Using a more restrictive time window for matching in situ and satellite imagery can reduce differences by 5–31% depending on correction technique. This work highlights the challenges of satellite retrievals over rivers and underscores the need for future optical and biogeochemical research aimed at improving our understanding of the absorbing and scattering properties of river water and their relationships to remote sensing reflectance. •Landsat-8 and Sentinel-2 images are compared over three large rivers.•Over 13,000 underway, ship-borne measurements were collected for validation.•Land surface reflectance product agreed within 4% of field data in turbid waters.•Choice of correction biases satellite-retrieved Chl-a and turbidity by 3–59%.