Evaluation of ACOLITE atmospheric correction methods for Landsat-8 and Sentinel-2 in the Río de la Plata turbid coastal waters

ABSTRACT

Landsat-8 (L8) and Sentinel-2 (S2) terrestrial satellite missions have shown to contain useful information for aquatic applications. However, quantitative retrieval of water quality parameters, such as turbidity, strongly depend on the performance of atmospheric correction algorithms. Among available processors, ACOLITE (https://odnature.naturalsciences.be/remsem/software-and-data/acolite) is simple to incorporate in imagery processing routines and has shown to have better performance than other processors for sediment-rich waters. Recently (in 2018), it incorporated a new default atmospheric correction approach, the dark spectrum fit (DSF), which remains to be tested in most of the Southern hemisphere coastal areas. In this work, we present new in-situ radiometric measurements collected in the northern coast of the Río de la Plata estuary, South America, during field campaigns along a 2-year period. The data set was used to evaluate the performance of ACOLITE’s DSF and exponential extrapolation (EXP) methods with L8 and S2 imagery, and to recalibrate a turbidity algorithm. The DSF did not perform very well, giving particularly poor results in the near infrared (NIR) bands. However, its performance was greatly improved with an optional sun glint correction (DSF+GC), although some positive bias was still present in the NIR bands. A GC seemed to be most important in dates with higher sun elevation (austral spring and summer), and should be strongly considered for other water bodies in the region and in similar or lower latitudes (35°S). Additionally, the EXP method gave good results in the green-NIR spectral region when a low (5th) percentile aerosol type was selected. Finally, the effect of the atmospheric correction on turbidity retrieval from satellite imagery was assessed: if the red and a NIR band were combined, the effect of the bias in the NIR region was negligible for the DSF+GC method; however, some impact was noticed for the lowest turbidity levels if a single NIR band was used.

Acknowledgements

This manuscript was possible thanks to the support of the Sistema Nacional de Investigadores (National Researchers System, Uruguay), the Project FMV-1-2017-1-136098 from Agencia Nacional de Investigación e Innovación (National Research and Innovation Agency, Uruguay), the fund for scientific equipment from Comisión Sectorial de Investigación Científica (Scientific Research Committee) of the Universidad de la República, Uruguay, and the fellowship for graduate studies granted to F. Maciel by the Comisión Acadmémica de Posgrado (Academic Graduate Committee) of the Universidad de la República, Uruguay. The authors also want to thank the assistance of Lucía Ponce de León, Matías González, and Rodrigo Mosquera during field measurements. Part of the data used in this work was collected during field campaigns in the frame of an outreach project for UTE (National Electric Administration of Uruguay).

Disclosure statement

Nopotential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the Comisión Sectorial de Investigación Científica []; Agencia Nacional de Investigación e Innovación, Uruguay [FMV-1-2017-1-136098].