Improved turbidity estimates in complex inland waters using combined NIR–SWIR atmospheric correction approach for Landsat 8 OLI data

ABSTRACT

Turbidity is one of the important water quality parameters, essentially a proxy to assess eutrophication state in inland coastal systems. In this article, a method of combined near-infrared–shortwave infrared (NIR–SWIR) atmospheric correction for Landsat 8 (L8) Operational Land Imager data is proposed to improve the turbidity retrieval in optically complex waters. From the extremely turbid to moderately turbid waters, the relative ranges in water-leaving reflectance in band 3 () are found to be 19–92% and 31–79% in band 4 (). The SWIR reflectances in and are 57% and 66% higher than that of standard NIR correction in extremely turbid waters. However, this method has resulted in ~30% higher reflectances than the NIR method in relatively less turbid waters; the latter method is still good in moderately turbid waters. Using Rayleigh corrected reflectances, a turbidity index, , was computed to discriminate the productive and/or turbid waters. The SWIR method was applied for water having T_ind > 1.5 threshold and the NIR method in the other regions. A new turbidity algorithm has been developed using L8 two band ratio () optimized with in situ turbidity data from four data buoys for 2014. The Landsat 8 band-weighted in situ reflectances for bands 3 and 4 are used to derive turbidity using the present algorithm and validated against in situ turbidity, providing a good coefficient of determination of R² = 0.87. As compared to the NIR-based correction, the turbidity obtained from the combined (NIR + SWIR) correction in extremely turbid waters is around 80–90% (absolute percentage difference (APD)) different. Whereas in the moderately turbid waters, the APD between the two corrections was around 50–75%. There are no obvious data discontinuities in using the combined approach. Comparisons were made with available single-band turbidity algorithms and found that the present turbidity algorithm performed well in the optically complex lagoon environment.

Acknowledgements

The authors express their sincere gratitude to Shri Tapan Misra, Director, SAC, for his encouragement. The authors are thankful to Dr Raj Kumar, Deputy Director, EPSA, and Shri Arun Kumar Sharma, Head, GSD, for providing overall guidance and keen interest. They are also thankful to Anurag Gupta, Moosa Ali, and CDA for assisting during fieldwork. USGS is gratefully thanked for providing Landsat 8 data and ACOLITE team for providing the software to process the data.

Disclosure statement

No potential conflict of interest was reported by the authors.