Retrieving water chlorophyll-a concentration in inland waters from Sentinel-2 imagery: Review of operability, performance and ways forward

Abstract

The fundamental role of water for life and the threats to water bodies around the world have highlighted the need for their conservation. Remote sensing is a tool that allows us to monitor water bodies in a rapid, systematic, accurate and economical way, being complementary to traditional field sampling methods. The main aim of this review is to synthesise the use of the Sentinel-2 satellite for chlorophyll-a monitoring, an indicator of the trophic state of aquatic ecosystems, and assess the role of each parameter on chlorophyll-a retrieval. To this end, indices, models, atmospheric corrections and field sampling details used so far in chlorophyll-a monitoring of aquatic ecosystems using Sentinel-2 imagery were analysed. Sentinel-2 was chosen because it has suitable features for monitoring water bodies (spatial, temporal and spectral resolution), despite not having been specifically designed for that purpose. The indices aphy(B4)/a*phy(B4), B7(1/B4-1/B5), B5-(B6 + B4)/2 and B3/B4 performed best in lakes and B2 + B3 + B4 + B5, B3/B6 and (B5-B4)/(B5 + B4) in reservoirs. The atmospheric correction ELM performed worse than Sen2Cor and ATCOR in lakes. In reservoirs, ATCOR performed best and C2XC and Dark Object Subtraction performed worse. Finally, classical machine learning and deep learning models outperformed traditional linear and non-linear models. An integrated vision of remote sensing with Ecology could improve some weaknesses found in the reviewed articles, such as the lack of methodological details in field sampling or knowledge of the dynamics and functioning of the ecosystem to achieve the most optimal sampling of the system. By doing so the field of remote sensing would have a higher aplicability. Some further investigations are needed on small water bodies (area 90% of the water bodies worldwide.

Keywords

• Remote sensing
• Inland aquatic ecosystems
• Water quality
• Atmospheric correction
• Spectral indices