Comparison of chlorophyll- a derived from sentinel-2 satellites, an unmanned aerial vehicle, and in situ hyperspectral sensor

Sentinel-2 satellite imagery has been widely used for water quality retrieval applications and provides higher spatial and temporal resolution than most similar Earth observation satellites (e.g., Landsat). However, accurately retrieving water quality parameters from Sentinel-2 can be challenging in smaller lakes with high variability of topographic features, bottom reflectance, fringing submerged vegetation, and interference by optically active, non-phytoplankton particles. The aim of this study was to compare chlorophyll a (Chl-a) retrievals from Sentinel-2 satellites, a DJI Phantom 4 Multispectral (P4M) unmanned aerial vehicle (UAV), and in-situ estimates of Chl-a from four small lakes in South-East Queensland, Australia, with diverse optical signatures. In-situ sampling was conducted at four lakes concurrently with P4M flights and simple spectral band combinations such as red and green ratios and the normalised difference chlorophyll index (NDCI) were applied to the reflectance data, and calibrated to Chl-a concentrations. The Sentinel-2 NDCI models showed good agreement (R2 = 0.85–0.88) with in-situ Chl-a at three lakes, that were more eutrophic. However, the P4M red/green ratio performed better in the remaining lake, which contained high concentrations of suspended sediment. Our findings highlight the advantages of UAV-based remote sensing that include distinguishing non-water pixels and facilitating integration with satellite observations.