Amongst the manifold human pressures impacting marine ecosystems, eutrophication and extreme weather events frequently result in microalgae coastal proliferation. Accurate observation of massive phytoplankton blooms in the coastal ocean is however challenging because phytoplankton composition and concentration can change over short time scales and because their spatial distribution typically displays small scale variability. While remote sensing can provide spatially and temporally resolved observations, the existing satellite missions present limitations in temporal, spatial, and/or spectral resolution when it comes to monitoring phytoplankton blooms in estuaries, bays, fjords, or coastal lagoons. Compared to previous missions, Sentinel-2 (S2) has been demonstrating improved capabilities for the detection of nearshore harmful algal blooms (HABs) due to its ability to synoptically observe inland and coastal waters every 5 days at a spatial resolution higher than 20 m in 10 spectral bands from the visible to short-wave infrared spectral domains. In addition, S2 provides accurate measurements of the water reflectance due to the development and implementation of atmospheric correction and deglinting algorithms specifically designed for optically complex coastal zones. Due to its enhanced observation capability, S2 therefore made it possible to detect and document several highly-concentrated phytoplankton blooms over a variety of coastal and inland waters during the last six years.
Despite such recent improvements, the majority of HAB remote sensing investigations generally focused on specific regional case studies. Therefore, only a modicum of bloom forming microalgae species has been separately documented so far. The objective of the present study is to propose a broader perspective for the remote sensing assessment of phytoplankton blooms, in order to better resolve the optical diversity of green, golden, red and brown seawater discolorations associated with the taxonomic and pigmentary diversity of phytoplankton coastal proliferation and/or accumulation. For that purpose, a database of high-biomass blooms and phytoplankton seawater discoloration events, both documented in situ and synchronous to S2 image acquisition, has been compiled from the IOC-ICES-PICES Harmful Algae Event Database (HAEDAT), from the French Phytoplankton and Phycotoxin Monitoring Network (REPHY), and from several phytoplankton events reported in the literature. Altogether, about 100 bloom records were compiled in more than 20 countries worldwide. For the selected blooms, available in situ information reported that the phytoplankton community was generally dominated by one or two microalgae species. The compiled S2 HABs database covered 25 different bloom forming phytoplankton species belonging mainly to the dinoflagellates, ciliates and cyanobacteria groups. The spectral characteristics of the remote-sensing reflectance (Rrs) were then extracted for each dominant species, and analyzed in order to evaluate S2 ability to distinguish phytoplankton optical types in the case of highly concentrated and quasi-monospecific coastal blooms.
Whatever the species, all Rrs spectra displayed a typical red-edge pattern with a valley at 665 nm and a peak at 705 nm, resulting from the interplay between the optical properties of pure seawater and chlorophyll-a at high concentration. Optical discrimination of the dominant bloom forming species could therefore not be based on the sole red-edge pattern. The variety of bloom optical types was better resolved when investigating Rrs spectral shape variability over all S2 spectral bands from 443 to 900 nm. Preliminary results from commonly used normalization and classification methods suggest that at least six bloom types could be distinctly identified using S2, including blooms dominated by Trichodesmium (a phycoerythryn-rich cyanobacteria), Noctiluca scintillans (a dinoflagellate causing orange seawater discoloration), and Mesodinium rubrum (a ciliate responsible of burgundy red tides). These species present typical pigment profiles and/or physiological characteristics that give them distinctive Rrs features. For the cyanobacteria other than Trichodesmium, blooms dominated by Nodularia spumigena and/or Aphanizomenon flosaquae displayed a Rrs spectrum whose shape significantly differs than those of blooms dominated by the Microcystis and/or Anabaenopsis genera. For the dinoflagellates other than Noctiluca scintillans, the Rrs variability was high at both intra- and interspecific levels, therefore generating a higher risk of confusion. However, blooms dominated by species such as Lingulodinium polyedrum or Prorocentrum sp. displayed a different Rrs shape than blooms dominated by Karenia sp. Blooms dominated by Lepidodinium chlorophorum or Margalefidinium polykrikoides could be confused with non-dinoflagellates blooms dominated by the haptophyte Phaeocystis globosa.
While the present compilation is inherently limited to the available concomitant in situ and satellite HAB datasets, it aims at shedding light on the visible tip of coastal eutrophication amenable to multispectral satellite detection. Improving the capability of remote sensing to detect massive red tides and phytoplankton seawater discoloration is indeed crucial because coastal algal proliferation is likely to increase in both frequency and amplitude in the next decades due to escalating nutrient runoffs from agriculture and nearshore urban development.
To our knowledge, this is the first time that a Rrs library of massive coastal blooms has been compiled from the S2 archive. The main reflectance types could guide a first guess identification of the dominant bloom-forming species in the absence of in situ information. Such spectral library also interestingly complements studies based on laboratory phytoplankton cultures, where accurate measurements of the inherent optical properties are performed at species level. It is however challenging to measure Rrs in the lab, as well as to extrapolate laboratory results to the scale of real phytoplankton events. By providing the bulk signature of typical bloom types that also contains the second order variability associated with non-algal colored constituents, the compiled S2 Rrs library could benefit to the development of optical models for coastal waters. From a satellite mission perspective, the current limitations of our S2-based reflectance classification are also useful to highlight which bloom forming species present a risk of confusion, and for which further work is required using either an enhanced spectral resolution and/or advanced clustering algorithms.