We propose to develop bio-optical algorithms and products using hyperspectral data under program element 2.4 to generate an improved application for the detection of freshwater cyanobacteria that are often toxic, belonging to a larger family of algal conditions referred to as harmful algal blooms (HABs). We propose to develop a new detection application for freshwater cyanobacteria HABs (CHABs) by integrating information from several different approaches, some of which expand upon current operational algorithms but which are not based on hyperspectral data. We will produce a suite of advanced bio-optical products, including from semi-analytic algorithms (SAAs). We believe that hyperspectral remote sensing will improve the ability of water quality algorithms to quantify CHABs. For the SAAs, we propose to develop separate products based on optical model inversion incorporating cyanobacteria properties. Cyanobacteria have unique absorption and scattering properties, which make them difficult to invert using standard SAAs models, but our approach will use these characteristics to create a unique identification of cyanobacteria populations, among other phytoplankton groups that are typically also present before, during and after CHAB events. Toxicity is often highly variable within CHABs is not directly detectable with optical methods, however, we note that CHAB surface scums, which we will detect, are often toxic.
The main application question we are addressing is: can we improve CHAB detection and assessment including early stages with hyperspectral data? We believe that a combination of approaches based on hyperspectral data will provide necessary discrimination to monitor CHAB populations, as well as other phytoplankton groups with many applications to freshwater quality. Thus, our ultimate application product will be a comprehensive CHAB assessment that includes confidence levels tied to bio-optical algorithm uncertainties. Our specific objectives are to:
Develop new adaptive CHAB indices based on present approaches but modified by dynamically selecting optimal bands per pixel depending on locations of spectral features expanding algorithm sensitivity and dynamic range.
Develop novel adaptive semi-analytic bio-optical models that use spectral libraries for inherent optical properties that themselves are associated with different algal groups and physiological conditions.
Develop a CHAB detection product application with confidence levels based on the combination of algorithm outputs and criteria established from objectives 1 and 2. Generate demonstration products from the proposed scheme using PACE analog hyperspectral in situ and remote sensing data over the case study areas.
We will develop and test this hyperspectral CHAB detection framework with a large in situ optical database collected in Lake Erie, Saginaw Bay and Green Bay three areas of the Great Lakes that experience annual CHAB events. This new framework will improve the quality of information obtained from satellites, such as that being disseminated through the HAB Tracker by NOAA's Great Lakes Environmental Research Laboratories (GLERL). The optical database contains hyperspectral measurements previously collected in the study region by the project team, and from NOAA GLERL's operational field monitoring programs that include routine water sampling, cell counts, optical profiles and radiometry. We will apply the new algorithms to hyperspectral imagery collected from aircraft and HICO, which had a dedicated Lake Erie sensing mission. The available data over the proposed study period are not costed to this proposal. We believe this is one of the most complete bio-optical data sets available in the world for a freshwater lake system, collected in an area with recurring CHABs. This site and data set is very compelling for developing and testing hyperspectral algorithms for CHABs, and we believe this proposed work is very responsive to element 2.4.