Sensors for Monitoring Harmful Algae, Cyanobacteria and Their Toxins

20 found in food, or similar. Further, the community should emphasize the importance of “connecting the dots” for businesses by demonstrating how this new technology will make them more competitive and profitable by averting/overcoming impacts related to HAB events. Breakout Session E – What are the approaches to testing of HAB technologies? • Q1: What level of verification testing is needed? ( i.e. what is “good enough” in the context of price versus performance for different data uses ) It is important that any verification testing be designed to match a well-articulated purpose – ‘Fit for Purpose’ and ‘Intended Use’ . For example, verification testing of technologies used for HAB forecasting support will be different from verification testing to support compliance with drinking water standards. In another example, PCR approaches may be best suited for “early warning” intended use, but once a bloom develops and the organism is known, the analytical task shifts to toxin detection rather than species identification. This can further be complicated when blooms contain more than one harmful species, particularly in marine systems. Also, verification testing may need to be system-specific. For example, HAB strains may vary lake to lake, so an assay validated (or verified) in one lake system may not be applicable in another lake system. This led to a discussion about the concept that “bloom” is a term that describes an intrinsically heterogeneous system that complicates the idea of verification testing. The testing framework can be challenged by a number of key biological factors, including independent variability in toxicity with respect to cell densities, unknown triggers of toxicity, timing of sample collection, and inherent influences from biodiversity, to name a few. There can also be platform functionality concerns for verification testing. One group discussed this using the ESP and IFCB technologies as examples. Questions arose such as: Are we testing cell counts versus speciation? Are the images themselves accurate? How well is the phytoplankton community being represented – chained organisms and those with higher mobility can be challenging for imaging systems; sampling intakes are working within the confines of inherently patchy surroundings; volumes sampled are often best guess estimates to not over/underwhelm detection capabilities. Are the processing algorithms (human and automated) accurate? Technologies provide surrogate measurements that require building a model to the parameter of interest. In the ideal scenario, these models could be built between remotely sensed or in situ fluorescence measurements with toxin concentration. In reality, that’s not yet feasible in large part due to the uncertainty and variability that exists in all data inputs assimilated by a given model (e.g. remotely sensed ocean color, temperature, in situ fluorescence data vs cell counts or pigment concentrations vs. toxin concentrations). Therefore, the realistic best-case scenario is probably to develop models to estimate the abundance of known toxin producers. Verification testing was discussed as data for building and ground-truthing a model. For remote sensing, verification data should be collected across a range of water body types and environmental conditions. Target parameters are typically cell counts, abundance, or extracted pigment concentrations, and samples should ideally be collected weekly within 1 km and 8 hrs of the satellite data collection. For in situ sensors, weekly measurements for extracted pigment concentrations or cell counts are probably ideal given the variability