Harmful algal blooms (HABs) are considered a major environmental problem and threat to the aquaculture industry in many European countries. Harmful algal species produce toxins that can be concentrated by filter-feeding shellfish and cause amnesia or paralysis when ingested. In high concentration HAB species may cause massive fin-fish suffocation and mortality through clogging the gill tissues. Harmful blooms result in economic losses through closure of shellfish grounds and direct damage of fish farms.
If the bloom is high in biomass, harmful algae can be observed from space using satellite sensors.The method in place for HAB detection has been developed by the Plymouth Marine Laboratory and is fully described in Kurekin et al. (2014). This method uses the specific characteristics of water leaving radiances and derived quantities of particular HAB species for automatic detection of HAB events. As it uses a fully automatic approach by allowing the classifier to recognize and train itself to the specific characteristics of a HAB, its performance depends on the choice and the accuracy of training data (the values of water reflectance, absorption and backscatter for different wavelengths). In this regard, laboratory measurements of HAB species under controlled conditions can provide a suitable training dataset.
Within AQUA-USERS, a simulated data set has been constructed based on measurements of cultures of harmful algal species successfully grown in laboratory conditions. For each species, real-time observations of backscattering, absorption and attenuation were obtained from the cultures. Water was collected after the optical measurements were completed for the ancillary measurements: pigment analysis (chlorophyll-a), microscopy, particulate organic carbon and particulate absorption on filter. Based on the experiment measurements the total absorption and backscattering properties were calculated for HAB species for different chlorophyll-a concentrations and results used to forward model the water reflectance, using the methodology proposed by Shang et al. (2014). Reflectance spectra derived were then used to train the HAB classifier.
The preliminary results from training the HAB classifier with simulated reflectance data of the species Karenia mikimotoi in MERIS images revealed promising results with improvement in false alarm rate for HAB classification when compared to results obtained using the classifier trained with a satellite derived data set.
The early warning of HABs using this method allows the aquaculture industry to take protective measures to minimise stock losses and reduce economical loses. Within the AQUA-USERS project we are extending this capability across several European countries.
Kurekin A, Miller P, Woerd J, (2014) Satellite discrimination of Karenia mikimotoi and Phaeocystis harmful algal blooms in European coastal waters: merged classification of ocean colour data. Harmful Algae, 31, 163-176.