An adaptive approach to detect high-biomass algal blooms from EO chlorophyll-a data in support of harmful algal bloom monitoring

摘要

High-biomass harmful algal blooms can kill farmed fish through toxicity, physical effects or de-oxygenation of the water column. These blooms often form over spatially large areas meaning that Earth observation is well placed to monitor and study them. In this letter, we present a statistical-based background subtraction technique that has been modified to detect high-biomass algal blooms. The method builds upon previous work and uses a statistical framework to combine spatial and temporal information to produce maps of bloom extent. Its statistical nature allows the approach to characterize the region of interest meaning that region-specific tuning is not needed. The accuracy of the approach has been evaluated using Moderate Resolution Imaging Spectroradiometer (MODIS) data and an in situ cell concentration dataset, resulting in a correct classification rate of 68.0% with a false alarm rate of 0.24 (n = 25). The method is then used to study the surface coverage of a large high-biomass harmful algal bloom of Karenia mikimotoi. The approach shows promise for the early warning of spatially large high-biomass algal blooms, providing valuable information to support in situ sampling campaigns.