Fuzzy modeling of cyanobacterial surface waterblooms: validation with NOAA-AVHRR satellite images

We combined a “traditional” dynamic simulation model based upon differential equations with fuzzy logic to describe the three main conditions governing surface waterbloom formation:

1. a preexisting population of cyanobacteria
2. buoyancy of the cells
3. stability of the water column

The attributes and membership functions of the fuzzy model were based on earlier field studies of diel changes in buoyancy and vertical distribution of cyanobacteria. The model was applied without further calibration to the large lake IJsselmeer (1200 km²) in the Netherlands, and we validated the model output using 12 years of NOAA-AVHRR (National Oceanic and Atmospheric Administration–Advanced Very High Resolution Radiometers) satellite images on which surface blooms are discernible as an enhanced vegetation index or increased surface water temperature. Existing surface blooms were predicted with high accuracy, but additional blooms were also predicted. A statistical test (Cohen's Kappa) showed that correct predictions of the absence or presence of surface blooms were highly unlikely to have occurred by chance only. The model can be used to predict the occurrence of surface waterblooms in advance on the basis of the long-term weather forecast, leaving time for appropriate management of the problem. The lake management has expressed interest in converting the present model into a fully operational–online–early warning system.