Predicting harmful algal blooms through a combined modelling approach

Abstract

The increasing intensity and frequency of harmful cyanobacterial blooms due to climate change and anthropogenically induced nutrient input is a well-studied and documented issue. Within the Netherlands, a protocol is in place to guarantee the safety of bathers based in part on cyano associated chlorophyll-a concentrations, measured in situ. Weather effects such as wind speed, precipitation or irradiance affect cyanobacteria presence. To increase the efficacy of the protocol measures and efficiency of the measurements a forecast through a combination of mechanistic models is proposed in this research. The combination of models should be able to translate weather forecasts into accurate and precise results of cyanobacteria presence. The combined models are ECMWF forecasts, the physical lake model General Lake Model (GLM) and the ecological model PCLake+. 50 different forecasts of ECMWF are used as input for GLM, which together provide input on water temperature, evaporation, wind speed, irradiation and precipitation for PCLake+. The important output variables of the models together will be a forecast on cyano associated chlorophyll-a and total chlorophyll-a concentration. Historical data as input for the models resulted in accurate model results for GLM. PCLake+ modelled a total chlorophyll-a concentration accurately but on cyano associated chlorophyll-a the accuracy decreased markedly. The effect of weather forecasts translates well to the total chlorophyll-a variance. Uncertainties concerning weather events present in early forecasts but removed in later forecasts resulted in a decrease in variance in the output. Future improvements on the combined models should focus on modelling the correct dominant group of algae. A suggested improvement to the model is an inclusion of a weather dependent groundwater model. Once the model is improved, the cyanobacteria forecasts can help improve the efficacy of Dutch water management practices.