Every summer someone asks me to predict how bad the blue-green algae conditions will be. I hem and haw and point out important variables like temperature and rainfall that produces nutrient loading. It seemed the early summer’s multiple 90^o F days and extensive spring rains with associated nutrient loading should produce ideal conditions for blue-green algae blooms early in the season, but that’s not what LCC’s volunteer blue-green algae monitors saw. July was relatively bloom free and the algae held off until mid-August. The truth is we do not fully know what initiates bloom conditions.

For an algae bloom to commence, a number of factors must occur. The predators of algae should be few enough in number to allow the bloom to take off. Nutrients, particularly phosphorus and nitrogen, must be available and in a form that the algae can use. The availability of nutrients is in turn influenced by the temperature and oxygen content of the water. Under certain conditions, for example low oxygen, nutrients stored in sediments transform from unavailable to available forms. Weather matters. More turbulent weather disrupts blooms by mixing the algae through the water column, while warm temperatures spur growth.

Predicting whether blooms will become toxic presents a whole other challenge. Some strains of the same species of algae produce toxins while others don’t, and different strains can often be found within the same bloom. Even strains capable of producing toxins don’t always do so.

Predicting bloom formation requires tracking all these variables and determining which are actually the most important. It is a feat that would challenge the prognosticating ability of Nostradamus. However, as computers become more powerful, our ability to analyze all the data we have continues to increase.

To help determine what factors are most important in predicting toxin production in blooms, researchers have been turning to a tool called Artificial Neural Networks (ANNs). ANNs mimic nerve networks in animals. Imagine a nerve cell. A long arm connects the bulbous portion of the cell to a suite of twig like projections, each of which can connect to another nerve cell. The multiple connections between cells create a complicated network. In artificial neural networks the bulbous portion of each cell is represented by a variable: phosphorus, nitrogen, temperature, etc. Each variable connects to multiple other variables. By using some mathematical wizardry, unexpected connections between variables can be discovered.

Using this technique, two independent teams of researchers have found a relationship between the ratio of nitrogen to phosphorus and potentially toxic blue-green algae blooms. Andrea Pearce earned her PhD from the University of Vermont applying ANNs to Lake Champlain water quality data. A team of Korean researchers used ANNs to answer questions about the initiation of blue-green algae blooms on the Daechung Reservoir.

Pearce found that bloom conditions usually develop when phosphorus is abundant and nitrogen scarce. Under these conditions, blue-green algae out-compete non-toxic green algae and diatoms for nitrogen while with excess nitrogen, the greens and diatoms prosper. The blue-greens can migrate vertically in the water column, and may gather essential nitrogen at the lake bottom where biological activity in the sediments makes it available. As a complicating factor, the relationship between nitrogen and potentially toxic algae only appears after water temperatures increase above about 66^o F.

The results from Korea were similar. They reported that water temperature and total dissolved nitrogen were the major determinants of blooms, with water temperature being more important than nutrient concentrations (once nutrients were already abundant). Interestingly, conditions three weeks prior to actual blooms seemed to best predict the blooms. Perhaps those early summer storms did foreshadow the mid-August blooms that developed this year.

Our tools for understanding algae blooms continue to evolve. Researchers on Lake Erie now release weekly forecasts of blue-green algae blooms, based on satellite imagery of existing conditions and wind and wave models that predict where they may move. The forecasts may not always be accurate, but recall we still have trouble forecasting the weather and those models are much more advanced than algae models. Each tool allows us to ask new questions and better analyze the results. Yet many of the secrets of the lake will stay mysterious.