Algal Bloom Prediction
GitHub Repository: CLA-Project
Algal Blooms pose a health threat to Madison's beachgoers and wildlife. Since the lakes of Madison are the lowest elevated points in the Yahara watershed, they are contaminated by the runoff from farmland and other phosphorus-containing sources. There has been plenty of research at UW-Madison to understand the behavior of these algal blooms, so this research project is one of many efforts on campus to create a predictive model for algal blooms on Madison's lakes. Some researchers attempt to create theory-based models based on ecological data, some researchers (such as myself) attempt to create a model purely using data science, and some researchers are developing a hybrid of these two model types. This project utilizes data gathered from volunteers at the Clean Lakes Alliance (CLA), weather data from the Daily Summaries at NOAA's website, and NTL-LTER Data gathered from a buoy sitting in Lake Mendota. Using this data, I have been developing a predicitve model for algal blooms on the largest lakes in the Yahara watershed: Mendota, Monona, Waubesa, Kegonsa, and Wingra. I have tried various machine learning models for this project in an effort to select the best one for making predicitions. Such models have included random forests, kernel SVMs (including linear, sigmoid, RBF, and polynomial), k-nearest neighbors, and neural networks.