Assistant Professor FAMU-FSU College of Engineering
Harmful Algal Blooms (HAB), colored sea surfaces brought on by massive plankton blooms, are widespread in offshore and coastal waters along a large portion of the world’s coastline. HABs frequently result in widespread fish and shellfish deaths as well as considerable losses for the farming and travel sectors in several nations. Our case study location, Biscayne Bay, which borders Miami-Dade County, is already exhibiting signs of stress because of anthropogenic factors like rapid development and expanding population. This study attempts to figure out how various environmental stressors and chlorophyll-a, a HAB representative, interact with each other. Water quality data, climatic data, and upstream land use are chosen as features from 1997 to 2020. All monthly data are linearly interpolated to daily to better comprehend the quantitative nexus for autoregressive characteristics of chlorophyll-a. Neural network, XGBoost, and SVM are utilized for establishing the time-lead prediction, 7-day, 14-day, 21-day, and 28-day, between chlorophyll-a and features. The results show that the neural network performs the fast and the best with 14-day lead prediction than other algorithms. Among the features, 14-day lead chlorophyll-a, developed area percent, and total phosphorus are important features for predicting HAB. We offer a paradigm and opportunity for data-driven methodologies to quantitatively predict HAB and prospective early warning systems essential for advancing a very hazy scientific understanding of HAB-affecting factors in coastal waters.
