Session: ASME Undergraduate Student Design Expo

Paper Number: 173404

Quantifying Lagged Rainfall and Temperature Signals as Triggers for Harmful Algal Blooms on the Florida Gulf Coast 

Harmful algal blooms (HABs) caused by the dinoflagellate Karenia brevis (K. brevis) are persistent ecological and public health hazards in the Gulf of Mexico. Despite decades of observational data, the timing and triggers of bloom development remain poorly understood. Changing precipitation and warming temperature add further uncertainties into the problem. To support system-level modeling and forecasting efforts in coastal environmental engineering, this research develops a data-driven framework that integrates time series and correlation analyses, detection of lagged hydrometeorological-bloom relationships, and supervised machine learning algorithms to identify and quantify the delayed impacts of short-term rainfall and temperature dynamics on harmful algal bloom formation, and to evaluate the predictive utility of hydrometeorological and water quality variables. We also evaluate the role of hurricanes on triggering K. brevis blooms.

This research analyzes K. brevis cell count records (1953–2023) in conjunction with daily precipitation and temperature data along the Florida Gulf Coast. Preliminary results show that bloom severity is strongly related to antecedent precipitation, followed by maximum air temperature. Cross-correlation analyses around Hurricane Michael (2018) identified a ~27-day lag in rainfall most strongly associated with the bloom intensity (Pearson’s r = 0.63, p-value < 1e-7), particularly for short 3–5 day cumulative rainfall windows. Maximum air temperature over 28–30 day periods, lagged by ~30 days, also correlated significantly with bloom counts (Pearson’s r ≈ 0.46, Kendall’s τ ≈ 0.47), suggesting a compounding thermal contribution.

Seasonal analyses of wet seasons of the region impacted by Hurricane Michael during June through October of 2021–2023 showed that total monthly rainfall, when lagged by two months, was strongly correlated with monthly K. brevis concentrations (Pearson’s r = 0.57, p-value = 0.003). This finding is consistent with a delayed nutrient runoff and bloom growth mechanism that unfolds over weeks to months.

To evaluate the predictive utility of these lag-based relationships, we trained a series of simplified random forest classifiers using individual precipitation features at key lag days identified through the cross-correlation analyses with Hurricane Michael, alongside in-situ hydrometeorological and physiochemical water quality variables such as water temperature, pH, and dissolved oxygen. Models using long antecedent (around 30 days) precipitation achieved high classification accuracy (>96%) and full recall for high-bloom events. These results suggest that short-term, lag-specific rainfall signals carry predictive value and can enable an efficient and interpretable framework to predict blooms.

Future research will focus on systematically classifying K. brevis blooms into discrete severity levels using established cell count thresholds across multiple Florida Gulf Coast counties. For each severe bloom event, I will examine precipitation patterns at lag intervals from 0 to 30 days prior, using corresponding measurements from the nearest weather stations. This approach will enable spatially and temporally resolved correlation analyses between localized rainfall and bloom severity, deepening understanding of bloom drivers in diverse coastal settings. I will also continue to incorporate machine learning techniques to enhance predictive evaluation and improve classification accuracy. Overall, this framework offers valuable insights for coastal environmental management and for the design of early-warning systems in marine ecosystems.

Presenting Author: Vicky Gao Belmont University

Presenting Author Biography: Vicky Gao is a junior studying Economics and Environmental Science at Belmont University. She is currently conducting NSF-funded research on harmful algal blooms and flood modeling using machine learning within the Resilient Infrastructure and Disaster Response Center (RIDER), a joint initiative of FSU and FAMU. Additionally, she collaborates with two professors on an ongoing independent research project investigating visitors’ willingness to contribute time and money to urban parks by modality. Vicky is passionate about interdisciplinary environmental research that combines economic valuation, data analytics, and ecological science to address complex sustainability challenges. Her research interests include urban water quality, especially as it intersects with the rising demand from artificial intelligence data centers. She hopes to pursue a graduate study in Natural Resource Economics, with a focus on applying quantitative methods to emerging environmental issues.

Authors:

Vicky Gao Belmont University
Sumon Hossain Rabby Florida State University and Florida A&M Joint College of Engineering Resilient Infrastructure and Disaster Response Center
Xiuming Sun Florida State University and Florida A&M Joint College of Engineering Resilient Infrastructure and Disaster Response Center
Ebrahim Ahmadisharaf Department of Civil and Environmental Engineering and Resilient Infrastructure & Disaster Response Center at Florida State University (FSU)