Remote Sensing and Artificial Intelligence-Based Modeling and Prediction of Harmful Algal Blooms in Lake Pontchartrain
Semester of Graduation
Master of Science in Civil Engineering (MSCE)
Civil and Environmental Engineering
Harmful Algal Blooms (HABs) and particularly toxic cyanobacterial harmful algal blooms (CyanoHABs) have become a growing threat to the environment, economy, communities, and human and animal health. This is particularly true for Lake Pontchartrain. Forecasting the occurrence of cyanobacterial HABs in Lake Pontchartrain is a process that currently does not exist, as nowcasting by the NCCOS Algal Bloom Monitoring System or the similar system implemented by the U.S. EPA is currently the only method to monitor HAB production in the Lake. This thesis made this process possible by identifying antecedent environmental conditions controlling CyanoHABs, describing the conditions using NASA satellite remote sensing data from the MODIS-Aqua, and finally simulating the conditions and associated CyanoHABs by developing eight forecasting models with the lead-time of 15-22 days for predicting NCCOS Clycano index value representing the level of CyanoHABs. Specifically, eight Random Forest models were created by using the WEKA platform and two years of time series data from 2021 – 2022 for NCCOS Clycano index values and corresponding satellite remote sensing data for chlorophyll-a concentration, sea surface temperature and reflectance bands. Additionally, the models were validated with data from 2019. Model forecasting results based on the training data indicate that all eight models are capable of forecasting Clycano index values with a very high correlation coefficient of 0.90 or higher and MAE of 10 and RMSE of 20 or lower. The theoretical significance of using the eight forecasting models is that the negative impact of cloud cover on the availability of remote sensing data can be minimized, greatly expanding the application of satellite remote sensing data. The practical significance of the eight forecasting models is that they make it possible to forecast CyanoHABs on a daily basis and thereby inform water quality programs of where and when CyanoHABs are likely to occur so that managers can proactively respond to CyanoHAB events, greatly reducing the CyanoHAB risk to the public health.
Smith, Ian, "Remote Sensing and Artificial Intelligence-Based Modeling and Prediction of Harmful Algal Blooms in Lake Pontchartrain" (2023). LSU Master's Theses. 5793.