Doctor of Philosophy (PhD)
Civil and Environmental Engineering
Oyster norovirus outbreaks pose increasing risks to human health and seafood industry worldwide. This study presents an Artificial Intelligence (AI)-based approach to identifying the primary cause of oyster norovirus outbreaks, nowcasting and forecasting the growing risk of oyster norovirus outbreaks in coastal waters. AI models were developed using Artificial Neural Networks (ANNs) and Genetic Programming (GP) methods and time series of epidemiological and environmental data. Input variable selection techniques, including Random Forests (RF) and Forwards Binary Logistic Regression (FBLR), were used to identify the significant model input variables among six independent environmental predictors including water temperature, solar radiation, gage height, salinity, wind, and rainfall and various combinations of the variables with different time lags. In terms of nowcasting, a risk-based GP model was developed to nowcast daily risks of oyster norovirus outbreaks along the Northern Gulf of Mexico coast, showing the true positive and negative rates of 78.53% and 88.82%, respectively. In terms of forecasting, an ANN model, called ANN-2Day, was presented. The forecasting model was capable of reproducing all historical oyster norovirus outbreaks with the true positive and negative rates of 100.00% and 99.84%, respectively. The sensitivity analysis results of the ANN-2Day model further indicated that oyster norovirus outbreaks were generally linked to the extreme combination of antecedent environmental conditions characterized by low water temperature, low solar radiation, low gage height, low salinity, strong wind, and heavy precipitation. In addition to the GP and ANN-2Day models, a remote sensing–based model was constructed using MODIS Aqua level 2 products. The remote sensing-based model enabled oyster management authorities to expand the prediction of norovirus outbreak risks from areas where monitoring data were accessible to other oyster harvest areas where monitoring stations are not available. In conclusion, the developed AI models enables public health agencies and oyster harvesters to better plan for management interventions and thus makes it possible to achieve a paradigm shift of their daily management and operation from primarily reacting to epidemic incidents of norovirus infection after they have occurred to eliminating (or at least reducing) the risk of costly incidents.
Shamkhali Chenar, Shima, "Development of Artificial Intelligence Approach to Nowcasting and Forecasting Oyster Norovirus Outbreaks along the U.S. Gulf Coast" (2017). LSU Doctoral Dissertations. 4159.
Available for download on Friday, November 09, 2018