Hybrid modeling and prediction of oyster norovirus outbreaks

Abstract

This paper presents a hybrid model for predicting oyster norovirus outbreaks by combining the Artificial Neural Networks (ANNs) and Principal Component Analysis (PCA) methods and using the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite remote-sensing data. Specifically, 10 years (2007–2016) of cloud-free MODIS Aqua data for water leaving reflectance and environmental data were extracted from the center of each oyster harvest area. Then, the PCA was utilized to compress the size of the MODIS Aqua data. An ANN model was trained using the first 4 years of the data from 2007 to 2010 and validated using the additional 6 years of independent datasets collected from 2011 to 2016. Results indicated that the hybrid PCA-ANN model was capable of reproducing the 10 years of historical oyster norovirus outbreaks along the Northern Gulf of Mexico coast with a sensitivity of 72.7% and specificity of 99.9%, respectively, demonstrating the efficacy of the hybrid model. HIGHLIGHTS A hybrid model is presented for the prediction of oyster norovirus outbreaks.; The model is based on Artificial Neural Networks and Principal Component Analysis.; The model greatly expands the spatial coverage of oyster safety monitoring programs.; The model expands the water quality monitoring frequency from 1 month to 1 day.; The model input data are satellite remote-sensing data that are freely available.;

Keywords

• ann model
• norovirus
• outbreak
• oyster
• pca
• remote sensing