A study of waterborne diseases during flooding using Radarsat-2 imagery and a back propagation neural network algorithm

Flood disasters are closely associated with an increased risk of infection, particularly from waterborne diseases. Most studies of waterborne diseases have relied on the direct determination of pathogens in contaminated water to assess disease risk. In contrast, this study aims to use an indirect assessment that employs a back propagation neural network (BPNN) for modelling diarrheal outbreaks using data from remote sensing and dissolved-oxygen (DO) measurements to reduce cost and time. Our study area is in Ayutthaya province, which was very severely affected by the catastrophic 2011 Thailand flood. BPNN was used to model the relationships among the parameters of the flood and the water quality and the risk of people becoming infected. Radarsat-2 scenes were utilized to estimate flood area and duration, while the flood water quality was derived from the interpolation of DO samples. The risk-ratio function was applied to the diarrheal morbidity to define the level of outbreak detection and the outbreak periods. Tests of the BPNN prediction model produced high prediction accuracy of diarrheal-outbreak risk with low prediction error and a high degree of correlation. With the promising accuracy of our approach, decision-makers can plan rapid and comprehensively preventive measures and countermeasures in advance.