Author(s): Il Won Seo, Soo Yeon Choi
Linked Author(s): Soo Yeon Choi
Keywords: Contact recreation, fecal coliform, Nakdong River, logistic regression model, hierarchical clustering method
Abstract: The level of fecal contamination in rivers is one of major concerns when it comes to safety of contact recreation in rivers, since fecal contamination is related to presence of pathogen in rivers. Therefore there is a need to predict the level of fecal indicator bacteria in real-time through fast and simple method. In this study, the level of fecal coliform among various fecal indicator bacteria was predicted in real-time using logistic regression model which is one of data-driven models. Using logistic regression model, it was is possible to predict directly whether the concentration of fecal coliform exceeded exceeds 200 CFU/100ml, which is the safe water quality criteria for contact recreation in terms of fecal contamination. Nakdong River was selected as a study site, since Nakdong River is becoming popular for recreational activity facilities thanks to Four Major River Restoration Project from 2010 to 2012. In this study, monitoring stations were classified into four groups, G1, G2, G3 and G4, regarding according to their locations and hierarchical clustering results in order to reduce heterogeneity in datasets. Logistic regression models were developed using datasets from monitoring stations in three groups, G2, G3, and G4. G1 was excluded because exceedance of fecal coliform safe criteria was a rare event in G1 section. Developed models were tested using test datasets. As results, three models showed relatively high accuracy of more than 80% in terms of total classification, specificity and sensitivity
Year: 2017