1. Title, Journal and Authors
Title : Development of Nowcasting System using Machine Learning Approaches to Predict Fecal Contamination Levels at Recreational Beaches in Korea
Authors : Yongeun Parka, Minjeong Kimb, Kyung Hwa Chob, Yakov Pachepskyc ,Seoung-Hwa Choid, Jeong-Goo Chod, Junho Jeone*
aSchool of Civil and Environmental Engineering, Konkuk Univ, Seoul 05029, Republic of Korea
bSchool of Urban and Environmental Engineering, Ulsan National Institute of Science and Technology, Ulsan 44919, Republic of Korea
cUSDA-ARS, Enviromental Microbial and Food Safety Lab, 10300 Baktunire Ave, Building 173, BARC-EAST, Beltsville, MD 20705
dBusan Metropolitan City Public Health and Enviroment Research Institute, Busan, 46616, Republic of Korea
eDep. of Environmental Engineering and Graduate School of FEED of Eco-Friendly Offshore Structure Changwon National Univ., Changwon, Gyeongsangnam-do 51140,
Repulic of Korea.
*Corresponding author (khcho@unist.ac.kr)
2. Summary
Microbial contamination and microbial water quality management in watersheds are critical issues in protecting populations from microbial risks. For this reason, it is important for human to avoide the contact with microbiologically contaminated water. To reach this goal, the research team create the forcast model to predict the microbial water quality in beach water.
This prediction model use a 2 different algorithm ANN(Articifical Neural Network) and SVR(Support Vector Regression) which is composed of three layers(the input layer, the hidden layer, and the output layer). This model maily focus on preannouncing the level of fecal contamination(enterococcus[ENT], Escherichia coli[E.coli]) based on an enviromental variable such as tidal level, wair and water temperature, solar radiation, wind direction and velocity, percipitation, discharge from the wastewater treatement plant. The input variables are statiscally evaluated. To compare the output with the true value, they use the fecal contamination data at Haeundae Beach(HA) and Gwangalli Beach(GW).
The perception, wind direction, and discharge volume from the wastewater treatment plant had positive relationships with two bacteria in both HA and GW, whereas solar raditiona had a negative one(p < 0.01). But there's no reliable relationship between other variables and fecal contamination.
Both ANN and SVR demonstrate the accuracy in predicting ENT and E.coli concentrations at HA however they had low prediction accuracy at GW. So research team compare the accuracy at GW of ANN with SVR, and the performance of the trained ANN is more correct than that of the trained SVR model(p < 0.05)
3. Contact
Sungryul Kim / Intern student
Environmental Systems Engineering Lab.
Gwangju Institute of Science and Technology
Phone : +82-10-2788-2169
E-mail : fuf1994@naver.com