Assessment of physically-based and data-driven models to predict microbialwater quality in open channels
Minyoung Kim , Charles P. Gerba , Christopher Y. Choi
DOI:
Received October 17, 2009,Revised January 08, 2010, Accepted , Available online
Volume 22,2010,Pages 851-857
- Summary
- References
- Related Articles
In the present study, a physically-based hydraulic modeling tool and a data-driven approach using artificial neural networks (ANNs)
were evaluated for their ability to simulate the fate and transport of microorganisms in a water system. To produce reliable data, a pipe
network was constructed and a series of experiments using a fecal coliform indicator (Escherichia coli 15597) was conducted. For
the physically-based model, morphological (pipe size, link length, slope, etc.) and hydraulic (flow rate) conditions were used as input
variables, and for ANNs, water quality parameters (conductivity, pH, and turbidity) were used. Both approaches accurately described
the fate and transport of microorganisms (physically-based model: correlation coe cient (R) in the range of 0.914 – 0.977 and ANNs:
R in the range of 0.949 – 0.980), with the exception of one case at a low flow rate (q = 31.56 cm3/sec). This study also indicated
that these approaches could be complementarily utilized to assess the vulnerability of water facilities and to establish emergency plans
based on hypothetical scenarios.
