Journal of Water and Health In Press, Uncorrected Proof © IWA Publishing 2012 | doi:10.2166/wh.2012.142
Methods for assessing long-term mean pathogen count in drinking water and risk management implications
James D. Englehardt, Nicholas J. Ashbolt, Chad Loewenstine, Erik R. Gadzinski and Albert Y. Ayenu-Prah jr.
University of Miami, PO Box 248294, Coral Gables, FL 33124-0630, USA. E-mail: firstname.lastname@example.org
USEPA Office of Research and Development, 26 West Martin Luther King Drive, Mail Code: 593, Cincinnati, OH 45268, USA
USEPA National Center for Environmental Assessment, 26 West Martin Luther King Drive, Cincinnati, OH 45268, USA and US Department of Justice, ERF, BLDG 27958-A, Quantico, VA 22135, USA
University of Miami, Coral Gables, FL 33124-0630, USA and HJ Foundation, 1385 NW 80th St., Miami, FL 33166, USA
CDM, 1515 Poydras St., Suite 1350, New Orleans, LA 70112, USA
First received 27 August 2011; accepted in revised form 15 January 2012. Available online 8 March 2012
Recently pathogen counts in drinking and source waters were shown theoretically to have the discrete Weibull (DW) or closely related discrete growth distribution (DGD). The result was demonstrated versus nine short-term and three simulated long-term water quality datasets. These distributions are highly skewed such that available datasets seldom represent the rare but important high-count events, making estimation of the long-term mean difficult. In this work the methods, and data record length, required to assess long-term mean microbial count are evaluated by simulation of representative DW and DGD waterborne pathogen count distributions. Also, microbial count data were analyzed spectrally for correlation and cycles. In general, longer data records were required for more highly skewed distributions, conceptually associated with more highly treated water. In particular, 500–1,000 random samples were required for reliable assessment of the population mean ±10%, though 50–100 samples produced an estimate within one log (45%) below. A simple correlated first order model was shown to produce count series with 1/f signal, and such periodicity over many scales was shown in empirical microbial count data, for consideration in sampling. A tiered management strategy is recommended, including a plan for rapid response to unusual levels of routinely-monitored water quality indicators.
Keywords: correlated; discrete; monitoring; sampling; scaling; Weibull
Full Text PDF