Premium
SETTING PREDICTION LIMITS FOR MERCURY CONCENTRATIONS IN FISH HAVING HIGH BIOACCUMULATION POTENTIAL
Author(s) -
BALTHIS W. L.,
VOIT E. O.,
MEABURN G. M.
Publication year - 1996
Publication title -
environmetrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.68
H-Index - 58
eISSN - 1099-095X
pISSN - 1180-4009
DOI - 10.1002/(sici)1099-095x(199607)7:4<429::aid-env225>3.0.co;2-c
Subject(s) - bioaccumulation , normality , environmental science , statistics , fish consumption , mercury (programming language) , fish <actinopterygii> , pollutant , mathematics , fishery , ecology , computer science , biology , programming language
The potential health risks associated with the consumption of contaminated fish have long been recognized by public health officials as cause for concern, and in the USA many states have developed strategies for issuing fish consumption advisories. The methods and criteria for establishing such advisories vary widely among the states, however, and the advice given to anglers may not be consistent between neighbouring states, even regarding the same body of water. Fish contaminant monitoring data are often used as the basis for adivsories, but few methods are available for the quantification and distributional characterization of contaminant levels in fish. Log‐normality of pollutant concentrations (i.e. normality of the logarithm of pollutant concentrations) is a common assumption, yet statistical tests of normality do not always confirm this assumption. An alternative to the log‐normal distribution is the S ‐distribution, which has been shown to approximate many statistical distributions with high accuracy, and often results in improved fit over the log‐normal. In this paper we evaluate the performance of the S ‐distribution in characterizing contaminant concentrations, and compare the results to those obtained using the log‐normal distribution. A method based on trends in mercury distribution parameters across length classes is presented and used to obtain 95 per‐cent prediction limits for mercury concentrations in king mackerel ( Scomberomorus cavalla ). It is shown that this method gives narrower prediction limits compared to those obtained using standard regression techniques.