Statistical analysis of lead, zinc, copper and cadmium concentrations in the sediment of the Jordan River, Utah

In the winter of 1996 the General Chemistry (Chemistry 121-1) course at the University of Utah included a service-learning project. The students attempted to determine the lead concentration at various sites along the Jordan River. The Jordan River is the major waterway in the Salt Lake Valley, and the investigated locations were selected to represent interesting features of the river.; On a series of field trips during a 2-week period, underwater sediment samples were collected by the students using a simple core sampler provided by the U.S. Geological Survey. In the laboratory, sample solutions were prepared from portions of these sediments taken by “wet grab” followed by acid digestion according to a simplified protocol based on Method 3050B promulgated by the Environmental Protection Agency. The test solutions were analyzed by inductively coupled plasma atomic emission spectroscopy in the Environmental Laboratory of Kennecott Utah Copper Company, providing not only the lead concentrations needed, but zinc, copper and cadmium concentrations as well.; The students' work was successful. In spite of the limitations and the shortcuts taken, the determined average metal concentrations as a set compare very reasonably to previously published results.; However, inspection of the data revealed an unexpectedly large number of large concentration values, outliers. The abundance of the “wild” concentrations necessitated a statistical evaluation of the analytical results.; The examination of the data uncovered that the results of the usual and customary parametric statistical treatments are false. This failure is due to the invalidity of the parametric model applied. To remedy the situation, robust methods were invoked. The performance of the robust estimates also proved to be severely impaired by the extent of deviation from the model implied.; Nonparametric statistics were used to circumvent the lack of an appropriate model. The metal concentrations in the sediment proved to be quite variable along the river and even between samples collected in close proximity of each other.; The data set in this investigation was large enough to draw attention to some interesting characteristics of the sampled system. It was not sufficiently extensive to resolve all the problems encountered.