Statistical Methods for Characterizing Ground‐Water Quality

Abstract

The benefits from ground‐water quality monitoring ultimately depend on the statistical methods used to analyze data. The methods must match both the information expectations of users and the characteristics of the water quality variables to which they are applied. The primary objective of regulatory ground‐water monitoring is detecting changes in quality. To select appropriate statistical tests for change, one must know whether the water quality variables of concern are seasonal, normally distributed, and serially dependent. This paper provides guidance in analyzing limited background data sets to determine these three characteristics. Recommended procedures to detect seasonality were pcriodograms, Student's t‐test, Mann‐Whitney test, analysis of variance, and Kruskal‐Wallis test. To test for normality, the skewness coefficient is recommended. To detect serial dependence, sample autocorrelation coefficients may be tested for significance.

Spatial and temporal changes in the structure of groundwater nitrate concentration time series (1935–1999) as demonstrated by autoregressive modelling

“…This may explain why short-term positive autocorrelation is evident in the nitrate residual time series for Cheddar, Gurney Slade and Sherborne (which has no known swallet contributions), with a month that has higher than average nitrate concentrations followed in general by one or two further months with higher than average concentrations. The strong influence of the previous months' values is typical of water quality processes (Harris et al, 1987), and has been noted in surface water nitrate concentration records in the United Kingdom (Worrall and Burt, 1999) and groundwater records in the United States (Lui, 1995). In such cases storage of nitrate is sufficient to maintain high concentrations for a sustained period of 1-2 months leaching during the winter months.…”

EFFECT OF SERIAL CORRECTION ON GROUND WATER QUALITY SAMPLING FREQUENCY1

“…These results show that the Cl's decrease markedly with increasing values of n and r. These Cl's can be used to give upper and lower estimates for sampling frequencies. Harris, et al (1987) calculated the power to detect serial correlation given a certain sample size and a Markov process. However, they used a normal approximation to the distribution which is good for large sample sizes but is less satisfactory with small sample sizes and large values of serial correlation.…”

Distribution of selected nitrogen compounds in ground water and the unsaturated zone, Superior-Hardy Special Protection Area, Nuckolls County, Nebraska, 1991-93

“…It may take about 2 years and more than 9 years for nitrate to reach ground water in the lowland and the upland, respectively. Montgomery, R.H., Loftis, J.C., and Harris, J., 1987 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 6.8 7.7 7.1 7.1 7.4 6.9 7.0 9.2 9.2 10.9 9.0 9.3 9.6 9.0 9.6 10.4 11.1 6.8 6.9 6.4 6.7 6.1 5.4 8.9 8.7 9.8 8.5…”