Water-Quality Trends for Streams and Reservoirs in the Research Triangle Area of North Carolina, 1983-95
Trend Analysis Methods
Trend analyses were used to determine whether the concentrations of a water-quality constituent had consistently increased or decreased over a particular time period. Because water-quality constituents vary due to factors such as season and streamflow amount, the seasonal Kendall trend test (Hirsch and others, 1982) was selected for most tests. The seasonal Kendall trend test is a non-parametric test that accounts for seasonal variations in concentrations by comparing only data from the same season. Because it is a non-parametric test, it does not require assumptions about normality and constant variance. Data are grouped according to the "season (2)" in which the sample was collected irrespective of the year. Within each "season," data are ordered according to year; comparisons are made between data-pair concentrations at year = t and year = t+1. An increasing trend exists when significantly more data pairs increase than decrease; a decreasing trend exists when significantly more data pairs decrease than increase; and if pairs decrease and increase at the same frequency, no trend exists.
Many constituent concentrations in streams are strongly related to the amount of streamflow. Typically, the concentrations of constituents associated with sediments or in suspended form increase as streamflow increases, whereas concentrations of constituents in the dissolved form decrease as streamflow increases. A particularly wet or dry year or period of years can mask an underlying trend. Stream data were adjusted for the effects of streamflow by using regression analysis of log-transformed streamflow and water-quality data (Helsel and Hirsch, 1992). If the streamflow-to-constituent concentration was not significant at a = 0.05, LOWESS--a non-parametric smoothing technique (Helsel and Hirsch, 1992)--was used. Data for sites having more than 10 years of record were adjusted for serial correlation. All trend analyses were conducted by using a seasonal Kendall-tau test accounting for monthly data (12 "seasons" per year). A Wilcoxon rank-sum test (Mann-Whitney test) was applied to total phosphorus data to determine differences between pre- and post-phosphate-detergent-ban concentrations.
2--The number of seasons is established by the user according to the available data and hydrologic factors. For the analysis presented herein, each month was considered a "season" because monthly data were available.
Servicing the USGS streamflow gage on the Eno River at Hillsborough.
Terry Middleton and John Taylor collecting a representative suspended-sediment sample with a depth integrating sampler.
Last modified: Wed Jun 10 10:42:33 1998
Return to Water-Quality Trends for Streams and Reservoirs in the Research Triangle Area of North Carolina, 1983-95