Laserfiche WebLink
<br />1927 <br />METHODS OF TREND ANALYSIS <br />. <br /> <br />PREliMINARY REPrl;'~ <br />~.." .~-:-. !:-:~-r- :--.......... - "'-. _ ..... . <br />.. ...- <br /> <br />Trend analysis can be used to determine if discharge or water quality have changed over time. In [his study, <br /> <br />step-trend analysis was used to determine discharge and specific-conductance trends in the Arkansas River. With a <br /> <br />step-trend analysis, it is assumed that data collected before a specific time are from a distinctly different data popu- <br /> <br />lation than the data collected after that time. The difference between the data populations is assumed to be one of <br /> <br />location (e.g. mean or median) but not necessarily of scale (e.g. variance or interqual1ile range). The step-trend anal- <br /> <br />ysis is much more specific than other trend analyses (e.g. monoconic trend analysis). because it requires that a par- <br /> <br />ticular fact, the time of the change, is known before any examination of the data. It is imperative that the decision <br /> <br />to use step-ttend ptocedures not be based on prior examination of the data, as this would bias the significance level <br /> <br />of the test. Significance levels, as represented by the individual p-values of each test, were for the two-sided trend <br /> <br />test, because no prior determination of the direction of trends was made. A significant test was defined at a 95-per- <br /> <br />cent confidence level (pSO.05). The step-trend procedures are pal1icularly well suited to this study, because the pur- <br /> <br />.pose of the study was to determine if discharge or specific conductance at a pal1icular location on the Arkansas River <br /> <br />changed after the construction of Pueblo Reservoir (1975) or after the implementation of the John Mal1in Reservoir <br /> <br />1980 Operating Plan. <br /> <br />Step-trend analysis was conducted with the nonparametric Mann-Whitney-Wilcoxon rank-sum test (Bradley, <br /> <br />1968). This non parametric procedure was selected, because the data (discharge and specific conductance) were gen- <br /> <br />erally not normally distributed, which waS determined based on graphical data analysis. It has been thoroughly dem- <br /> <br />onstrated that non-parametric procedures have more power (or efficiency) than parametric procedures in cases <br /> <br />where there is a substantial departUre from normality (Helsel and Hirsch, 1988). <br /> <br />Qualitative comparisons of discharge and specific-conductance data were made with the aid of box plots. Box- <br /> <br />plots are useful because variability between data sets and unusually large or small values within a data set can easily <br /> <br />be seen. Boxplots were constructed to compare monthly differences in discharge and specific-conductance data at <br /> <br />pal1icular mainstem sites before and after implementation of reservoir operations. Box p~ots contain the following <br /> <br />.formation (fig 2). The horizontal line inside the box represents the median value (50 percent of the data are greater <br /> <br />7 <br />