WSP12246 - Laserfiche WebLink

Home Browse Search

2AO"~ tested because that period was when all of the salinity- control work was done in the Grand Valley and lower Gunnison River Basin (through September 1993). Waler years 1986-93 also were tested for trends because Stage II S(llinity work began in the Grand Valley in 1986 and involved more extensive salinity- control work than Stage I, and salinity-control work began in the Lower Gunnison Basin Unit in 1988. The greatest effect of the saJinity-conlrol projects on dissolved solids in the Colorado River should have occurred after 1986. Several variables that represenl various meas- ures of salinity were tested for trends. The seasonal Kendall test and the ESTREND program were used for trend analysis on periodic dissolved-solids, calcium, magnesium, sodium, and sulfate concentrations and on monthly dissolved-solids loads. Linear-regression analysis was used for the trend analysis of annual and seasonal dissolved-solids loads and of the annual Grand Valley dissolved-solids load. The trend results on unadjusted data and flow-adjusted data for water years 1970-93 were compared to LOWESS smooth curves for the same period. If statistically significant trends of decreasing concentrations or loads were reported, then the LOWESS smooth curves might aid in delineation of when the !rends had occurred, either before or after initiation of the salinity-control projects. The trend results for 1980-93 and 1986-93 were com- pared to the results for] 970-93 and to the LOWESS curves in attempts to identify trends in concentrations or loads that could be related to the salinity-control work. All trend results are presented, and then the rela- lion of the !rend results to salinity-control projects are discussed at the end of this section of the report. Trend results are listed in tables that include the trend-slope magnitude, trend slope as a percent rate of change, the p value of the test, and the significance level of the slope. Hypolhesis testing by statistical methods requires the selection of an alpha level, which also is referred to as the significance level of the test, for making the decision to reject or not to reject thc null hypothesis of no trend. The alpha level is the probabil- ity of incorrectly rejecting the null hypothesis whcn, in fact, the null hypothesis is true. The alpha level also is called the type I error (Helsel and Hirsch, 1992). In terms of trend analysis, the alpha level is the probabil- ity of reporting a trend when, in fact, there is not a trend. Thus, the alpha level is the risk level that an investigator is willing to accept for making a type I error. For example, an alpha level of 0.05 (5 percent) implies that in 95 percent of the cases, the test will cor- rectly indicate no trend when there, actually is no trend. Selection of a very small alpha level would minimize a type I error; however, that selection increases the chance of committing a type II error, which is failing to reject the null hypothesis when, in fact, it is false (or reporting no trend when there actually might be a trend). An alpha level commonly used in hypothesis testing is 0.05. The null hypothesis then is evaluated by comparing the p value from the statistical test of the data to 0.05. When the p value is less than 0.05, the null hypothesis is rejected; if the p value is greater than 0.05, then the null hypothesis is not rejected. As described in many statistical texts, such as Helsel and Hirsch (1992), not rejecting the null hypothesis is not the same as saying the hypothesis is actually proven. All that can be said is that, based on the data available, the null hypothesis cannot be rejected. Instead of selecting only one alpha level for describing the signif- icance of the trends, a range of alpha levels was used for comparison to the p values by using a scheme sim- ilar to one used by Liebermann and others (1988). The monotonic Ircnd resulls wcre considered highly signif- icant (HS) if the p value was less than or equal to 0.01; significant (S) if the p value exceeded 0.01 and was less than or equal to 0.05; marginally significant (MS) if the p value exceeded 0.05 and was less than or equal to 0.10; and the trend was not significanl (NS) if the p value exceeded 0.10. Trends in Dissolved-Solids Concentrations Periodic dissolved-solids concentrations repre- sent discrete water samples collected at varyi ng fre- quencies at gaging stations 09095500, 09152500, and 09163500 during water years 1970-93. Monotonic trend results for periodic dissolved-solids concentra- tions and for flow-adjusted concentrations for the three stations for 1970-93, 1980-93, and 1986-93 are summarized in table 2. Most trends in unadjusted dissolved-solids concentrations were not significant (table 2). The trend results indicate increasing dissolved-solids concentrations for 1980-93 and 1986-93 at all three stations, although only the trend for 1980-93 for station 09095500 and ihe trend for 1986-93 for station 09152500 were significant. Increasing concentrations probably reflect the effect of streamflow during 1980-93, when above-average 14 Trend Analysis of Selected Water~auality Data Associated With Sallnlty~Control Projects In the Grand Valley. In the lower Gunnison River Basin, and at Meeker Dome, Western Colorado