25.01.2014 Views

Statistical Analysis of Trends in the Red River Over a 45 Year Period

Statistical Analysis of Trends in the Red River Over a 45 Year Period

Statistical Analysis of Trends in the Red River Over a 45 Year Period

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

concentrations <strong>of</strong> major anions and cations, have similar means; however, <strong>the</strong> south<br />

floodway at St. Norbert station exhibits a high degree <strong>of</strong> variability. A comparison <strong>of</strong><br />

specific conductance shows virtually identical spatial patterns. Specific conductance<br />

<strong>in</strong>creases slightly from <strong>the</strong> Emerson to Selkirk water monitor<strong>in</strong>g stations. Specific<br />

conductance measures <strong>the</strong> amount <strong>of</strong> dissolved ions <strong>in</strong> <strong>the</strong> water, when <strong>the</strong>re is an<br />

<strong>in</strong>crease <strong>of</strong> base flow relative to run <strong>of</strong>f, <strong>the</strong> specific conductance <strong>in</strong>creases. Specific<br />

conductivity is lowest <strong>in</strong> <strong>the</strong> spr<strong>in</strong>g season when <strong>the</strong> snow melts and measurements<br />

were taken on <strong>the</strong> field samples as opposed to laboratory samples. The summary<br />

statistics for constituents are shown <strong>in</strong> Table 3.2.<br />

When test<strong>in</strong>g for seasonality; <strong>the</strong> existence <strong>of</strong> seasonal patterns <strong>in</strong> water chemistry<br />

was analyzed us<strong>in</strong>g <strong>the</strong> non-parametric Kruskal-Wallis test (Conover, 1999).<br />

The null hypo<strong>the</strong>sis for this test was that <strong>the</strong> populations for each season have <strong>the</strong><br />

same median; versus <strong>the</strong> alternative hypo<strong>the</strong>sis that not all medians are <strong>the</strong> same.<br />

In order to test for trends <strong>in</strong> water quality parameters, <strong>the</strong> Mann-Kendall test for<br />

trend and Sen’s slope (Hirsch, et al.,1982) was implemented to help evaluate <strong>the</strong><br />

correlation <strong>of</strong> selected constituent concentrations with time. This test does not<br />

depend on <strong>the</strong> assumption <strong>of</strong> a particular parametric form for <strong>the</strong> underly<strong>in</strong>g distribution<br />

and hence is a “non-parametric” method. WQSTAT PLUS v.1.56, (NIC<br />

Environmental Division developed with assistance from Colorado State University<br />

faculty), is <strong>the</strong> program used for <strong>the</strong> non-parametric methods and <strong>the</strong>re are certa<strong>in</strong><br />

data requirements for this program:<br />

1. The application <strong>of</strong> <strong>the</strong> Kruskal-Wallis test for seasonality requires a m<strong>in</strong>imum<br />

sample size <strong>of</strong> four data po<strong>in</strong>ts <strong>in</strong> each “hydrologic season”<br />

2. For <strong>the</strong> trend analysis statistics, (Sen’s Slope and <strong>the</strong> Mann-Kendall Test)if<br />

<strong>the</strong>re are fewer than 41 data po<strong>in</strong>ts an exact test procedure is performed<br />

3. If 41 or more data po<strong>in</strong>ts are available, <strong>the</strong> normal approximation test is used<br />

by this program (equivalently a Chi-Square test)<br />

4. If <strong>the</strong> Seasonal Mann-Kendall test is required, that test requires a m<strong>in</strong>imum<br />

sample size <strong>of</strong> four data po<strong>in</strong>ts <strong>in</strong> each “hydrologic season”<br />

17

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!