iron, and sulfate were performed in the Soil and 

 Water Analysis Laboratory of Utah State Univer- 

 sity, Logan, Utah. 



For reasons mostly unknown, 1 3 of the installed 

 samplers ceased to function during the study 

 period. The results of the remaining 32 are re- 

 ported here. The data were screened to assure 

 that each soil-water sampler used in the analysis 

 was represented in both the 1974-75 and 1977- 

 79 sample periods. 



A lot of variation exists within the data; the data 

 may also be biased due to the differential ex- 

 change properties of the ceramic cups (Grover 

 and Lamborn 1979; Severson and Grigal 1976; 

 Hansen and Harris 1975). Probable sources of 

 data variation include variable intake rates of the 

 ceramic cups, a long extraction time (Severson 

 and Grigal 1 976), a non homogenous soil solution, 

 and the phenomenon whereby ions in a soil solu- 

 tion do not vary inversely with soil-water content 

 (Reitemeier 1 946). 



Although the data are variable, there are 491 

 observations, including ionic concentrations of 

 copper, iron, and sulfate. 



Multivariate cluster analysis operates under 

 less restrictive assumptions than some other sta- 

 tistical techniques — analysis of variance for in- 

 stance (Davis 1973). This is probably advanta- 

 geous in field problems involving complex and 

 interacting processes that are impressible to iso- 

 late and study individually. This is the case in the 

 present instance, where it is necessary to sort out, 

 a posteriori, the influence of revegetation. 



The data contain 491 samples from 32 sample 

 tubes; 21 tubes on the revegetated area and 1 1 

 tubes on the untreated (bare) area. The samplers 

 on the revegetated area average 21 ft (6.4 m) 

 deep, with a range from 4 to 61 ft (1.2 to 18.6 m). 

 The samplers on the untreated area averaged 1 6 ft 

 (4.9 m) deep, with a range from 4 to 44 ft (1.2 to 

 1 3.4 m). The data were not separated by depth.' 



Sampling dates were: 



7/25/74 6/15/77 



8/29/74 9/1/77 



3/12/75 10/19/77 



5/29/75 6/14/78 



6/25/75 10/14/78 



7/17/75 6/14/79 



^An analysis of variance test on both the treated and 

 untreated areas suggests that there is no difference in the data 

 due to depth, probability = 0.9. 



RESULTS 



On-site precipitation was measured with an 

 aluminum standpipe storage gage fitted with an 

 Alter windshield. For the water years (Oct. 1 to 

 Sept. 30) 1 975, 1 976, 1 977, and 1 978, the preci- 

 pitation amounts were 25.5 inches (648 mm), 20 

 inches (508 mm), 23.5 inches (597 mm), and 16 

 inches (406 mm), respectively. There are no long- 

 term precipitation records available for this site, 

 but the 4-year measured precipitation is probably 

 less than a 30-year average. 



Considering the basic statistics of the samples, 

 number of observations, mean, standard devia- 

 tion, and range, there does not appear to be any 

 difference in the ionic concentrations of the soil 

 water due to revegetation (table 1 ). However, both 

 the revegetated and untreated areas appear as 

 though some improvement in water quality has 

 taken place over time. At least the sample ionic 

 means show some reduction. However, the large 

 sample standard deviations increase the uncer- 

 tainty of any statistical inference purporting to 

 show differences between sample means. 



Cluster Analysis 



Hierarchial cluster analysis was performed at 

 the computer center of Utah State University, 

 Logan (Marshall and Romesburg, undated). The 

 data for each ion were standardized by subtract- 

 ing the mean from the observed value and dividing 

 the result by the standard deviation. This pro- 

 cedure ensures that each value in the data matrix 

 is weighted equally (Sneath and Sokal 1 973). The 

 resemblance coefficient was the m - space Eucli- 

 dian distance, d..,, which is a measure of dissimi- 

 larity. 



n 



m 



N' 



k=1 



1/2 



where Xy^ denotes the kth variable measured on 

 / th object. In all, m variables are measured for each 

 object and cy,y is the distance between the /th and 

 /th objects. As you would expect, a small distance 

 indicates that the objects are similar while a large 

 distance indicates dissimilarity. The m-space Euc- 

 lidian distance ranges from zero to infinity. The 

 clustering technique used was the unweighted 

 pair-group method using arithmetic averages 

 (UPGMA). 



2 



