The samples were clustered using a computer using a program written by 

 the DNRC and a rather awesome forty-six foot dendrogram was constructed by 

 hand. In order to aid interpretation, the dominant species and their cover 

 classes were entered on one axis along with plot numbers. The dendrogram 

 showed both discrete types and areas of gradation. 



This dendrogram is too large to be included in this publication. It 

 may be found in DNRC files in Helena. 



Interpretation 



Sokol and Sneath (1963) recognize the advantage of a program that, 

 for given data, yields an unequivocal similarity scheme, but this should 

 not be confused with a thoroughly objective process. Anyone using the same 

 data and techniques in this study, could produce a dendrogram identical to 

 the one used for classification. It would be possible to pick one or more 

 similarity levels and conclude that there dire as many types as there are 

 stems intercepted by a line cutting across that similarity level. The argu- 

 ment for doing so would be to maintain the objectivity of the data analysis. 



Due to the number of choices necessary for sampling and data analysis, 

 however, the supposed objectivity of the data analysis is not there to main- 

 tain. (Appendix C contains a discussion of this topic.) And so it seems 

 that the major advantage of the multivariate analysis is that choices and 

 assumptions are closely presented, and the data are graphically portrayed. 



In practice, the dendrogram guided the making of a preliminary class- 

 ification, and was used as a reference for any revisions. In comparing 

 clusters, the significance of changes in coverage or dominance had to be 

 considered along with the possibility of associated changes in site, pro- 

 ductivity, structure, or diversity. No single similarity level is appro- 

 priate for classification; rather, the similarity levels were most meaning- 

 ful in the context of the clusters being considered. Vegetation authorities 

 at Montana State University and the University of Montana reviewed the 

 preliminary classification and eventually a classification for the grass- 

 land types was chosen. 



Samples with shrub layer presented a special problem because the under- 

 story was often similar to other types, but the addition of a shrub layer 

 seemed to be important not only from a phytosociological standpoint, but 

 also from the aspect of animal ecologists and managers who might use the 

 vegetation classes. For example, the Symphoricarpos occidental is - Rosa 

 arkansana community type is very similar floristically to the Shepherdia 

 argentea/Symphoricarpos occidentalis - Rosa arkansana community type but 

 for the addition of Shepherdia argentea . They may share over one hundred 

 percent (absolute coverage) of their species in common. A similar sit- 

 uation occurs when Cornus stolonifera , Fraxirius pennsylvanica , or Populus 

 deltoides are added. These species represent important additions to the 

 lower strata of plants. Daubenmire (1963) emphasizes the importance of 

 taking synusiae (or layers) into account. 



14 



