Comparisons of techniques for ordinating and classifying old-growth floodplain forests in Southern Illinois

摘要: Comparisons are made among the responses of several ordination and classification techniques using a set 136 plot samples from diverse old-growth bottomland forest in Southern Illinois. Of evaluated (PO, PCA, GO RA), RA provided most interpretable ordering species samples. The was divided into three segments which correspond to major vegetation units within study area itself more broadly Floodplain Forest Region. Among used (MINFO, MDISP CLUSTER), MINFO produced clusters groupings similar those identified by RA. Classification that assume strict linearity or samples, is correlation coefficient standard distance similarity measures, less satisfactory terms their ability recover informative patterns data than logarithmic terms. Discriminant analysis interpret groups various environmental variables as predictors This approach successfully an interpretation group structure other techniques. woods at Horseshoe Lake appear be influenced complex flooding-aeration gradient.