y f- 



96 



than one dependent variable because of the risk of error caused by 

 unassessed changes in the covariable. 



None of the multiple regression methods produced a useful 

 model for predicting emergent biomass. All stepwise multiple 

 regression models for emergent biomass were subject to random 

 error. The reason that diatom taxa related so poorly to emergent 

 vegetation might be that the stems of emergent vegetation do not 

 provide the large surface areas for periphyton attachment that 

 submerged and floating-leaved macrophytes provide. Surface area 

 has been shown to be a primary factor influencing epiphytic biomass 

 (Cattaneo and Kalff 1980). Plants with many small or finely- 

 dissected leaves and more lateral growth patterns such as Hydrilla, 

 Valisneria and Myriophyllum would, therefore, demonstrate clearer 

 relationships with periphytic biomass than would emergent 

 macrophytes having vertical growth forms. Hoyer and Canfield 

 (1986), in addition, have shown that the mean surface area to 

 biomass ratio of submerged plants is higher than for emergent 

 plants, which permits a greater biomass of periphyton to be 

 supported on submerged plants than on an equal weight of emergent 

 plants. 



The best model for predicting percent-volume infestation was 

 equation 3.7 that was derived by canonical correspondence analysis 

 and explained 60% of the variance in that macrophyte variable. 

 Equation 3.2, a mutlivariate equation that was based on three diatom 

 taxa, also explained 52% of the variance in percent volume 

 infestation. Percent-area coverage is best predicted with equation 

 3.4 that was based on 4 diatom taxa and explained approximately 



