24 

 variables accounted for 71% of variation in number of trout/km. The habitat 

 variables in that model were water velocity, late summer flow/mean flow, 

 solid overhead cover, max summer water temperature, and vegetative substrate 

 ratings (Appendix III). 



For the multiple regression models involving urban or control stations only, 

 the r values were even higher and the solid-overhead-cover variable dropped out 

 (Appendix III), as it had been the most significantly differing variable between 

 the urban and control categories of stream (Table 2). Percent pool-and-turbulence 

 cover and percent eroding banks entered only into the urban model. Water velocity 

 was the first variable to enter into all three models (Appendix III) after channel 

 width was omitted. 

 Other Observations on Fish Populations 



The species of fish collected in the study sites were rainbow trout (Salmo 

 gairdneri ) , brook trout (Salvelinus fontinalis ) , brown trout (Salmo trutta) , 

 longnose sucker (Catostomus catostomus ) , white sucker (Catostomus commersoni) , 

 mountain sucker (Catostomus platyrhynchus) , and mottled sculpin (Cottus bairdi ) . 



Standing crop per unit stream area ranged from 34 kg/ha in station 5 of 

 Mathew Bird Creek to a high of 305 kg/ha in station 13 of Bozeman Creek. In 

 terms of numbers of over-20-cm trout per kilometer, abundance ranged from 30 

 to 611 in the same stations (Table 5). Grouping of the stations into longer 

 stream sections reduced the confidence intervals of trout population estimates 

 (Table 6), enabling clearer indication of relationships between different parts 

 of the stream in Bozeman Creek (Figures 4 and 5). 



In Bozeman Creek, number/km for station 36, a control station of greater- 

 than-average length and lying just below a mink farm where mink feed and manure 

 are reportedly introduced into the stream, was significantly higher than in 



