Water samples collected every 8 d from three sites on the Welsh River Dee over 29 months were examined chemically and morphologically and their heterotrophic activities assessed. Bacterial population estimates were regressed on 21 independent environmental variables using multiple linear regression analysis. Prediction equations calculated from 1975–1976 data accounted for 70 of the total variation from an upland site but only 40 for a lowland site. The validity of the prediction equations was examined by correlating the predicted and observed count data of 1977–1978. The correlations of plate count values were significant for the upper (r= 0.64,P<0.001) and middle reach (r= 0.62,P<0.01) sites but not significant for the lowland site. The direct count correlation was significant for all three sites. The permanganate value, a measure of organic matter, was the most dominant variable in the prediction equations and was shown by principal component analysis (PCA) to be linked to particulate matter and river flow. Prediction equations constructed from the PCA data accounted for less variation than the original equations but, in general, there was an agreement between the two methods of multivariate analysis. The logarithm of the heterotrophic potential (Vmax) from the lowland site was shown to be a linear function of temperature (r= 0.7,P<0.001). A specific activity index calculated for the lowland site data gave a median value of 8.4 times 10‐10μg/h/bacterial cell and was highly sensitive to temperature change (r= 0.74,P<0.001) whereas the median value for the upland site was 2.6 times 10‐10μg/h/bacterial cell and did not respond significantly to temperature change (r= 0.4
展开▼
