of 1.5 mm. Thus, we excluded meiofaunal species; but note that meiofauna
are not commonly seen in samples from these streams (Thompson &
Townsend,2005). Macroinvertebrates were measured (to the nearest 0.25 mm),
identified to the best level of taxonomic resolution possible (Winterbourn &
Gregson,1989; Thompson & Townsend,1999) and ascribed to a functional
feeding group.Physicochemical conditions
Several ecologically relevant aspects of habitat were measured (Table5.1). Algal
production was estimated using a radioactive carbon isotope (7 mL of (^14 C)
NaHCO 3 (185Mbq ml^1 )) using stream substrata placed in portable chambers
(Fuller & Bucher, 1991 ; Thompson & Townsend,1999). Means and standard
deviations of substratum grain size were calculated for 100 random substratum
particles at each site. For ten of the sites, detailed information was also available
on disturbance regime (Townsendet al., 1997; these are ten of the sites in
Fig. 5.1). It is important to note, however, that for the sites considered here
there had been no major disturbance events in the 90 days preceding their
sampling.Statistical analyses
Size frequency histograms were generated for visual comparison of size distri-
butions, and overlain with mean, upper quartile and lower quartile values for
each stream. For statistical analysis the data for each site were converted to the
proportion of the total number of macroinvertebrates that occurred in 1 mm
size bins. Seasonal data at two sites were analyzed separately: size frequency
histograms were constructed and visually compared.
To test for effects of the various predictors on body size, mean, lower and
upper quartiles were calculated for each site for all macroinvertebrates com-
bined and for each functional feeding group. Where fewer than ten individuals
occurred in a functional feeding group for a site, these data were excluded from
analysis because the real size range of the taxon was not considered to have been
adequately sampled. Analysis of variance was used to test for effects of land-use
category and fish regime on each of the body-size parameters. Means and
quartiles of body size were plotted and regressed separately against primary
production, substratum particle size and variability, and disturbance. The sum-
mer data were also used in a principal components analysis to generate an
ordination of sites and to explore multivariate influences. This ordination was
overlain with information on primary production, substratum particle size and
variability and, for ten sites, disturbance. For those sites with complete data sets,
an optimal subsets Mantel test (BIOENV in Primer 5) was used to ascertain what
combination of ecological attributes was best able to predict distribution of
individuals across size classes.86 C. R. TOWNSEND AND R. M. THOMPSON