1. Benthic invertebrates were surveyed in 213 headwater streams in southern Ontario to test the utility of Percent Model Affinity (PMA) for assessing the relative impact of agriculture on invertebrate communities. Qualitative samples were collected from reference (= forested, sixty-nine samples) and agricultural (195 samples) streams between May 1990 and August 1993. The percentage composition of samples from reference streams was averaged by season and type of substratum, and the most homogeneous groupings were used as expected communities for the calculation of PMA. 2. The lower confidence limit about the mean PMA for each expected community was calculated as for a one-tailed t-test comparing a single observation with the mean of a sample. Samples from agricultural streams whose PMA scores were outside these confidence limits were judged to indicate significant impact. 3. The sensitivity of PMA to differences in land use increased with taxonomic resolution: 40% of agricultural sites were significantly different from reference communities at the ordinal level, 79.5% were significantly different at the lowest practical taxonomic level. Most of the increase in discrimination between reference and agricultural sites was due to Chironomidae. Identification to species, rather than genus, yielded only a small increase in discrimination between reference and agricultural sites, largely because of taxonomic difficulties and because most members of several diverse genera were very similar in their distributions. 4. All of the samples from streams which drained orchards and vinyards (n = 7) were significantly different from the expected communities, as were 89% of those from streams draining corn (n = 94). A larger percentage of streams draining hay (78%, 26) were impacted than were streams bordered by pasture (66%, n = 51), but the average magnitude of impact was greater in pasture streams. Only six of eleven samples from streams draining tobacco indicated significant impact. 5. The effect of land use adjacent to the sampling site was greatest early in the growing season. Comparison of natural channels and ditches draining agricultural land suggests that channelization and subsurface tile drainage eliminate most of the benefits of riparian buffer strips and magnify the effects of farming practices on the benthic fauna.
