Effects of different sampling effort and taxonomic resolution on assessment metrics

Abstract

I investigated the effect of different degrees of sampling effort and taxonomic resolution on macroinvertebrate assessment indicator metrics. Specifically, I examined the relationship between a typical rapid assessment approach involving a low-effort sampling (LES) and a more thorough and intensive sampling method, or a high-effort sampling (HES) across a set of watersheds with varying degrees of agricultural impact. Seven macroinvertebrate indicator metrics were significantly different (α = 0.05) between low- and high-effort samplings, but response patterns of the indicator metrics were strongly correlated between effort levels. Likewise, metrics based on different levels of taxonomic resolution were significantly correlated each other; however, based on t-tests, metric scores between family- and genus-levels were inconsistent. Normalizing regression models using landscape features were employed to assign standardized impairment status to each site. The models for indicator metrics using HES or genus level identification had higher R2 and F-ratios than did LES or family level identification. However, there was no significant difference (α= 0.05) in normalized indicator scores between levels of sampling effort or taxonomic resolution. Normalized impairment classifications were also significantly (α= 0.05) correlated among all combinations of sampling effort and taxonomic resolution. Assessments produced using very different degrees of effort yielded almost identical results when metric results were normalized. LES and family level identification is more cost- vi effective since they require less effort and time. However, the higher statistical significance and precision of HES or genus level resolution suggests that where precision is a determining factor, a higher level of sampling effort should be considered.