Premium
Estimating Sampling Effort Required for Characterizing Species Richness and Site‐to‐Site Similarity in Fish Assemblage Surveys of Wadeable Illinois Streams
Author(s) -
Holtrop Ann M.,
Cao Yong,
Dolan Chad R.
Publication year - 2010
Publication title -
transactions of the american fisheries society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.696
H-Index - 86
eISSN - 1548-8659
pISSN - 0002-8487
DOI - 10.1577/t09-078.1
Subject(s) - species richness , sampling (signal processing) , jaccard index , similarity (geometry) , riffle , ecology , variance (accounting) , streams , statistics , multivariate statistics , environmental science , habitat , hydrology (agriculture) , biology , mathematics , computer science , geology , computer network , accounting , geotechnical engineering , filter (signal processing) , cluster analysis , artificial intelligence , business , image (mathematics) , computer vision
For fish assemblage surveys, estimates of sampling effort required for capturing all or most species at a site often vary greatly among sites, reflecting substantial among‐site differences in mean species detectability (MSD). Different sampling methods used in individual studies may also contribute to the variation in estimates of required sampling effort among regions. This high variability presents a great challenge to standardize sampling effort across sites or to generalize the estimates from particular studies. It is also unclear whether the sampling effort required for capturing species is sufficient for accurately estimating site‐to‐site similarity in species composition, which is critical for evaluating β‐diversity and obtaining stable resolutions from multivariate analysis. We used a recently proposed hierarchical model to estimate the total species richness (TSR est ) at each of 11 wadeable stream sites sampled with an electric seine in Illinois. Accordingly, a sampling effort of 59–144 channel widths (CWs) is required for capturing all species and an effort of 40–72 CWs is required for capturing 90% of TSR est . We then modeled the MSD–environment relationship with random forests regression; the model accounted for 52.2% of the variance in MSD estimated at one subreach but less of the variance in MSD at higher levels of sampling effort. The MSD mainly decreased with increasing watershed size and percentage of clay‐dominant substrates and increased with percentage of rock‐dominant substrates and percentage of riffle habitats. We also estimated the true values of the Jaccard coefficient (JC) and Sørensen coefficient (SC) for all pairs of sites with the hierarchical model. The same accuracy required less sampling effort to estimate JC and SC than to estimate TSR est (e.g., only 7–60 CWs were required for 90% accuracy of JC and SC). Our findings provide the possibility of setting site‐specific sampling effort based on predicted MSD and can be used as a guide for electric seine surveys of stream fish in Illinois and potentially other midwestern U.S. states.