Acoustic methods are used to estimate the density of pelagic fish in large lakes with results of midwater trawling used to assign species composition. Apportionment in lakes having mixed species can be challenging because only a small fraction of the water sampled acoustically is sampled with trawl gear. Here we describe a new method where single echo detections (SEDs) are assigned to species based on classification tree models developed from catch data that separate species based on fish size and the spatial habitats they occupy. During the summer of 2011, we conducted a spatially-balanced lake-wide acoustic and midwater trawl survey of Lake Superior. A total of 51 sites in four bathymetric depth strata (0-30. m, 30-100. m, 100-200. m, and >200. m) were sampled. We developed classification tree models for each stratum and found fish length was the most important variable for separating species. To apply these trees to the acoustic data, we needed to identify a target strength to length (TS-to-L) relationship appropriate for all abundant Lake Superior pelagic species. We tested performance of 7 general (i.e., multi-species) relationships derived from three published studies. The best-performing relationship was identified by comparing predicted and observed catch compositions using a second independent Lake Superior data set. Once identified, the relationship was used to predict lengths of SEDs from the lake-wide survey, and the classification tree models were used to assign each SED to a species. Exotic rainbow smelt (Osmerus mordax) were the most common species at bathymetric depths <100. m with their population estimated at 755 million (3.4. kt). Kiyi (Coregonus kiyi) were the most abundant species at depths >100. m (384 million; 6.0. kt). Cisco (Coregonus artedi) were widely distributed over all strata with their population estimated at 182 million (44. kt). The apportionment method we describe should be transferable to other large lakes provided fish are not tightly aggregated, and an appropriate TS-to-L relationship for abundant pelagic fish species can be determined.
- Acoustic estimates of fish density
- Apportionment of acoustic density estimates to species
- Classification trees
- Lake Superior
- Laurentian Great Lakes
- Midwater trawl catches