Do different sampling designs produce differences in the metrics of curimba, prochilodus lineatus (Characiformes: Prochilodontidae)?

Abstract

Probabilistic sampling, a classical recommendation for unbiased data, has been often overlooked. Fish sampling in reservoirs, unlike most rivers, can be accomplished by a variety of probabilistic sampling designs due to few restrictions in the choice of sampling sites (SSs). We evaluated whether population metrics of Prochilodus lineatus differ between fixed SSs (do not change with each sampling trip) and variable SSs (change with each sampling trip) in the Volta Grande (VGR) and Jaguara (JR) reservoirs of the Grande River, Paraná River basin, Brazil. We used gillnets to sample fish at nine probabilistic fixed SSs and nine probabilistic variable SSs per sampling trip. We conducted 25 sampling trips in VGR and 22 in JR during 2011 to 2015, and performed 35 analyses of 8 metrics for each reservoir. We found sampling design influenced only 7% of the analyses involving three metrics, apparently caused by biased data of fixed SSs. We attribute the small percentage of analyses influenced by sampling design to the low spatial heterogeneity of the reservoirs. The choice of the most appropriate sampling design seems to depend on the type of variation (temporal or spatial) that one wishes to detect, and the availability of time and financial resources.