Use of Resampling to Evaluate a Simple Random Sampling Design for General Monitoring of Fishes in Texas Reservoirs

Abstract

Traditional formulae used to estimate sample sizes to achieve a desired level of precision have been proven to underestimate required sample sizes. Further, because these formulae typically focus on precision, they are unable to provide the information needed to evaluate the risk of not attaining other sampling objectives. Resampling approaches provide a framework for estimating realistic sample sizes, can handle complex objectives, and will provide probability-based statements that can be used to assess risk. We used a resampling approach to assess the performance of a simple random sampling design for the general monitoring of fishes with electrofishing, gill nets, and trap nets in Texas reservoirs. We evaluated median relative standard error (RSE = 100 × SE of the estimate/estimate) of catch-per-unit-effort (CPUE) estimates of stock length fish with current sampling methods. We estimated the effort necessary to achieve RSEs of (at worst) 25% or 15% (RSE₂₅ and RSE₁₅, respectively) at various probability levels. Resampling estimates of effort needed to achieve RSE₂₅ were compared with estimates calculated from one commonly used formula (N = (CV/RSE)²) for all target species. Further, we estimated the effort necessary to collect 100 stock length fish of each target species. Median RSE ranged from 22 for largemouth bass Micropterus salmoides to 43 for crappies Pomoxis spp. The median number of stock length fish collected using the current sampling effort ranged from 16 for blue catfish Ictalurus furcatus to 43 for palmetto bass (male white bass Morone chrysops × female striped bass M. saxatilis) per sampling event. Estimated sampling effort needed to achieve RSE₂₅ for largemouth bass and to collect 100 stock length largemouth bass at the 75th percentile of samples was 29 and 62 5-min stations, respectively. Estimated sampling effort needed to achieve RSE₂₅ at the 75th percentile of samples for other target species ranged from 32 net-nights for channel catfish I. punctatus to 59 net-nights for crappies. Estimated sampling effort needed to collect 100 stock length fish from gill nets and trap nets at the 75th percentile of samples ranged from 72 net-nights for crappies to over 100 net-nights for blue catfish. Simple random sampling generally resulted in low catches of stock length fish and a poor precision of CPUE estimates, especially for target species collected with gill nets and trap nets. Sampling effort estimates (from N = (CV/RSE)²) underestimated the sampling effort needed for all species examined. Unlike traditional sample size formulae, the resampling approach presented here allows the user to determine sampling effort with an associated probability of success in achieving a particular level of precision or in collecting a given number of fish.