Length, Weight, and Yield in Channel Catfish, Lake Diane, MI

TL;DR

This study analyzes the length-weight relationships and fillet yield of channel catfish in Lake Diane, Michigan, providing new models with lower uncertainty and confirming the fishery's health and suitability for management.

Contribution

The paper introduces an improved length-weight model for channel catfish that reduces parameter uncertainty compared to traditional models.

Findings

Catfish in Lake Diane are generally above standard weight.

The improved model yields more precise parameter estimates.

Fillet yield averages around 28.1% with high correlation.

Abstract

Background: Channel catfish (Ictalurus punctatus) are important to both commercial aquaculture and recreational fisheries. Little published data is available on length-weight relationships of channel catfish in Michigan. Though there is no record of public or private stocking, channel catfish appeared in Lake Diane between 1984 and 1995 and it has developed into an excellent fishery. Results: NLLS regression yields parameter estimates of b = 3.2293 and a = 0.00522. The improved model yields the same estimate for the exponent, b, and a length estimate (parameter L1) of 45.23 cm. Estimates of uncertainty and covariance are smaller for the improved model, but the correlation coefficient is r = 0.995 in both cases. LLS regression produced different parameter values, a = 0.01356 and b = 2.9726, and a smaller correlation coefficient, r = 0.980. On average, catfish in the sample weighed 106.0% of the standard weight, (Brown et al.) and the linear regression (no slope) of fillet yield vs. total weight suggests a typical fillet yield of 28.1% with r = 0.989. Conclusion: Most of the fish in the sample were above the standard weight, heavier than the 75th percentile for their length. Channel catfish are doing well in Lake Diane and the population is well matched to the food supply. Management should attempt to maintain current population levels. In this case, the improved length-weight model, W(L) = (L/L1)^b, provided lower uncertainties in parameter estimates and smaller covariance than the traditional model.